Merge pull request #12715 from open-webui/dev

0.6.3

.github/ISSUE_TEMPLATE/bug_report.md

@@ -1,80 +0,0 @@
----
-name: Bug report
-about: Create a report to help us improve
-title: ''
-labels: ''
-assignees: ''
----
-
-# Bug Report
-
-## Important Notes
-
-- **Before submitting a bug report**: Please check the Issues or Discussions section to see if a similar issue or feature request has already been posted. It's likely we're already tracking it! If you’re unsure, start a discussion post first. This will help us efficiently focus on improving the project.
-
-- **Collaborate respectfully**: We value a constructive attitude, so please be mindful of your communication. If negativity is part of your approach, our capacity to engage may be limited. We’re here to help if you’re open to learning and communicating positively. Remember, Open WebUI is a volunteer-driven project managed by a single maintainer and supported by contributors who also have full-time jobs. We appreciate your time and ask that you respect ours.
-
-- **Contributing**: If you encounter an issue, we highly encourage you to submit a pull request or fork the project. We actively work to prevent contributor burnout to maintain the quality and continuity of Open WebUI.
-
-- **Bug reproducibility**: If a bug cannot be reproduced with a `:main` or `:dev` Docker setup, or a pip install with Python 3.11, it may require additional help from the community. In such cases, we will move it to the "issues" Discussions section due to our limited resources. We encourage the community to assist with these issues. Remember, it’s not that the issue doesn’t exist; we need your help!
-
-Note: Please remove the notes above when submitting your post. Thank you for your understanding and support!
-
----
-
-## Installation Method
-
-[Describe the method you used to install the project, e.g., git clone, Docker, pip, etc.]
-
-## Environment
-
-- **Open WebUI Version:** [e.g., v0.3.11]
-- **Ollama (if applicable):** [e.g., v0.2.0, v0.1.32-rc1]
-
-- **Operating System:** [e.g., Windows 10, macOS Big Sur, Ubuntu 20.04]
-- **Browser (if applicable):** [e.g., Chrome 100.0, Firefox 98.0]
-
-**Confirmation:**
-
-- [ ] I have read and followed all the instructions provided in the README.md.
-- [ ] I am on the latest version of both Open WebUI and Ollama.
-- [ ] I have included the browser console logs.
-- [ ] I have included the Docker container logs.
-- [ ] I have provided the exact steps to reproduce the bug in the "Steps to Reproduce" section below.
-
-## Expected Behavior:
-
-[Describe what you expected to happen.]
-
-## Actual Behavior:
-
-[Describe what actually happened.]
-
-## Description
-
-**Bug Summary:**
-[Provide a brief but clear summary of the bug]
-
-## Reproduction Details
-
-**Steps to Reproduce:**
-[Outline the steps to reproduce the bug. Be as detailed as possible.]
-
-## Logs and Screenshots
-
-**Browser Console Logs:**
-[Include relevant browser console logs, if applicable]
-
-**Docker Container Logs:**
-[Include relevant Docker container logs, if applicable]
-
-**Screenshots/Screen Recordings (if applicable):**
-[Attach any relevant screenshots to help illustrate the issue]
-
-## Additional Information
-
-[Include any additional details that may help in understanding and reproducing the issue. This could include specific configurations, error messages, or anything else relevant to the bug.]
-
-## Note
-
-If the bug report is incomplete or does not follow the provided instructions, it may not be addressed. Please ensure that you have followed the steps outlined in the README.md and troubleshooting.md documents, and provide all necessary information for us to reproduce and address the issue. Thank you!

.github/ISSUE_TEMPLATE/bug_report.yaml

@@ -0,0 +1,146 @@
+name: Bug Report
+description: Create a detailed bug report to help us improve Open WebUI.
+title: 'issue: '
+labels: ['bug', 'triage']
+assignees: []
+body:
+  - type: markdown
+    attributes:
+      value: |
+        # Bug Report
+
+        ## Important Notes
+
+        - **Before submitting a bug report**: Please check the [Issues](https://github.com/open-webui/open-webui/issues) or [Discussions](https://github.com/open-webui/open-webui/discussions) sections to see if a similar issue has already been reported. If unsure, start a discussion first, as this helps us efficiently focus on improving the project.
+
+        - **Respectful collaboration**: Open WebUI is a volunteer-driven project with a single maintainer and contributors who also have full-time jobs. Please be constructive and respectful in your communication.
+
+        - **Contributing**: If you encounter an issue, consider submitting a pull request or forking the project. We prioritize preventing contributor burnout to maintain Open WebUI's quality.
+
+        - **Bug Reproducibility**: If a bug cannot be reproduced using a `:main` or `:dev` Docker setup or with `pip install` on Python 3.11, community assistance may be required. In such cases, we will move it to the "[Issues](https://github.com/open-webui/open-webui/discussions/categories/issues)" Discussions section. Your help is appreciated!
+
+  - type: checkboxes
+    id: issue-check
+    attributes:
+      label: Check Existing Issues
+      description: Confirm that you’ve checked for existing reports before submitting a new one.
+      options:
+        - label: I have searched the existing issues and discussions.
+          required: true
+        - label: I am using the latest version of Open WebUI.
+          required: true
+
+  - type: dropdown
+    id: installation-method
+    attributes:
+      label: Installation Method
+      description: How did you install Open WebUI?
+      options:
+        - Git Clone
+        - Pip Install
+        - Docker
+        - Other
+    validations:
+      required: true
+
+  - type: input
+    id: open-webui-version
+    attributes:
+      label: Open WebUI Version
+      description: Specify the version (e.g., v0.3.11)
+    validations:
+      required: true
+
+  - type: input
+    id: ollama-version
+    attributes:
+      label: Ollama Version (if applicable)
+      description: Specify the version (e.g., v0.2.0, or v0.1.32-rc1)
+    validations:
+      required: false
+
+  - type: input
+    id: operating-system
+    attributes:
+      label: Operating System
+      description: Specify the OS (e.g., Windows 10, macOS Sonoma, Ubuntu 22.04)
+    validations:
+      required: true
+
+  - type: input
+    id: browser
+    attributes:
+      label: Browser (if applicable)
+      description: Specify the browser/version (e.g., Chrome 100.0, Firefox 98.0)
+    validations:
+      required: false
+
+  - type: checkboxes
+    id: confirmation
+    attributes:
+      label: Confirmation
+      description: Ensure the following prerequisites have been met.
+      options:
+        - label: I have read and followed all instructions in `README.md`.
+          required: true
+        - label: I am using the latest version of **both** Open WebUI and Ollama.
+          required: true
+        - label: I have included the browser console logs.
+          required: true
+        - label: I have included the Docker container logs.
+          required: true
+        - label: I have listed steps to reproduce the bug in detail.
+          required: true
+
+  - type: textarea
+    id: expected-behavior
+    attributes:
+      label: Expected Behavior
+      description: Describe what should have happened.
+    validations:
+      required: true
+
+  - type: textarea
+    id: actual-behavior
+    attributes:
+      label: Actual Behavior
+      description: Describe what actually happened.
+    validations:
+      required: true
+
+  - type: textarea
+    id: reproduction-steps
+    attributes:
+      label: Steps to Reproduce
+      description: Providing clear, step-by-step instructions helps us reproduce and fix the issue faster. If we can't reproduce it, we can't fix it.
+      placeholder: |
+        1. Go to '...'
+        2. Click on '...'
+        3. Scroll down to '...'
+        4. See the error message '...'
+    validations:
+      required: true
+
+  - type: textarea
+    id: logs-screenshots
+    attributes:
+      label: Logs & Screenshots
+      description: Include relevant logs, errors, or screenshots to help diagnose the issue.
+      placeholder: 'Attach logs from the browser console, Docker logs, or error messages.'
+    validations:
+      required: true
+
+  - type: textarea
+    id: additional-info
+    attributes:
+      label: Additional Information
+      description: Provide any extra details that may assist in understanding the issue.
+    validations:
+      required: false
+
+  - type: markdown
+    attributes:
+      value: |
+        ## Note
+        If the bug report is incomplete or does not follow instructions, it may not be addressed. Ensure that you've followed all the **README.md** and **troubleshooting.md** guidelines, and provide all necessary information for us to reproduce the issue.  
+        Thank you for contributing to Open WebUI!

.github/ISSUE_TEMPLATE/config.yml

@@ -0,0 +1 @@
+blank_issues_enabled: false

.github/ISSUE_TEMPLATE/feature_request.md

@@ -1,35 +0,0 @@
----
-name: Feature request
-about: Suggest an idea for this project
-title: ''
-labels: ''
-assignees: ''
----
-
-# Feature Request
-
-## Important Notes
-
-- **Before submitting a report**: Please check the Issues or Discussions section to see if a similar issue or feature request has already been posted. It's likely we're already tracking it! If you’re unsure, start a discussion post first. This will help us efficiently focus on improving the project.
-
-- **Collaborate respectfully**: We value a constructive attitude, so please be mindful of your communication. If negativity is part of your approach, our capacity to engage may be limited. We’re here to help if you’re open to learning and communicating positively. Remember, Open WebUI is a volunteer-driven project managed by a single maintainer and supported by contributors who also have full-time jobs. We appreciate your time and ask that you respect ours.
-
-- **Contributing**: If you encounter an issue, we highly encourage you to submit a pull request or fork the project. We actively work to prevent contributor burnout to maintain the quality and continuity of Open WebUI.
-
-- **Bug reproducibility**: If a bug cannot be reproduced with a `:main` or `:dev` Docker setup, or a pip install with Python 3.11, it may require additional help from the community. In such cases, we will move it to the "issues" Discussions section due to our limited resources. We encourage the community to assist with these issues. Remember, it’s not that the issue doesn’t exist; we need your help!
-
-Note: Please remove the notes above when submitting your post. Thank you for your understanding and support!
-
----
-
-**Is your feature request related to a problem? Please describe.**
-A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
-
-**Describe the solution you'd like**
-A clear and concise description of what you want to happen.
-
-**Describe alternatives you've considered**
-A clear and concise description of any alternative solutions or features you've considered.
-
-**Additional context**
-Add any other context or screenshots about the feature request here.

.github/ISSUE_TEMPLATE/feature_request.yaml

@@ -0,0 +1,64 @@
+name: Feature Request
+description: Suggest an idea for this project
+title: 'feat: '
+labels: ['triage']
+body:
+  - type: markdown
+    attributes:
+      value: |
+        ## Important Notes
+        ### Before submitting
+        Please check the [Issues](https://github.com/open-webui/open-webui/issues) or [Discussions](https://github.com/open-webui/open-webui/discussions) to see if a similar request has been posted.
+        It's likely we're already tracking it! If you’re unsure, start a discussion post first.
+        This will help us efficiently focus on improving the project.
+
+        ### Collaborate respectfully
+        We value a **constructive attitude**, so please be mindful of your communication. If negativity is part of your approach, our capacity to engage may be limited. We're here to help if you're **open to learning** and **communicating positively**.
+
+        Remember:
+        - Open WebUI is a **volunteer-driven project**
+        - It's managed by a **single maintainer**
+        - It's supported by contributors who also have **full-time jobs**
+
+        We appreciate your time and ask that you **respect ours**.
+
+
+        ### Contributing
+        If you encounter an issue, we highly encourage you to submit a pull request or fork the project. We actively work to prevent contributor burnout to maintain the quality and continuity of Open WebUI.
+
+        ### Bug reproducibility
+        If a bug cannot be reproduced with a `:main` or `:dev` Docker setup, or a `pip install` with Python 3.11, it may require additional help from the community. In such cases, we will move it to the "[issues](https://github.com/open-webui/open-webui/discussions/categories/issues)" Discussions section due to our limited resources. We encourage the community to assist with these issues. Remember, it’s not that the issue doesn’t exist; we need your help!
+
+  - type: checkboxes
+    id: existing-issue
+    attributes:
+      label: Check Existing Issues
+      description: Please confirm that you've checked for existing similar requests
+      options:
+        - label: I have searched the existing issues and discussions.
+          required: true
+  - type: textarea
+    id: problem-description
+    attributes:
+      label: Problem Description
+      description: Is your feature request related to a problem? Please provide a clear and concise description of what the problem is.
+      placeholder: "Ex. I'm always frustrated when..."
+    validations:
+      required: true
+  - type: textarea
+    id: solution-description
+    attributes:
+      label: Desired Solution you'd like
+      description: Clearly describe what you want to happen.
+    validations:
+      required: true
+  - type: textarea
+    id: alternatives-considered
+    attributes:
+      label: Alternatives Considered
+      description: A clear and concise description of any alternative solutions or features you've considered.
+  - type: textarea
+    id: additional-context
+    attributes:
+      label: Additional Context
+      description: Add any other context or screenshots about the feature request here.

.github/dependabot.yml

@@ -1,12 +1,26 @@
 version: 2
 updates:
+  - package-ecosystem: uv
+    directory: '/'
+    schedule:
+      interval: monthly
+    target-branch: 'dev'
+
   - package-ecosystem: pip
     directory: '/backend'
     schedule:
       interval: monthly
     target-branch: 'dev'
+
+  - package-ecosystem: npm
+    directory: '/'
+    schedule:
+      interval: monthly
+    target-branch: 'dev'
+
   - package-ecosystem: 'github-actions'
     directory: '/'
     schedule:
       # Check for updates to GitHub Actions every week
       interval: monthly
+    target-branch: 'dev'

.github/pull_request_template.md

@@ -9,9 +9,9 @@
 - [ ] **Changelog:** Ensure a changelog entry following the format of [Keep a Changelog](https://keepachangelog.com/) is added at the bottom of the PR description.
 - [ ] **Documentation:** Have you updated relevant documentation [Open WebUI Docs](https://github.com/open-webui/docs), or other documentation sources?
 - [ ] **Dependencies:** Are there any new dependencies? Have you updated the dependency versions in the documentation?
-- [ ] **Testing:** Have you written and run sufficient tests for validating the changes?
+- [ ] **Testing:** Have you written and run sufficient tests to validate the changes?
 - [ ] **Code review:** Have you performed a self-review of your code, addressing any coding standard issues and ensuring adherence to the project's coding standards?
-- [ ] **Prefix:** To cleary categorize this pull request, prefix the pull request title, using one of the following:
+- [ ] **Prefix:** To clearly categorize this pull request, prefix the pull request title using one of the following:
   - **BREAKING CHANGE**: Significant changes that may affect compatibility
   - **build**: Changes that affect the build system or external dependencies
   - **ci**: Changes to our continuous integration processes or workflows
@@ -22,7 +22,7 @@
   - **i18n**: Internationalization or localization changes
   - **perf**: Performance improvement
   - **refactor**: Code restructuring for better maintainability, readability, or scalability
-  - **style**: Changes that do not affect the meaning of the code (white-space, formatting, missing semi-colons, etc.)
+  - **style**: Changes that do not affect the meaning of the code (white space, formatting, missing semi-colons, etc.)
   - **test**: Adding missing tests or correcting existing tests
   - **WIP**: Work in progress, a temporary label for incomplete or ongoing work
 

.github/workflows/docker-build.yaml

@@ -14,7 +14,7 @@ env:
 
 jobs:
   build-main-image:
-    runs-on: ubuntu-latest
+    runs-on: ${{ matrix.platform == 'linux/arm64' && 'ubuntu-24.04-arm' || 'ubuntu-latest' }}
     permissions:
       contents: read
       packages: write
@@ -111,7 +111,7 @@ jobs:
           retention-days: 1
 
   build-cuda-image:
-    runs-on: ubuntu-latest
+    runs-on: ${{ matrix.platform == 'linux/arm64' && 'ubuntu-24.04-arm' || 'ubuntu-latest' }}
     permissions:
       contents: read
       packages: write
@@ -211,7 +211,7 @@ jobs:
           retention-days: 1
 
   build-ollama-image:
-    runs-on: ubuntu-latest
+    runs-on: ${{ matrix.platform == 'linux/arm64' && 'ubuntu-24.04-arm' || 'ubuntu-latest' }}
     permissions:
       contents: read
       packages: write

.github/workflows/format-backend.yaml

@@ -5,10 +5,18 @@ on:
     branches:
       - main
       - dev
+    paths:
+      - 'backend/**'
+      - 'pyproject.toml'
+      - 'uv.lock'
   pull_request:
     branches:
       - main
       - dev
+    paths:
+      - 'backend/**'
+      - 'pyproject.toml'
+      - 'uv.lock'
 
 jobs:
   build:
@@ -17,7 +25,9 @@ jobs:
 
     strategy:
       matrix:
-        python-version: [3.11]
+        python-version:
+          - 3.11.x
+          - 3.12.x
 
     steps:
       - uses: actions/checkout@v4
@@ -25,7 +35,7 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v5
         with:
-          python-version: ${{ matrix.python-version }}
+          python-version: '${{ matrix.python-version }}'
 
       - name: Install dependencies
         run: |

.github/workflows/format-build-frontend.yaml

@@ -5,10 +5,18 @@ on:
     branches:
       - main
       - dev
+    paths-ignore:
+      - 'backend/**'
+      - 'pyproject.toml'
+      - 'uv.lock'
   pull_request:
     branches:
       - main
       - dev
+    paths-ignore:
+      - 'backend/**'
+      - 'pyproject.toml'
+      - 'uv.lock'
 
 jobs:
   build:
@@ -21,7 +29,7 @@ jobs:
       - name: Setup Node.js
         uses: actions/setup-node@v4
         with:
-          node-version: '22' # Or specify any other version you want to use
+          node-version: '22'
 
       - name: Install Dependencies
         run: npm install

.github/workflows/release-pypi.yml

@@ -19,7 +19,7 @@ jobs:
         uses: actions/checkout@v4
       - uses: actions/setup-node@v4
         with:
-          node-version: 18
+          node-version: 22
       - uses: actions/setup-python@v5
         with:
           python-version: 3.11

CHANGELOG.md

@@ -5,6 +5,326 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.6.3] - 2025-04-12
+
+### Added
+
+- 🧪 **Auto-Artifact Detection Toggle**: Automatically detects artifacts in results—but now you can disable this behavior under advanced settings for full control.
+- 🖼️ **Widescreen Mode for Shared Chats**: Shared link conversations now support widescreen layouts—perfect for presentations or easier review across wider displays.
+- 🔁 **Reindex Knowledge Files on Demand**: Admins can now trigger reindexing of all knowledge files after changing embeddings—ensuring immediate alignment with new models for optimal RAG performance.
+- 📄 **OpenAPI YAML Format Support**: External tools can now use YAML-format OpenAPI specs—making integration simpler for developers familiar with YAML-based configurations.
+- 💬 **Message Content Copy Behavior**: Copy action now excludes 'details' tags—streamlining clipboard content when sharing or pasting summaries elsewhere.
+- 🧭 **Sougou Web Search Integration**: New search engine option added—enhancing global relevance and diversity of search sources for multilingual users.
+- 🧰 **Frontend Web Loader Engine Configuration**: Admins can now set preferred web loader engine for RAG workflows directly from the frontend—offering more control across setups.
+- 👥 **Multi-Model Chat Permission Control**: Admins can manage access to multi-model chats per user group—allowing tighter governance in team environments.
+- 🧱 **Persistent Configuration Can Be Disabled**: New environment variable lets advanced users and hosts turn off persistent configs—ideal for volatile or stateless deployments.
+- 🧠 **Elixir Code Highlighting Support**: Elixir syntax is now beautifully rendered in code blocks—perfect for developers using this language in AI or automation projects.
+- 🌐 **PWA External Manifest URL Support**: You can now define an external manifest.json—integrate Open WebUI seamlessly in managed or proxy-based PWA environments like Cloudflare Zero Trust.
+- 🧪 **Azure AI Speech-to-Text Provider Integration**: Easily transcribe large audio files (up to 200MB) with high accuracy using Microsoft's Azure STT—fully configurable in Audio Settings.
+- 🔏 **PKCE (Code Challenge Method) Support for OIDC**: Enhance your OIDC login security with Proof Key for Code Exchange—ideal for zero-trust and native client apps.
+- ✨ **General UI/UX Enhancements**: Numerous refinements across layout, styling, and tool interactions—reducing visual noise and improving overall usability across key workflows.
+- 🌍 **Translation Updates Across Multiple Languages**: Refined Catalan, Russian, Chinese (Simplified & Traditional), Hungarian, and Spanish translations for clearer navigation and instructions globally.
+
+### Fixed
+
+- 💥 **Chat Completion Error with Missing Models Resolved**: Fixed internal server error when referencing a model that doesn’t exist—ensuring graceful fallback and clear error guidance.
+- 🔧 **Correct Knowledge Base Citations Restored**: Citations generated by RAG workflows now show accurate references—ensuring verifiability in outputs from sourced content.
+- 🎙️ **Broken OGG/WebM Audio Upload Handling for OpenAI Fixed**: Uploading OGG or WebM files now converts properly to WAV before transcription—restoring accurate AI speech recognition workflows.
+- 🔐 **Tool Server 'Session' Authentication Restored**: Previously broken session auth on external tool servers is now fully functional—ensuring secure and seamless access to connected tools.
+- 🌐 **Folder-Based Chat Rename Now Updates Correctly**: Renaming chats in folders now reflects instantly everywhere—improving chat organization and clarity.
+- 📜 **KaTeX Overflow Displays Fixed**: Math expressions now stay neatly within message bounds—preserving layout consistency even with long formulas.
+- 🚫 **Stopping Ongoing Chat Fixed**: You can now return to an active (ongoing) chat and stop generation at any time—ensuring full control over sessions.
+- 🔧 **TOOL_SERVERS / TOOL_SERVER_CONNECTIONS Indexing Issue Fixed**: Fixed a mismatch between tool lists and their access paths—restoring full function and preventing confusion in tool management.
+- 🔐 **LDAP Login Handles Multiple Emails**: When LDAP returns multiple email attributes, the first valid one is now used—ensuring login success and account consistency.
+- 🧩 **Model Visibility Toggle Fix**: Toggling model visibility now works even for untouched models—letting admins smoothly manage user access across base models.
+- ⚙️ **Cross-Origin manifest.json Now Loads Properly**: Compatibility issues with Cloudflare Zero Trust (and others) resolved, allowing manifest.json to load behind authenticated proxies.
+
+### Changed
+
+- 🔒 **Default Access Scopes Set to Private for All Resources**: Models, tools, and knowledge are now private by default when created—ensuring better baseline security and visibility controls.
+- 🧱 **General Backend Refactoring for Stability**: Numerous invisible improvements enhance backend scalability, security, and maintainability—powering upcoming features with a stronger foundation.
+- 🧩 **Stable Dependency Upgrades**: Updated key platform libraries—Chromadb (0.6.3), pgvector (0.4.0), Azure Identity (1.21.0), and Youtube Transcript API (1.0.3)—for improved compatibility, functionality, and security.
+
+## [0.6.2] - 2025-04-06
+
+### Added
+
+- 🌍 **Improved Global Language Support**: Expanded and refined translations across multiple languages to enhance clarity and consistency for international users.
+
+### Fixed
+
+- 🛠️ **Accurate Tool Descriptions from OpenAPI Servers**: External tools now use full endpoint descriptions instead of summaries when generating tool specifications—helping AI models understand tool purpose more precisely and choose the right tool more accurately in tool workflows.
+- 🔧 **Precise Web Results Source Attribution**: Fixed a key issue where all web search results showed the same source ID—now each result gets its correct and distinct source, ensuring accurate citations and traceability.
+- 🔍 **Clean Web Search Retrieval**: Web search now retains only results from URLs where real content was successfully fetched—improving accuracy and removing empty or broken links from citations.
+- 🎵 **Audio File Upload Response Restored**: Resolved an issue where uploading audio files did not return valid responses, restoring smooth file handling for transcription and audio-based workflows.
+
+### Changed
+
+- 🧰 **General Backend Refactoring**: Multiple behind-the-scenes improvements streamline backend performance, reduce complexity, and ensure a more stable, maintainable system overall—making everything smoother without changing your workflow.
+
+## [0.6.1] - 2025-04-05
+
+### Added
+
+- 🛠️ **Global Tool Servers Configuration**: Admins can now centrally configure global external tool servers from Admin Settings > Tools, allowing seamless sharing of tool integrations across all users without manual setup per user.
+- 🔐 **Direct Tool Usage Permission for Users**: Introduced a new user-level permission toggle that grants non-admin users access to direct external tools, empowering broader team collaboration while maintaining control.
+- 🧠 **Mistral OCR Content Extraction Support**: Added native support for Mistral OCR as a high-accuracy document loader, drastically improving text extraction from scanned documents in RAG workflows.
+- 🖼️ **Tools Indicator UI Redesign**: Enhanced message input now smartly displays both built-in and external tools via a unified dropdown, making it simpler and more intuitive to activate tools during conversations.
+- 📄 **RAG Prompt Improved and More Coherent**: Default RAG system prompt has been revised to be more clear and citation-focused—admins can leave the template field empty to use this new gold-standard prompt.
+- 🧰 **Performance & Developer Improvements**: Major internal restructuring of several tool-related components, simplifying styling and merging external/internal handling logic, resulting in better maintainability and performance.
+- 🌍 **Improved Translations**: Updated translations for Tibetan, Polish, Chinese (Simplified & Traditional), Arabic, Russian, Ukrainian, Dutch, Finnish, and French to improve clarity and consistency across the interface.
+
+### Fixed
+
+- 🔑 **External Tool Server API Key Bug Resolved**: Fixed a critical issue where authentication headers were not being sent when calling tools from external OpenAPI tool servers, ensuring full security and smooth tool operations.
+- 🚫 **Conditional Export Button Visibility**: UI now gracefully hides export buttons when there's nothing to export in models, prompts, tools, or functions, improving visual clarity and reducing confusion.
+- 🧪 **Hybrid Search Failure Recovery**: Resolved edge case in parallel hybrid search where empty or unindexed collections caused backend crashes—these are now cleanly skipped to ensure system stability.
+- 📂 **Admin Folder Deletion Fix**: Addressed an issue where folders created in the admin workspace couldn't be deleted, restoring full organizational flexibility for admins.
+- 🔐 **Improved Generic Error Feedback on Login**: Authentication errors now show simplified, non-revealing messages for privacy and improved UX, especially with federated logins.
+- 📝 **Tool Message with Images Improved**: Enhanced how tool-generated messages with image outputs are shown in chat, making them more readable and consistent with the overall UI design.
+- ⚙️ **Auto-Exclusion for Broken RAG Collections**: Auto-skips document collections that fail to fetch data or return "None", preventing silent errors and streamlining retrieval workflows.
+- 📝 **Docling Text File Handling Fix**: Fixed file parsing inconsistency that broke docling-based RAG functionality for certain plain text files, ensuring wider file compatibility.
+
+## [0.6.0] - 2025-03-31
+
+### Added
+
+- 🧩 **External Tool Server Support via OpenAPI**: Connect Open WebUI to any OpenAPI-compatible REST server instantly—offering immediate integration with thousands of developer tools, SDKs, and SaaS systems for powerful extensibility. Learn more: https://github.com/open-webui/openapi-servers
+- 🛠️ **MCP Server Support via MCPO**: You can now convert and expose your internal MCP tools as interoperable OpenAPI HTTP servers within Open WebUI for seamless, plug-n-play AI toolchain creation. Learn more: https://github.com/open-webui/mcpo
+- 📨 **/messages Chat API Endpoint Support**: For power users building external AI systems, new endpoints allow precise control of messages asynchronously—feed long-running external responses into Open WebUI chats without coupling with the frontend.
+- 📝 **Client-Side PDF Generation**: PDF exports are now generated fully client-side for drastically improved output quality—perfect for saving conversations or documents.
+- 💼 **Enforced Temporary Chats Mode**: Admins can now enforce temporary chat sessions by default to align with stringent data retention and compliance requirements.
+- 🌍 **Public Resource Sharing Permission Controls**: Fine-grained user group permissions now allow enabling/disabling public sharing for models, knowledge, prompts, and tools—ideal for privacy, team control, and internal deployments.
+- 📦 **Custom pip Options for Tools/Functions**: You can now specify custom pip installation options with "PIP_OPTIONS", "PIP_PACKAGE_INDEX_OPTIONS" environment variables—improving compatibility, support for private indexes, and better control over Python environments.
+- 🔢 **Editable Message Counter**: You can now double-click the message count number and jump straight to editing the index—quickly navigate complex chats or regenerate specific messages precisely.
+- 🧠 **Embedding Prefix Support Added**: Add custom prefixes to your embeddings for instruct-style tokens, enabling stronger model alignment and more consistent RAG performance.
+- 🙈 **Ability to Hide Base Models**: Optionally hide base models from the UI, helping users streamline model visibility and limit access to only usable endpoints..
+- 📚 **Docling Content Extraction Support**: Open WebUI now supports Docling as a content extraction engine, enabling smarter and more accurate parsing of complex file formats—ideal for advanced document understanding and Retrieval-Augmented Generation (RAG) workflows.
+- 🗃️ **Redis Sentinel Support Added**: Enhance deployment redundancy with support for Redis Sentinel for highly available, failover-safe Redis-based caching or pub/sub.
+- 📚 **JSON Schema Format for Ollama**: Added support for defining the format using JSON schema in Ollama-compatible models, improving flexibility and validation of model outputs.
+- 🔍 **Chat Sidebar Search "Clear” Button**: Quickly clear search filters in chat sidebar using the new ✖️ button—streamline your chat navigation with one click.
+- 🗂️ **Auto-Focus + Enter Submit for Folder Name**: When creating a new folder, the system automatically enters rename mode with name preselected—simplifying your org workflow.
+- 🧱 **Markdown Alerts Rendering**: Blockquotes with syntax hinting (e.g. ⚠️, ℹ️, ✅) now render styled Markdown alert banners, making messages and documentation more visually structured.
+- 🔁 **Hybrid Search Runs in Parallel Now**: Hybrid (BM25 + embedding) search components now run in parallel—dramatically reducing response times and speeding up document retrieval.
+- 📋 **Cleaner UI for Tool Call Display**: Optimized the visual layout of called tools inside chat messages for better clarity and reduced visual clutter.
+- 🧪 **Playwright Timeout Now Configurable**: Default timeout for Playwright processes is now shorter and adjustable via environment variables—making web scraping more robust and tunable to environments.
+- 📈 **OpenTelemetry Support for Observability**: Open WebUI now integrates with OpenTelemetry, allowing you to connect with tools like Grafana, Jaeger, or Prometheus for detailed performance insights and real-time visibility—entirely opt-in and fully self-hosted. Even if enabled, no data is ever sent to us, ensuring your privacy and ownership over all telemetry data.
+- 🛠 **General UI Enhancements & UX Polish**: Numerous refinements across sidebar, code blocks, modal interactions, button alignment, scrollbar visibility, and folder behavior improve overall fluidity and usability of the interface.
+- 🧱 **General Backend Refactoring**: Numerous backend components have been refactored to improve stability, maintainability, and performance—ensuring a more consistent and reliable system across all features.
+- 🌍 **Internationalization Language Support Updates**: Added Estonian and Galician languages, improved Spanish (fully revised), Traditional Chinese, Simplified Chinese, Turkish, Catalan, Ukrainian, and German for a more localized and inclusive interface.
+
+### Fixed
+
+- 🧑‍💻 **Firefox Input Height Bug**: Text input in Firefox now maintains proper height, ensuring message boxes look consistent and behave predictably.
+- 🧾 **Tika Blank Line Bug**: PDFs processed with Apache Tika 3.1.0.0 no longer introduce excessive blank lines—improving RAG output quality and visual cleanliness.
+- 🧪 **CSV Loader Encoding Issues**: CSV files with unknown encodings now automatically detect character sets, resolving import errors in non-UTF-8 datasets.
+- ✅ **LDAP Auth Config Fix**: Path to certificate file is now optional for LDAP setups, fixing authentication trouble for users without preconfigured cert paths.
+- 📥 **File Deletion in Bypass Mode**: Resolved issue where files couldn’t be deleted from knowledge when “bypass embedding” mode was enabled.
+- 🧩 **Hybrid Search Result Sorting & Deduplication Fixed**: Fixed citation and sorting issues in RAG hybrid and reranker modes, ensuring retrieved documents are shown in correct order per score.
+- 🧷 **Model Export/Import Broken for a Single Model**: Fixed bug where individual models couldn’t be exported or re-imported, restoring full portability.
+- 📫 **Auth Redirect Fix**: Logged-in users are now routed properly without unnecessary login prompts when already authenticated.
+
+### Changed
+
+- 🧠 **Prompt Autocompletion Disabled By Default**: Autocomplete suggestions while typing are now disabled unless explicitly re-enabled in user preferences—reduces distractions while composing prompts for advanced users.
+- 🧾 **Normalize Citation Numbering**: Source citations now properly begin from "1" instead of "0"—improving consistency and professional presentation in AI outputs.
+- 📚 **Improved Error Handling from Pipelines**: Pipelines now show the actual returned error message from failed tasks rather than generic "Connection closed"—making debugging far more user-friendly.
+
+### Removed
+
+- 🧾 **ENABLE_AUDIT_LOGS Setting Removed**: Deprecated setting “ENABLE_AUDIT_LOGS” has been fully removed—now controlled via “AUDIT_LOG_LEVEL” instead.
+
+## [0.5.20] - 2025-03-05
+
+### Added
+
+- **⚡ Toggle Code Execution On/Off**: You can now enable or disable code execution, providing more control over security, ensuring a safer and more customizable experience.
+
+### Fixed
+
+- **📜 Pinyin Keyboard Enter Key Now Works Properly**: Resolved an issue where the Enter key for Pinyin keyboards was not functioning as expected, ensuring seamless input for Chinese users.
+- **🖼️ Web Manifest Loading Issue Fixed**: Addressed inconsistencies with 'site.webmanifest', guaranteeing proper loading and representation of the app across different browsers and devices.
+- **📦 Non-Root Container Issue Resolved**: Fixed a critical issue where the UI failed to load correctly in non-root containers, ensuring reliable deployment in various environments.
+
+## [0.5.19] - 2025-03-04
+
+### Added
+
+- **📊 Logit Bias Parameter Support**: Fine-tune conversation dynamics by adjusting the Logit Bias parameter directly in chat settings, giving you more control over model responses.
+- **⌨️ Customizable Enter Behavior**: You can now configure Enter to send messages only when combined with Ctrl (Ctrl+Enter) via Settings > Interface, preventing accidental message sends.
+- **📝 Collapsible Code Blocks**: Easily collapse long code blocks to declutter your chat, making it easier to focus on important details.
+- **🏷️ Tag Selector in Model Selector**: Quickly find and categorize models with the new tag filtering system in the Model Selector, streamlining model discovery.
+- **📈 Experimental Elasticsearch Vector DB Support**: Now supports Elasticsearch as a vector database, offering more flexibility for data retrieval in Retrieval-Augmented Generation (RAG) workflows.
+- **⚙️ General Reliability Enhancements**: Various stability improvements across the WebUI, ensuring a smoother, more consistent experience.
+- **🌍 Updated Translations**: Refined multilingual support for better localization and accuracy across various languages.
+
+### Fixed
+
+- **🔄 "Stream" Hook Activation**: Fixed an issue where the "Stream" hook only worked when globally enabled, ensuring reliable real-time filtering.
+- **📧 LDAP Email Case Sensitivity**: Resolved an issue where LDAP login failed due to email case sensitivity mismatches, improving authentication reliability.
+- **💬 WebSocket Chat Event Registration**: Fixed a bug preventing chat event listeners from being registered upon sign-in, ensuring real-time updates work properly.
+
+## [0.5.18] - 2025-02-27
+
+### Fixed
+
+- **🌐 Open WebUI Now Works Over LAN in Insecure Context**: Resolved an issue preventing Open WebUI from functioning when accessed over a local network in an insecure context, ensuring seamless connectivity.
+- **🔄 UI Now Reflects Deleted Connections Instantly**: Fixed an issue where deleting a connection did not update the UI in real time, ensuring accurate system state visibility.
+- **🛠️ Models Now Display Correctly with ENABLE_FORWARD_USER_INFO_HEADERS**: Addressed a bug where models were not visible when ENABLE_FORWARD_USER_INFO_HEADERS was set, restoring proper model listing.
+
+## [0.5.17] - 2025-02-27
+
+### Added
+
+- **🚀 Instant Document Upload with Bypass Embedding & Retrieval**: Admins can now enable "Bypass Embedding & Retrieval" in Admin Settings > Documents, significantly speeding up document uploads and ensuring full document context is retained without chunking.
+- **🔎 "Stream" Hook for Real-Time Filtering**: The new "stream" hook allows dynamic real-time message filtering. Learn more in our documentation (https://docs.openwebui.com/features/plugin/functions/filter).
+- **☁️ OneDrive Integration**: Early support for OneDrive storage integration has been introduced, expanding file import options.
+- **📈 Enhanced Logging with Loguru**: Backend logging has been improved with Loguru, making debugging and issue tracking far more efficient.
+- **⚙️ General Stability Enhancements**: Backend and frontend refactoring improves performance, ensuring a smoother and more reliable user experience.
+- **🌍 Updated Translations**: Refined multilingual support for better localization and accuracy across various languages.
+
+### Fixed
+
+- **🔄 Reliable Model Imports from the Community Platform**: Resolved import failures, allowing seamless integration of community-shared models without errors.
+- **📊 OpenAI Usage Statistics Restored**: Fixed an issue where OpenAI usage metrics were not displaying correctly, ensuring accurate tracking of usage data.
+- **🗂️ Deduplication for Retrieved Documents**: Documents retrieved during searches are now intelligently deduplicated, meaning no more redundant results—helping to keep information concise and relevant.
+
+### Changed
+
+- **📝 "Full Context Mode" Renamed for Clarity**: The "Full Context Mode" toggle in Web Search settings is now labeled "Bypass Embedding & Retrieval" for consistency across the UI.
+
+## [0.5.16] - 2025-02-20
+
+### Fixed
+
+- **🔍 Web Search Retrieval Restored**: Resolved a critical issue that broke web search retrieval by reverting deduplication changes, ensuring complete and accurate search results once again.
+
+## [0.5.15] - 2025-02-20
+
+### Added
+
+- **📄 Full Context Mode for Local Document Search (RAG)**: Toggle full context mode from Admin Settings > Documents to inject entire document content into context, improving accuracy for models with large context windows—ideal for deep context understanding.
+- **🌍 Smarter Web Search with Agentic Workflows**: Web searches now intelligently gather and refine multiple relevant terms, similar to RAG handling, delivering significantly better search results for more accurate information retrieval.
+- **🔎 Experimental Playwright Support for Web Loader**: Web content retrieval is taken to the next level with Playwright-powered scraping for enhanced accuracy in extracted web data.
+- **☁️ Experimental Azure Storage Provider**: Early-stage support for Azure Storage allows more cloud storage flexibility directly within Open WebUI.
+- **📊 Improved Jupyter Code Execution with Plots**: Interactive coding now properly displays inline plots, making data visualization more seamless inside chat interactions.
+- **⏳ Adjustable Execution Timeout for Jupyter Interpreter**: Customize execution timeout (default: 60s) for Jupyter-based code execution, allowing longer or more constrained execution based on your needs.
+- **▶️ "Running..." Indicator for Jupyter Code Execution**: A visual indicator now appears while code execution is in progress, providing real-time status updates on ongoing computations.
+- **⚙️ General Backend & Frontend Stability Enhancements**: Extensive refactoring improves reliability, performance, and overall user experience for a more seamless Open WebUI.
+- **🌍 Translation Updates**: Various international translation refinements ensure better localization and a more natural user interface experience.
+
+### Fixed
+
+- **📱 Mobile Hover Issue Resolved**: Users can now edit responses smoothly on mobile without interference, fixing a longstanding hover issue.
+- **🔄 Temporary Chat Message Duplication Fixed**: Eliminated buggy behavior where messages were being unnecessarily repeated in temporary chat mode, ensuring a smooth and consistent conversation flow.
+
+## [0.5.14] - 2025-02-17
+
+### Fixed
+
+- **🔧 Critical Import Error Resolved**: Fixed a circular import issue preventing 'override_static' from being correctly imported in 'open_webui.config', ensuring smooth system initialization and stability.
+
+## [0.5.13] - 2025-02-17
+
+### Added
+
+- **🌐 Full Context Mode for Web Search**: Enable highly accurate web searches by utilizing full context mode—ideal for models with large context windows, ensuring more precise and insightful results.
+- **⚡ Optimized Asynchronous Web Search**: Web searches now load significantly faster with optimized async support, providing users with quicker, more efficient information retrieval.
+- **🔄 Auto Text Direction for RTL Languages**: Automatic text alignment based on language input, ensuring seamless conversation flow for Arabic, Hebrew, and other right-to-left scripts.
+- **🚀 Jupyter Notebook Support for Code Execution**: The "Run" button in code blocks can now use Jupyter for execution, offering a powerful, dynamic coding experience directly in the chat.
+- **🗑️ Message Delete Confirmation Dialog**: Prevent accidental deletions with a new confirmation prompt before removing messages, adding an additional layer of security to your chat history.
+- **📥 Download Button for SVG Diagrams**: SVG diagrams generated within chat can now be downloaded instantly, making it easier to save and share complex visual data.
+- **✨ General UI/UX Improvements and Backend Stability**: A refined interface with smoother interactions, improved layouts, and backend stability enhancements for a more reliable, polished experience.
+
+### Fixed
+
+- **🛠️ Temporary Chat Message Continue Button Fixed**: The "Continue Response" button for temporary chats now works as expected, ensuring an uninterrupted conversation flow.
+
+### Changed
+
+- **📝 Prompt Variable Update**: Deprecated square bracket '[]' indicators for prompt variables; now requires double curly brackets '{{}}' for consistency and clarity.
+- **🔧 Stability Enhancements**: Error handling improved in chat history, ensuring smoother operations when reviewing previous messages.
+
+## [0.5.12] - 2025-02-13
+
+### Added
+
+- **🛠️ Multiple Tool Calls Support for Native Function Mode**: Functions now can call multiple tools within a single response, unlocking better automation and workflow flexibility when using native function calling.
+
+### Fixed
+
+- **📝 Playground Text Completion Restored**: Addressed an issue where text completion in the Playground was not functioning.
+- **🔗 Direct Connections Now Work for Regular Users**: Fixed a bug where users with the 'user' role couldn't establish direct API connections, enabling seamless model usage for all user tiers.
+- **⚡ Landing Page Input No Longer Lags with Long Text**: Improved input responsiveness on the landing page, ensuring fast and smooth typing experiences even when entering long messages.
+- **🔧 Parameter in Functions Fixed**: Fixed an issue where the reserved parameters wasn’t recognized within functions, restoring full functionality for advanced task-based automation.
+
+## [0.5.11] - 2025-02-13
+
+### Added
+
+- **🎤 Kokoro-JS TTS Support**: A new on-device, high-quality text-to-speech engine has been integrated, vastly improving voice generation quality—everything runs directly in your browser.
+- **🐍 Jupyter Notebook Support in Code Interpreter**: Now, you can configure Code Interpreter to run Python code not only via Pyodide but also through Jupyter, offering a more robust coding environment for AI-driven computations and analysis.
+- **🔗 Direct API Connections for Private & Local Inference**: You can now connect Open WebUI to your private or localhost API inference endpoints. CORS must be enabled, but this unlocks direct, on-device AI infrastructure support.
+- **🔍 Advanced Domain Filtering for Web Search**: You can now specify which domains should be included or excluded from web searches, refining results for more relevant information retrieval.
+- **🚀 Improved Image Generation Metadata Handling**: Generated images now retain metadata for better organization and future retrieval.
+- **📂 S3 Key Prefix Support**: Fine-grained control over S3 storage file structuring with configurable key prefixes.
+- **📸 Support for Image-Only Messages**: Send messages containing only images, facilitating more visual-centric interactions.
+- **🌍 Updated Translations**: German, Spanish, Traditional Chinese, and Catalan translations updated for better multilingual support.
+
+### Fixed
+
+- **🔧 OAuth Debug Logs & Username Claim Fixes**: Debug logs have been added for OAuth role and group management, with fixes ensuring proper OAuth username retrieval and claim handling.
+- **📌 Citations Formatting & Toggle Fixes**: Inline citation toggles now function correctly, and citations with more than three sources are now fully visible when expanded.
+- **📸 ComfyUI Maximum Seed Value Constraint Fixed**: The maximum allowed seed value for ComfyUI has been corrected, preventing unintended behavior.
+- **🔑 Connection Settings Stability**: Addressed connection settings issues that were causing instability when saving configurations.
+- **📂 GGUF Model Upload Stability**: Fixed upload inconsistencies for GGUF models, ensuring reliable local model handling.
+- **🔧 Web Search Configuration Bug**: Fixed issues where web search filters and settings weren't correctly applied.
+- **💾 User Settings Persistence Fix**: Ensured user-specific settings are correctly saved and applied across sessions.
+- **🔄 OpenID Username Retrieval Enhancement**: Usernames are now correctly picked up and assigned for OpenID Connect (OIDC) logins.
+
+## [0.5.10] - 2025-02-05
+
+### Fixed
+
+- **⚙️ System Prompts Now Properly Templated via API**: Resolved an issue where system prompts were not being correctly processed when used through the API, ensuring template variables now function as expected.
+- **📝 '<thinking>' Tag Display Issue Fixed**: Fixed a bug where the 'thinking' tag was disrupting content rendering, ensuring clean and accurate text display.
+- **💻 Code Interpreter Stability with Custom Functions**: Addressed failures when using the Code Interpreter with certain custom functions like Anthropic, ensuring smoother execution and better compatibility.
+
+## [0.5.9] - 2025-02-05
+
+### Fixed
+
+- **💡 "Think" Tag Display Issue**: Resolved a bug where the "Think" tag was not functioning correctly, ensuring proper visualization of the model's reasoning process before delivering responses.
+
+## [0.5.8] - 2025-02-05
+
+### Added
+
+- **🖥️ Code Interpreter**: Models can now execute code in real time to refine their answers dynamically, running securely within a sandboxed browser environment using Pyodide. Perfect for calculations, data analysis, and AI-assisted coding tasks!
+- **💬 Redesigned Chat Input UI**: Enjoy a sleeker and more intuitive message input with improved feature selection, making it easier than ever to toggle tools, enable search, and interact with AI seamlessly.
+- **🛠️ Native Tool Calling Support (Experimental)**: Supported models can now call tools natively, reducing query latency and improving contextual responses. More enhancements coming soon!
+- **🔗 Exa Search Engine Integration**: A new search provider has been added, allowing users to retrieve up-to-date and relevant information without leaving the chat interface.
+- **🌍 Localized Dates & Times**: Date and time formats now match your system locale, ensuring a more natural, region-specific experience.
+- **📎 User Headers for External Embedding APIs**: API calls to external embedding services now include user-related headers.
+- **🌍 "Always On" Web Search Toggle**: A new option under Settings > Interface allows users to enable Web Search by default—transform Open WebUI into your go-to search engine, ensuring AI-powered results with every query.
+- **🚀 General Performance & Stability**: Significant improvements across the platform for a faster, more reliable experience.
+- **🖼️ UI/UX Enhancements**: Numerous design refinements improving readability, responsiveness, and accessibility.
+- **🌍 Improved Translations**: Chinese, Korean, French, Ukrainian and Serbian translations have been updated with refined terminologies for better clarity.
+
+### Fixed
+
+- **🔄 OAuth Name Field Fallback**: Resolves OAuth login failures by using the email field as a fallback when a name is missing.
+- **🔑 Google Drive Credentials Restriction**: Ensures only authenticated users can access Google Drive credentials for enhanced security.
+- **🌐 DuckDuckGo Search Rate Limit Handling**: Fixes issues where users would encounter 202 errors due to rate limits when using DuckDuckGo for web search.
+- **📁 File Upload Permission Indicator**: Users are now notified when they lack permission to upload files, improving clarity on system restrictions.
+- **🔧 Max Tokens Issue**: Fixes cases where 'max_tokens' were not applied correctly, ensuring proper model behavior.
+- **🔍 Validation for RAG Web Search URLs**: Filters out invalid or unsupported URLs when using web-based retrieval augmentation.
+- **🖋️ Title Generation Bug**: Fixes inconsistencies in title generation, ensuring proper chat organization.
+
+### Removed
+
+- **⚡ Deprecated Non-Web Worker Pyodide Execution**: Moves entirely to browser sandboxing for better performance and security.
+
 ## [0.5.7] - 2025-01-23
 
 ### Added

Dockerfile

@@ -132,7 +132,7 @@ RUN if [ "$USE_OLLAMA" = "true" ]; then \
 # install python dependencies
 COPY --chown=$UID:$GID ./backend/requirements.txt ./requirements.txt
 
-RUN pip3 install uv && \
+RUN pip3 install --no-cache-dir uv && \
     if [ "$USE_CUDA" = "true" ]; then \
     # If you use CUDA the whisper and embedding model will be downloaded on first use
     pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/$USE_CUDA_DOCKER_VER --no-cache-dir && \

README.md

@@ -13,10 +13,15 @@
 
 **Open WebUI is an [extensible](https://docs.openwebui.com/features/plugin/), feature-rich, and user-friendly self-hosted AI platform designed to operate entirely offline.** It supports various LLM runners like **Ollama** and **OpenAI-compatible APIs**, with **built-in inference engine** for RAG, making it a **powerful AI deployment solution**.
 
-For more information, be sure to check out our [Open WebUI Documentation](https://docs.openwebui.com/).
-
 ![Open WebUI Demo](./demo.gif)
 
+> [!TIP]  
+> **Looking for an [Enterprise Plan](https://docs.openwebui.com/enterprise)?** – **[Speak with Our Sales Team Today!](mailto:sales@openwebui.com)**
+>
+> Get **enhanced capabilities**, including **custom theming and branding**, **Service Level Agreement (SLA) support**, **Long-Term Support (LTS) versions**, and **more!**
+
+For more information, be sure to check out our [Open WebUI Documentation](https://docs.openwebui.com/).
+
 ## Key Features of Open WebUI ⭐
 
 - 🚀 **Effortless Setup**: Install seamlessly using Docker or Kubernetes (kubectl, kustomize or helm) for a hassle-free experience with support for both `:ollama` and `:cuda` tagged images.
@@ -174,7 +179,7 @@ docker run --rm --volume /var/run/docker.sock:/var/run/docker.sock containrrr/wa
 
 In the last part of the command, replace `open-webui` with your container name if it is different.
 
-Check our Migration Guide available in our [Open WebUI Documentation](https://docs.openwebui.com/tutorials/migration/).
+Check our Updating Guide available in our [Open WebUI Documentation](https://docs.openwebui.com/getting-started/updating).
 
 ### Using the Dev Branch 🌙
 

backend/open_webui/__init__.py

@@ -73,8 +73,15 @@ def serve(
             os.environ["LD_LIBRARY_PATH"] = ":".join(LD_LIBRARY_PATH)
 
     import open_webui.main  # we need set environment variables before importing main
+    from open_webui.env import UVICORN_WORKERS  # Import the workers setting
 
-    uvicorn.run(open_webui.main.app, host=host, port=port, forwarded_allow_ips="*")
+    uvicorn.run(
+        open_webui.main.app,
+        host=host,
+        port=port,
+        forwarded_allow_ips="*",
+        workers=UVICORN_WORKERS,
+    )
 
 
 @app.command()

backend/open_webui/constants.py

@@ -31,6 +31,7 @@ class ERROR_MESSAGES(str, Enum):
     USERNAME_TAKEN = (
         "Uh-oh! This username is already registered. Please choose another username."
     )
+    PASSWORD_TOO_LONG = "Uh-oh! The password you entered is too long. Please make sure your password is less than 72 bytes long."
     COMMAND_TAKEN = "Uh-oh! This command is already registered. Please choose another command string."
     FILE_EXISTS = "Uh-oh! This file is already registered. Please choose another file."
 
@@ -57,7 +58,7 @@ class ERROR_MESSAGES(str, Enum):
     )
 
     FILE_NOT_SENT = "FILE_NOT_SENT"
-    FILE_NOT_SUPPORTED = "Oops! It seems like the file format you're trying to upload is not supported. Please upload a file with a supported format (e.g., JPG, PNG, PDF, TXT) and try again."
+    FILE_NOT_SUPPORTED = "Oops! It seems like the file format you're trying to upload is not supported. Please upload a file with a supported format and try again."
 
     NOT_FOUND = "We could not find what you're looking for :/"
     USER_NOT_FOUND = "We could not find what you're looking for :/"

backend/open_webui/env.py

@@ -65,10 +65,8 @@ except Exception:
 # LOGGING
 ####################################
 
-log_levels = ["CRITICAL", "ERROR", "WARNING", "INFO", "DEBUG"]
-
 GLOBAL_LOG_LEVEL = os.environ.get("GLOBAL_LOG_LEVEL", "").upper()
-if GLOBAL_LOG_LEVEL in log_levels:
+if GLOBAL_LOG_LEVEL in logging.getLevelNamesMapping():
     logging.basicConfig(stream=sys.stdout, level=GLOBAL_LOG_LEVEL, force=True)
 else:
     GLOBAL_LOG_LEVEL = "INFO"
@@ -78,6 +76,7 @@ log.info(f"GLOBAL_LOG_LEVEL: {GLOBAL_LOG_LEVEL}")
 
 if "cuda_error" in locals():
     log.exception(cuda_error)
+    del cuda_error
 
 log_sources = [
     "AUDIO",
@@ -92,6 +91,7 @@ log_sources = [
     "RAG",
     "WEBHOOK",
     "SOCKET",
+    "OAUTH",
 ]
 
 SRC_LOG_LEVELS = {}
@@ -99,19 +99,19 @@ SRC_LOG_LEVELS = {}
 for source in log_sources:
     log_env_var = source + "_LOG_LEVEL"
     SRC_LOG_LEVELS[source] = os.environ.get(log_env_var, "").upper()
-    if SRC_LOG_LEVELS[source] not in log_levels:
+    if SRC_LOG_LEVELS[source] not in logging.getLevelNamesMapping():
         SRC_LOG_LEVELS[source] = GLOBAL_LOG_LEVEL
     log.info(f"{log_env_var}: {SRC_LOG_LEVELS[source]}")
 
 log.setLevel(SRC_LOG_LEVELS["CONFIG"])
 
-
 WEBUI_NAME = os.environ.get("WEBUI_NAME", "Open WebUI")
 if WEBUI_NAME != "Open WebUI":
     WEBUI_NAME += " (Open WebUI)"
 
 WEBUI_FAVICON_URL = "https://openwebui.com/favicon.png"
 
+TRUSTED_SIGNATURE_KEY = os.environ.get("TRUSTED_SIGNATURE_KEY", "")
 
 ####################################
 # ENV (dev,test,prod)
@@ -129,7 +129,6 @@ else:
     except Exception:
         PACKAGE_DATA = {"version": "0.0.0"}
 
-
 VERSION = PACKAGE_DATA["version"]
 
 
@@ -160,7 +159,6 @@ try:
 except Exception:
     changelog_content = (pkgutil.get_data("open_webui", "CHANGELOG.md") or b"").decode()
 
-
 # Convert markdown content to HTML
 html_content = markdown.markdown(changelog_content)
 
@@ -191,7 +189,6 @@ for version in soup.find_all("h2"):
 
     changelog_json[version_number] = version_data
 
-
 CHANGELOG = changelog_json
 
 ####################################
@@ -208,7 +205,6 @@ ENABLE_FORWARD_USER_INFO_HEADERS = (
     os.environ.get("ENABLE_FORWARD_USER_INFO_HEADERS", "False").lower() == "true"
 )
 
-
 ####################################
 # WEBUI_BUILD_HASH
 ####################################
@@ -243,7 +239,6 @@ if FROM_INIT_PY:
 
     DATA_DIR = Path(os.getenv("DATA_DIR", OPEN_WEBUI_DIR / "data"))
 
-
 STATIC_DIR = Path(os.getenv("STATIC_DIR", OPEN_WEBUI_DIR / "static"))
 
 FONTS_DIR = Path(os.getenv("FONTS_DIR", OPEN_WEBUI_DIR / "static" / "fonts"))
@@ -255,7 +250,6 @@ if FROM_INIT_PY:
         os.getenv("FRONTEND_BUILD_DIR", OPEN_WEBUI_DIR / "frontend")
     ).resolve()
 
-
 ####################################
 # Database
 ####################################
@@ -320,7 +314,6 @@ RESET_CONFIG_ON_START = (
     os.environ.get("RESET_CONFIG_ON_START", "False").lower() == "true"
 )
 
-
 ENABLE_REALTIME_CHAT_SAVE = (
     os.environ.get("ENABLE_REALTIME_CHAT_SAVE", "False").lower() == "true"
 )
@@ -329,7 +322,23 @@ ENABLE_REALTIME_CHAT_SAVE = (
 # REDIS
 ####################################
 
-REDIS_URL = os.environ.get("REDIS_URL", "redis://localhost:6379/0")
+REDIS_URL = os.environ.get("REDIS_URL", "")
+REDIS_SENTINEL_HOSTS = os.environ.get("REDIS_SENTINEL_HOSTS", "")
+REDIS_SENTINEL_PORT = os.environ.get("REDIS_SENTINEL_PORT", "26379")
+
+####################################
+# UVICORN WORKERS
+####################################
+
+# Number of uvicorn worker processes for handling requests
+UVICORN_WORKERS = os.environ.get("UVICORN_WORKERS", "1")
+try:
+    UVICORN_WORKERS = int(UVICORN_WORKERS)
+    if UVICORN_WORKERS < 1:
+        UVICORN_WORKERS = 1
+except ValueError:
+    UVICORN_WORKERS = 1
+    log.info(f"Invalid UVICORN_WORKERS value, defaulting to {UVICORN_WORKERS}")
 
 ####################################
 # WEBUI_AUTH (Required for security)
@@ -356,14 +365,22 @@ WEBUI_SECRET_KEY = os.environ.get(
     ),  # DEPRECATED: remove at next major version
 )
 
-WEBUI_SESSION_COOKIE_SAME_SITE = os.environ.get(
-    "WEBUI_SESSION_COOKIE_SAME_SITE",
-    os.environ.get("WEBUI_SESSION_COOKIE_SAME_SITE", "lax"),
+WEBUI_SESSION_COOKIE_SAME_SITE = os.environ.get("WEBUI_SESSION_COOKIE_SAME_SITE", "lax")
+
+WEBUI_SESSION_COOKIE_SECURE = (
+    os.environ.get("WEBUI_SESSION_COOKIE_SECURE", "false").lower() == "true"
 )
 
-WEBUI_SESSION_COOKIE_SECURE = os.environ.get(
-    "WEBUI_SESSION_COOKIE_SECURE",
-    os.environ.get("WEBUI_SESSION_COOKIE_SECURE", "false").lower() == "true",
+WEBUI_AUTH_COOKIE_SAME_SITE = os.environ.get(
+    "WEBUI_AUTH_COOKIE_SAME_SITE", WEBUI_SESSION_COOKIE_SAME_SITE
+)
+
+WEBUI_AUTH_COOKIE_SECURE = (
+    os.environ.get(
+        "WEBUI_AUTH_COOKIE_SECURE",
+        os.environ.get("WEBUI_SESSION_COOKIE_SECURE", "false"),
+    ).lower()
+    == "true"
 )
 
 if WEBUI_AUTH and WEBUI_SECRET_KEY == "":
@@ -376,6 +393,11 @@ ENABLE_WEBSOCKET_SUPPORT = (
 WEBSOCKET_MANAGER = os.environ.get("WEBSOCKET_MANAGER", "")
 
 WEBSOCKET_REDIS_URL = os.environ.get("WEBSOCKET_REDIS_URL", REDIS_URL)
+WEBSOCKET_REDIS_LOCK_TIMEOUT = os.environ.get("WEBSOCKET_REDIS_LOCK_TIMEOUT", 60)
+
+WEBSOCKET_SENTINEL_HOSTS = os.environ.get("WEBSOCKET_SENTINEL_HOSTS", "")
+
+WEBSOCKET_SENTINEL_PORT = os.environ.get("WEBSOCKET_SENTINEL_PORT", "26379")
 
 AIOHTTP_CLIENT_TIMEOUT = os.environ.get("AIOHTTP_CLIENT_TIMEOUT", "")
 
@@ -387,19 +409,33 @@ else:
     except Exception:
         AIOHTTP_CLIENT_TIMEOUT = 300
 
-AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST = os.environ.get(
-    "AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST", ""
+AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST = os.environ.get(
+    "AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST",
+    os.environ.get("AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST", "10"),
 )
 
-if AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST == "":
-    AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST = None
+if AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST == "":
+    AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST = None
 else:
     try:
-        AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST = int(
-            AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST
+        AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST = int(AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST)
+    except Exception:
+        AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST = 10
+
+
+AIOHTTP_CLIENT_TIMEOUT_TOOL_SERVER_DATA = os.environ.get(
+    "AIOHTTP_CLIENT_TIMEOUT_TOOL_SERVER_DATA", "10"
+)
+
+if AIOHTTP_CLIENT_TIMEOUT_TOOL_SERVER_DATA == "":
+    AIOHTTP_CLIENT_TIMEOUT_TOOL_SERVER_DATA = None
+else:
+    try:
+        AIOHTTP_CLIENT_TIMEOUT_TOOL_SERVER_DATA = int(
+            AIOHTTP_CLIENT_TIMEOUT_TOOL_SERVER_DATA
         )
     except Exception:
-        AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST = 5
+        AIOHTTP_CLIENT_TIMEOUT_TOOL_SERVER_DATA = 10
 
 ####################################
 # OFFLINE_MODE
@@ -409,3 +445,54 @@ OFFLINE_MODE = os.environ.get("OFFLINE_MODE", "false").lower() == "true"
 
 if OFFLINE_MODE:
     os.environ["HF_HUB_OFFLINE"] = "1"
+
+####################################
+# AUDIT LOGGING
+####################################
+# Where to store log file
+AUDIT_LOGS_FILE_PATH = f"{DATA_DIR}/audit.log"
+# Maximum size of a file before rotating into a new log file
+AUDIT_LOG_FILE_ROTATION_SIZE = os.getenv("AUDIT_LOG_FILE_ROTATION_SIZE", "10MB")
+# METADATA | REQUEST | REQUEST_RESPONSE
+AUDIT_LOG_LEVEL = os.getenv("AUDIT_LOG_LEVEL", "NONE").upper()
+try:
+    MAX_BODY_LOG_SIZE = int(os.environ.get("MAX_BODY_LOG_SIZE") or 2048)
+except ValueError:
+    MAX_BODY_LOG_SIZE = 2048
+
+# Comma separated list for urls to exclude from audit
+AUDIT_EXCLUDED_PATHS = os.getenv("AUDIT_EXCLUDED_PATHS", "/chats,/chat,/folders").split(
+    ","
+)
+AUDIT_EXCLUDED_PATHS = [path.strip() for path in AUDIT_EXCLUDED_PATHS]
+AUDIT_EXCLUDED_PATHS = [path.lstrip("/") for path in AUDIT_EXCLUDED_PATHS]
+
+####################################
+# OPENTELEMETRY
+####################################
+
+ENABLE_OTEL = os.environ.get("ENABLE_OTEL", "False").lower() == "true"
+OTEL_EXPORTER_OTLP_ENDPOINT = os.environ.get(
+    "OTEL_EXPORTER_OTLP_ENDPOINT", "http://localhost:4317"
+)
+OTEL_SERVICE_NAME = os.environ.get("OTEL_SERVICE_NAME", "open-webui")
+OTEL_RESOURCE_ATTRIBUTES = os.environ.get(
+    "OTEL_RESOURCE_ATTRIBUTES", ""
+)  # e.g. key1=val1,key2=val2
+OTEL_TRACES_SAMPLER = os.environ.get(
+    "OTEL_TRACES_SAMPLER", "parentbased_always_on"
+).lower()
+
+####################################
+# TOOLS/FUNCTIONS PIP OPTIONS
+####################################
+
+PIP_OPTIONS = os.getenv("PIP_OPTIONS", "").split()
+PIP_PACKAGE_INDEX_OPTIONS = os.getenv("PIP_PACKAGE_INDEX_OPTIONS", "").split()
+
+
+####################################
+# PROGRESSIVE WEB APP OPTIONS
+####################################
+
+EXTERNAL_PWA_MANIFEST_URL = os.environ.get("EXTERNAL_PWA_MANIFEST_URL")

backend/open_webui/functions.py

@@ -2,6 +2,7 @@ import logging
 import sys
 import inspect
 import json
+import asyncio
 
 from pydantic import BaseModel
 from typing import AsyncGenerator, Generator, Iterator
@@ -76,11 +77,13 @@ async def get_function_models(request):
         if hasattr(function_module, "pipes"):
             sub_pipes = []
 
-            # Check if pipes is a function or a list
-
+            # Handle pipes being a list, sync function, or async function
             try:
                 if callable(function_module.pipes):
-                    sub_pipes = function_module.pipes()
+                    if asyncio.iscoroutinefunction(function_module.pipes):
+                        sub_pipes = await function_module.pipes()
+                    else:
+                        sub_pipes = function_module.pipes()
                 else:
                     sub_pipes = function_module.pipes
             except Exception as e:
@@ -220,6 +223,9 @@ async def generate_function_chat_completion(
     extra_params = {
         "__event_emitter__": __event_emitter__,
         "__event_call__": __event_call__,
+        "__chat_id__": metadata.get("chat_id", None),
+        "__session_id__": metadata.get("session_id", None),
+        "__message_id__": metadata.get("message_id", None),
         "__task__": __task__,
         "__task_body__": __task_body__,
         "__files__": files,
@@ -250,7 +256,7 @@ async def generate_function_chat_completion(
 
         params = model_info.params.model_dump()
         form_data = apply_model_params_to_body_openai(params, form_data)
-        form_data = apply_model_system_prompt_to_body(params, form_data, user)
+        form_data = apply_model_system_prompt_to_body(params, form_data, metadata, user)
 
     pipe_id = get_pipe_id(form_data)
     function_module = get_function_module_by_id(request, pipe_id)

backend/open_webui/main.py

@@ -45,6 +45,9 @@ from starlette.middleware.sessions import SessionMiddleware
 from starlette.responses import Response, StreamingResponse
 
 
+from open_webui.utils import logger
+from open_webui.utils.audit import AuditLevel, AuditLoggingMiddleware
+from open_webui.utils.logger import start_logger
 from open_webui.socket.main import (
     app as socket_app,
     periodic_usage_pool_cleanup,
@@ -81,22 +84,45 @@ from open_webui.routers.retrieval import (
     get_rf,
 )
 
-from open_webui.internal.db import Session
+from open_webui.internal.db import Session, engine
 
 from open_webui.models.functions import Functions
 from open_webui.models.models import Models
 from open_webui.models.users import UserModel, Users
+from open_webui.models.chats import Chats
 
 from open_webui.config import (
+    LICENSE_KEY,
     # Ollama
     ENABLE_OLLAMA_API,
     OLLAMA_BASE_URLS,
     OLLAMA_API_CONFIGS,
     # OpenAI
     ENABLE_OPENAI_API,
+    ONEDRIVE_CLIENT_ID,
     OPENAI_API_BASE_URLS,
     OPENAI_API_KEYS,
     OPENAI_API_CONFIGS,
+    # Direct Connections
+    ENABLE_DIRECT_CONNECTIONS,
+    # Tool Server Configs
+    TOOL_SERVER_CONNECTIONS,
+    # Code Execution
+    ENABLE_CODE_EXECUTION,
+    CODE_EXECUTION_ENGINE,
+    CODE_EXECUTION_JUPYTER_URL,
+    CODE_EXECUTION_JUPYTER_AUTH,
+    CODE_EXECUTION_JUPYTER_AUTH_TOKEN,
+    CODE_EXECUTION_JUPYTER_AUTH_PASSWORD,
+    CODE_EXECUTION_JUPYTER_TIMEOUT,
+    ENABLE_CODE_INTERPRETER,
+    CODE_INTERPRETER_ENGINE,
+    CODE_INTERPRETER_PROMPT_TEMPLATE,
+    CODE_INTERPRETER_JUPYTER_URL,
+    CODE_INTERPRETER_JUPYTER_AUTH,
+    CODE_INTERPRETER_JUPYTER_AUTH_TOKEN,
+    CODE_INTERPRETER_JUPYTER_AUTH_PASSWORD,
+    CODE_INTERPRETER_JUPYTER_TIMEOUT,
     # Image
     AUTOMATIC1111_API_AUTH,
     AUTOMATIC1111_BASE_URL,
@@ -115,11 +141,16 @@ from open_webui.config import (
     IMAGE_STEPS,
     IMAGES_OPENAI_API_BASE_URL,
     IMAGES_OPENAI_API_KEY,
+    IMAGES_GEMINI_API_BASE_URL,
+    IMAGES_GEMINI_API_KEY,
     # Audio
     AUDIO_STT_ENGINE,
     AUDIO_STT_MODEL,
     AUDIO_STT_OPENAI_API_BASE_URL,
     AUDIO_STT_OPENAI_API_KEY,
+    AUDIO_STT_AZURE_API_KEY,
+    AUDIO_STT_AZURE_REGION,
+    AUDIO_STT_AZURE_LOCALES,
     AUDIO_TTS_API_KEY,
     AUDIO_TTS_ENGINE,
     AUDIO_TTS_MODEL,
@@ -129,12 +160,20 @@ from open_webui.config import (
     AUDIO_TTS_VOICE,
     AUDIO_TTS_AZURE_SPEECH_REGION,
     AUDIO_TTS_AZURE_SPEECH_OUTPUT_FORMAT,
+    PLAYWRIGHT_WS_URL,
+    PLAYWRIGHT_TIMEOUT,
+    FIRECRAWL_API_BASE_URL,
+    FIRECRAWL_API_KEY,
+    WEB_LOADER_ENGINE,
     WHISPER_MODEL,
+    DEEPGRAM_API_KEY,
     WHISPER_MODEL_AUTO_UPDATE,
     WHISPER_MODEL_DIR,
     # Retrieval
     RAG_TEMPLATE,
     DEFAULT_RAG_TEMPLATE,
+    RAG_FULL_CONTEXT,
+    BYPASS_EMBEDDING_AND_RETRIEVAL,
     RAG_EMBEDDING_MODEL,
     RAG_EMBEDDING_MODEL_AUTO_UPDATE,
     RAG_EMBEDDING_MODEL_TRUST_REMOTE_CODE,
@@ -154,40 +193,57 @@ from open_webui.config import (
     CHUNK_SIZE,
     CONTENT_EXTRACTION_ENGINE,
     TIKA_SERVER_URL,
+    DOCLING_SERVER_URL,
+    DOCUMENT_INTELLIGENCE_ENDPOINT,
+    DOCUMENT_INTELLIGENCE_KEY,
+    MISTRAL_OCR_API_KEY,
     RAG_TOP_K,
+    RAG_TOP_K_RERANKER,
     RAG_TEXT_SPLITTER,
     TIKTOKEN_ENCODING_NAME,
     PDF_EXTRACT_IMAGES,
     YOUTUBE_LOADER_LANGUAGE,
     YOUTUBE_LOADER_PROXY_URL,
     # Retrieval (Web Search)
-    RAG_WEB_SEARCH_ENGINE,
-    RAG_WEB_SEARCH_RESULT_COUNT,
-    RAG_WEB_SEARCH_CONCURRENT_REQUESTS,
-    RAG_WEB_SEARCH_DOMAIN_FILTER_LIST,
+    ENABLE_WEB_SEARCH,
+    WEB_SEARCH_ENGINE,
+    BYPASS_WEB_SEARCH_EMBEDDING_AND_RETRIEVAL,
+    WEB_SEARCH_RESULT_COUNT,
+    WEB_SEARCH_CONCURRENT_REQUESTS,
+    WEB_SEARCH_TRUST_ENV,
+    WEB_SEARCH_DOMAIN_FILTER_LIST,
     JINA_API_KEY,
     SEARCHAPI_API_KEY,
     SEARCHAPI_ENGINE,
+    SERPAPI_API_KEY,
+    SERPAPI_ENGINE,
     SEARXNG_QUERY_URL,
     SERPER_API_KEY,
     SERPLY_API_KEY,
     SERPSTACK_API_KEY,
     SERPSTACK_HTTPS,
     TAVILY_API_KEY,
+    TAVILY_EXTRACT_DEPTH,
     BING_SEARCH_V7_ENDPOINT,
     BING_SEARCH_V7_SUBSCRIPTION_KEY,
     BRAVE_SEARCH_API_KEY,
+    EXA_API_KEY,
+    PERPLEXITY_API_KEY,
+    SOUGOU_API_SID,
+    SOUGOU_API_SK,
     KAGI_SEARCH_API_KEY,
     MOJEEK_SEARCH_API_KEY,
+    BOCHA_SEARCH_API_KEY,
     GOOGLE_PSE_API_KEY,
     GOOGLE_PSE_ENGINE_ID,
     GOOGLE_DRIVE_CLIENT_ID,
     GOOGLE_DRIVE_API_KEY,
+    ONEDRIVE_CLIENT_ID,
     ENABLE_RAG_HYBRID_SEARCH,
     ENABLE_RAG_LOCAL_WEB_FETCH,
-    ENABLE_RAG_WEB_LOADER_SSL_VERIFICATION,
-    ENABLE_RAG_WEB_SEARCH,
+    ENABLE_WEB_LOADER_SSL_VERIFICATION,
     ENABLE_GOOGLE_DRIVE_INTEGRATION,
+    ENABLE_ONEDRIVE_INTEGRATION,
     UPLOAD_DIR,
     # WebUI
     WEBUI_AUTH,
@@ -205,6 +261,7 @@ from open_webui.config import (
     ENABLE_CHANNELS,
     ENABLE_COMMUNITY_SHARING,
     ENABLE_MESSAGE_RATING,
+    ENABLE_USER_WEBHOOKS,
     ENABLE_EVALUATION_ARENA_MODELS,
     USER_PERMISSIONS,
     DEFAULT_USER_ROLE,
@@ -251,6 +308,7 @@ from open_webui.config import (
     TASK_MODEL,
     TASK_MODEL_EXTERNAL,
     ENABLE_TAGS_GENERATION,
+    ENABLE_TITLE_GENERATION,
     ENABLE_SEARCH_QUERY_GENERATION,
     ENABLE_RETRIEVAL_QUERY_GENERATION,
     ENABLE_AUTOCOMPLETE_GENERATION,
@@ -265,8 +323,14 @@ from open_webui.config import (
     reset_config,
 )
 from open_webui.env import (
+    AUDIT_EXCLUDED_PATHS,
+    AUDIT_LOG_LEVEL,
     CHANGELOG,
+    REDIS_URL,
+    REDIS_SENTINEL_HOSTS,
+    REDIS_SENTINEL_PORT,
     GLOBAL_LOG_LEVEL,
+    MAX_BODY_LOG_SIZE,
     SAFE_MODE,
     SRC_LOG_LEVELS,
     VERSION,
@@ -280,6 +344,8 @@ from open_webui.env import (
     BYPASS_MODEL_ACCESS_CONTROL,
     RESET_CONFIG_ON_START,
     OFFLINE_MODE,
+    ENABLE_OTEL,
+    EXTERNAL_PWA_MANIFEST_URL,
 )
 
 
@@ -297,14 +363,23 @@ from open_webui.utils.middleware import process_chat_payload, process_chat_respo
 from open_webui.utils.access_control import has_access
 
 from open_webui.utils.auth import (
+    get_license_data,
+    get_http_authorization_cred,
     decode_token,
     get_admin_user,
     get_verified_user,
 )
-from open_webui.utils.oauth import oauth_manager
+from open_webui.utils.oauth import OAuthManager
 from open_webui.utils.security_headers import SecurityHeadersMiddleware
 
-from open_webui.tasks import stop_task, list_tasks  # Import from tasks.py
+from open_webui.tasks import (
+    list_task_ids_by_chat_id,
+    stop_task,
+    list_tasks,
+)  # Import from tasks.py
+
+from open_webui.utils.redis import get_sentinels_from_env
+
 
 if SAFE_MODE:
     print("SAFE MODE ENABLED")
@@ -321,19 +396,23 @@ class SPAStaticFiles(StaticFiles):
             return await super().get_response(path, scope)
         except (HTTPException, StarletteHTTPException) as ex:
             if ex.status_code == 404:
-                return await super().get_response("index.html", scope)
+                if path.endswith(".js"):
+                    # Return 404 for javascript files
+                    raise ex
+                else:
+                    return await super().get_response("index.html", scope)
             else:
                 raise ex
 
 
 print(
     rf"""
-  ___                    __        __   _     _   _ ___
- / _ \ _ __   ___ _ __   \ \      / /__| |__ | | | |_ _|
-| | | | '_ \ / _ \ '_ \   \ \ /\ / / _ \ '_ \| | | || |
-| |_| | |_) |  __/ | | |   \ V  V /  __/ |_) | |_| || |
- \___/| .__/ \___|_| |_|    \_/\_/ \___|_.__/ \___/|___|
-      |_|
+ ██████╗ ██████╗ ███████╗███╗   ██╗    ██╗    ██╗███████╗██████╗ ██╗   ██╗██╗
+██╔═══██╗██╔══██╗██╔════╝████╗  ██║    ██║    ██║██╔════╝██╔══██╗██║   ██║██║
+██║   ██║██████╔╝█████╗  ██╔██╗ ██║    ██║ █╗ ██║█████╗  ██████╔╝██║   ██║██║
+██║   ██║██╔═══╝ ██╔══╝  ██║╚██╗██║    ██║███╗██║██╔══╝  ██╔══██╗██║   ██║██║
+╚██████╔╝██║     ███████╗██║ ╚████║    ╚███╔███╔╝███████╗██████╔╝╚██████╔╝██║
+ ╚═════╝ ╚═╝     ╚══════╝╚═╝  ╚═══╝     ╚══╝╚══╝ ╚══════╝╚═════╝  ╚═════╝ ╚═╝
 
 
 v{VERSION} - building the best open-source AI user interface.
@@ -345,21 +424,46 @@ https://github.com/open-webui/open-webui
 
 @asynccontextmanager
 async def lifespan(app: FastAPI):
+    start_logger()
     if RESET_CONFIG_ON_START:
         reset_config()
 
+    if LICENSE_KEY:
+        get_license_data(app, LICENSE_KEY)
+
     asyncio.create_task(periodic_usage_pool_cleanup())
     yield
 
 
 app = FastAPI(
+    title="Open WebUI",
     docs_url="/docs" if ENV == "dev" else None,
     openapi_url="/openapi.json" if ENV == "dev" else None,
     redoc_url=None,
     lifespan=lifespan,
 )
 
-app.state.config = AppConfig()
+oauth_manager = OAuthManager(app)
+
+app.state.config = AppConfig(
+    redis_url=REDIS_URL,
+    redis_sentinels=get_sentinels_from_env(REDIS_SENTINEL_HOSTS, REDIS_SENTINEL_PORT),
+)
+
+app.state.WEBUI_NAME = WEBUI_NAME
+app.state.LICENSE_METADATA = None
+
+
+########################################
+#
+# OPENTELEMETRY
+#
+########################################
+
+if ENABLE_OTEL:
+    from open_webui.utils.telemetry.setup import setup as setup_opentelemetry
+
+    setup_opentelemetry(app=app, db_engine=engine)
 
 
 ########################################
@@ -388,6 +492,23 @@ app.state.config.OPENAI_API_CONFIGS = OPENAI_API_CONFIGS
 
 app.state.OPENAI_MODELS = {}
 
+########################################
+#
+# TOOL SERVERS
+#
+########################################
+
+app.state.config.TOOL_SERVER_CONNECTIONS = TOOL_SERVER_CONNECTIONS
+app.state.TOOL_SERVERS = []
+
+########################################
+#
+# DIRECT CONNECTIONS
+#
+########################################
+
+app.state.config.ENABLE_DIRECT_CONNECTIONS = ENABLE_DIRECT_CONNECTIONS
+
 ########################################
 #
 # WEBUI
@@ -423,6 +544,7 @@ app.state.config.MODEL_ORDER_LIST = MODEL_ORDER_LIST
 app.state.config.ENABLE_CHANNELS = ENABLE_CHANNELS
 app.state.config.ENABLE_COMMUNITY_SHARING = ENABLE_COMMUNITY_SHARING
 app.state.config.ENABLE_MESSAGE_RATING = ENABLE_MESSAGE_RATING
+app.state.config.ENABLE_USER_WEBHOOKS = ENABLE_USER_WEBHOOKS
 
 app.state.config.ENABLE_EVALUATION_ARENA_MODELS = ENABLE_EVALUATION_ARENA_MODELS
 app.state.config.EVALUATION_ARENA_MODELS = EVALUATION_ARENA_MODELS
@@ -453,11 +575,12 @@ app.state.config.LDAP_CIPHERS = LDAP_CIPHERS
 
 app.state.AUTH_TRUSTED_EMAIL_HEADER = WEBUI_AUTH_TRUSTED_EMAIL_HEADER
 app.state.AUTH_TRUSTED_NAME_HEADER = WEBUI_AUTH_TRUSTED_NAME_HEADER
+app.state.EXTERNAL_PWA_MANIFEST_URL = EXTERNAL_PWA_MANIFEST_URL
 
+app.state.USER_COUNT = None
 app.state.TOOLS = {}
 app.state.FUNCTIONS = {}
 
-
 ########################################
 #
 # RETRIEVAL
@@ -466,17 +589,23 @@ app.state.FUNCTIONS = {}
 
 
 app.state.config.TOP_K = RAG_TOP_K
+app.state.config.TOP_K_RERANKER = RAG_TOP_K_RERANKER
 app.state.config.RELEVANCE_THRESHOLD = RAG_RELEVANCE_THRESHOLD
 app.state.config.FILE_MAX_SIZE = RAG_FILE_MAX_SIZE
 app.state.config.FILE_MAX_COUNT = RAG_FILE_MAX_COUNT
 
+
+app.state.config.RAG_FULL_CONTEXT = RAG_FULL_CONTEXT
+app.state.config.BYPASS_EMBEDDING_AND_RETRIEVAL = BYPASS_EMBEDDING_AND_RETRIEVAL
 app.state.config.ENABLE_RAG_HYBRID_SEARCH = ENABLE_RAG_HYBRID_SEARCH
-app.state.config.ENABLE_RAG_WEB_LOADER_SSL_VERIFICATION = (
-    ENABLE_RAG_WEB_LOADER_SSL_VERIFICATION
-)
+app.state.config.ENABLE_WEB_LOADER_SSL_VERIFICATION = ENABLE_WEB_LOADER_SSL_VERIFICATION
 
 app.state.config.CONTENT_EXTRACTION_ENGINE = CONTENT_EXTRACTION_ENGINE
 app.state.config.TIKA_SERVER_URL = TIKA_SERVER_URL
+app.state.config.DOCLING_SERVER_URL = DOCLING_SERVER_URL
+app.state.config.DOCUMENT_INTELLIGENCE_ENDPOINT = DOCUMENT_INTELLIGENCE_ENDPOINT
+app.state.config.DOCUMENT_INTELLIGENCE_KEY = DOCUMENT_INTELLIGENCE_KEY
+app.state.config.MISTRAL_OCR_API_KEY = MISTRAL_OCR_API_KEY
 
 app.state.config.TEXT_SPLITTER = RAG_TEXT_SPLITTER
 app.state.config.TIKTOKEN_ENCODING_NAME = TIKTOKEN_ENCODING_NAME
@@ -502,17 +631,26 @@ app.state.config.YOUTUBE_LOADER_LANGUAGE = YOUTUBE_LOADER_LANGUAGE
 app.state.config.YOUTUBE_LOADER_PROXY_URL = YOUTUBE_LOADER_PROXY_URL
 
 
-app.state.config.ENABLE_RAG_WEB_SEARCH = ENABLE_RAG_WEB_SEARCH
-app.state.config.RAG_WEB_SEARCH_ENGINE = RAG_WEB_SEARCH_ENGINE
-app.state.config.RAG_WEB_SEARCH_DOMAIN_FILTER_LIST = RAG_WEB_SEARCH_DOMAIN_FILTER_LIST
+app.state.config.ENABLE_WEB_SEARCH = ENABLE_WEB_SEARCH
+app.state.config.WEB_SEARCH_ENGINE = WEB_SEARCH_ENGINE
+app.state.config.WEB_SEARCH_DOMAIN_FILTER_LIST = WEB_SEARCH_DOMAIN_FILTER_LIST
+app.state.config.WEB_SEARCH_RESULT_COUNT = WEB_SEARCH_RESULT_COUNT
+app.state.config.WEB_SEARCH_CONCURRENT_REQUESTS = WEB_SEARCH_CONCURRENT_REQUESTS
+app.state.config.WEB_LOADER_ENGINE = WEB_LOADER_ENGINE
+app.state.config.WEB_SEARCH_TRUST_ENV = WEB_SEARCH_TRUST_ENV
+app.state.config.BYPASS_WEB_SEARCH_EMBEDDING_AND_RETRIEVAL = (
+    BYPASS_WEB_SEARCH_EMBEDDING_AND_RETRIEVAL
+)
 
 app.state.config.ENABLE_GOOGLE_DRIVE_INTEGRATION = ENABLE_GOOGLE_DRIVE_INTEGRATION
+app.state.config.ENABLE_ONEDRIVE_INTEGRATION = ENABLE_ONEDRIVE_INTEGRATION
 app.state.config.SEARXNG_QUERY_URL = SEARXNG_QUERY_URL
 app.state.config.GOOGLE_PSE_API_KEY = GOOGLE_PSE_API_KEY
 app.state.config.GOOGLE_PSE_ENGINE_ID = GOOGLE_PSE_ENGINE_ID
 app.state.config.BRAVE_SEARCH_API_KEY = BRAVE_SEARCH_API_KEY
 app.state.config.KAGI_SEARCH_API_KEY = KAGI_SEARCH_API_KEY
 app.state.config.MOJEEK_SEARCH_API_KEY = MOJEEK_SEARCH_API_KEY
+app.state.config.BOCHA_SEARCH_API_KEY = BOCHA_SEARCH_API_KEY
 app.state.config.SERPSTACK_API_KEY = SERPSTACK_API_KEY
 app.state.config.SERPSTACK_HTTPS = SERPSTACK_HTTPS
 app.state.config.SERPER_API_KEY = SERPER_API_KEY
@@ -520,12 +658,22 @@ app.state.config.SERPLY_API_KEY = SERPLY_API_KEY
 app.state.config.TAVILY_API_KEY = TAVILY_API_KEY
 app.state.config.SEARCHAPI_API_KEY = SEARCHAPI_API_KEY
 app.state.config.SEARCHAPI_ENGINE = SEARCHAPI_ENGINE
+app.state.config.SERPAPI_API_KEY = SERPAPI_API_KEY
+app.state.config.SERPAPI_ENGINE = SERPAPI_ENGINE
 app.state.config.JINA_API_KEY = JINA_API_KEY
 app.state.config.BING_SEARCH_V7_ENDPOINT = BING_SEARCH_V7_ENDPOINT
 app.state.config.BING_SEARCH_V7_SUBSCRIPTION_KEY = BING_SEARCH_V7_SUBSCRIPTION_KEY
+app.state.config.EXA_API_KEY = EXA_API_KEY
+app.state.config.PERPLEXITY_API_KEY = PERPLEXITY_API_KEY
+app.state.config.SOUGOU_API_SID = SOUGOU_API_SID
+app.state.config.SOUGOU_API_SK = SOUGOU_API_SK
 
-app.state.config.RAG_WEB_SEARCH_RESULT_COUNT = RAG_WEB_SEARCH_RESULT_COUNT
-app.state.config.RAG_WEB_SEARCH_CONCURRENT_REQUESTS = RAG_WEB_SEARCH_CONCURRENT_REQUESTS
+
+app.state.config.PLAYWRIGHT_WS_URL = PLAYWRIGHT_WS_URL
+app.state.config.PLAYWRIGHT_TIMEOUT = PLAYWRIGHT_TIMEOUT
+app.state.config.FIRECRAWL_API_BASE_URL = FIRECRAWL_API_BASE_URL
+app.state.config.FIRECRAWL_API_KEY = FIRECRAWL_API_KEY
+app.state.config.TAVILY_EXTRACT_DEPTH = TAVILY_EXTRACT_DEPTH
 
 app.state.EMBEDDING_FUNCTION = None
 app.state.ef = None
@@ -567,6 +715,35 @@ app.state.EMBEDDING_FUNCTION = get_embedding_function(
     app.state.config.RAG_EMBEDDING_BATCH_SIZE,
 )
 
+########################################
+#
+# CODE EXECUTION
+#
+########################################
+
+app.state.config.ENABLE_CODE_EXECUTION = ENABLE_CODE_EXECUTION
+app.state.config.CODE_EXECUTION_ENGINE = CODE_EXECUTION_ENGINE
+app.state.config.CODE_EXECUTION_JUPYTER_URL = CODE_EXECUTION_JUPYTER_URL
+app.state.config.CODE_EXECUTION_JUPYTER_AUTH = CODE_EXECUTION_JUPYTER_AUTH
+app.state.config.CODE_EXECUTION_JUPYTER_AUTH_TOKEN = CODE_EXECUTION_JUPYTER_AUTH_TOKEN
+app.state.config.CODE_EXECUTION_JUPYTER_AUTH_PASSWORD = (
+    CODE_EXECUTION_JUPYTER_AUTH_PASSWORD
+)
+app.state.config.CODE_EXECUTION_JUPYTER_TIMEOUT = CODE_EXECUTION_JUPYTER_TIMEOUT
+
+app.state.config.ENABLE_CODE_INTERPRETER = ENABLE_CODE_INTERPRETER
+app.state.config.CODE_INTERPRETER_ENGINE = CODE_INTERPRETER_ENGINE
+app.state.config.CODE_INTERPRETER_PROMPT_TEMPLATE = CODE_INTERPRETER_PROMPT_TEMPLATE
+
+app.state.config.CODE_INTERPRETER_JUPYTER_URL = CODE_INTERPRETER_JUPYTER_URL
+app.state.config.CODE_INTERPRETER_JUPYTER_AUTH = CODE_INTERPRETER_JUPYTER_AUTH
+app.state.config.CODE_INTERPRETER_JUPYTER_AUTH_TOKEN = (
+    CODE_INTERPRETER_JUPYTER_AUTH_TOKEN
+)
+app.state.config.CODE_INTERPRETER_JUPYTER_AUTH_PASSWORD = (
+    CODE_INTERPRETER_JUPYTER_AUTH_PASSWORD
+)
+app.state.config.CODE_INTERPRETER_JUPYTER_TIMEOUT = CODE_INTERPRETER_JUPYTER_TIMEOUT
 
 ########################################
 #
@@ -581,6 +758,9 @@ app.state.config.ENABLE_IMAGE_PROMPT_GENERATION = ENABLE_IMAGE_PROMPT_GENERATION
 app.state.config.IMAGES_OPENAI_API_BASE_URL = IMAGES_OPENAI_API_BASE_URL
 app.state.config.IMAGES_OPENAI_API_KEY = IMAGES_OPENAI_API_KEY
 
+app.state.config.IMAGES_GEMINI_API_BASE_URL = IMAGES_GEMINI_API_BASE_URL
+app.state.config.IMAGES_GEMINI_API_KEY = IMAGES_GEMINI_API_KEY
+
 app.state.config.IMAGE_GENERATION_MODEL = IMAGE_GENERATION_MODEL
 
 app.state.config.AUTOMATIC1111_BASE_URL = AUTOMATIC1111_BASE_URL
@@ -609,6 +789,11 @@ app.state.config.STT_ENGINE = AUDIO_STT_ENGINE
 app.state.config.STT_MODEL = AUDIO_STT_MODEL
 
 app.state.config.WHISPER_MODEL = WHISPER_MODEL
+app.state.config.DEEPGRAM_API_KEY = DEEPGRAM_API_KEY
+
+app.state.config.AUDIO_STT_AZURE_API_KEY = AUDIO_STT_AZURE_API_KEY
+app.state.config.AUDIO_STT_AZURE_REGION = AUDIO_STT_AZURE_REGION
+app.state.config.AUDIO_STT_AZURE_LOCALES = AUDIO_STT_AZURE_LOCALES
 
 app.state.config.TTS_OPENAI_API_BASE_URL = AUDIO_TTS_OPENAI_API_BASE_URL
 app.state.config.TTS_OPENAI_API_KEY = AUDIO_TTS_OPENAI_API_KEY
@@ -643,6 +828,7 @@ app.state.config.ENABLE_SEARCH_QUERY_GENERATION = ENABLE_SEARCH_QUERY_GENERATION
 app.state.config.ENABLE_RETRIEVAL_QUERY_GENERATION = ENABLE_RETRIEVAL_QUERY_GENERATION
 app.state.config.ENABLE_AUTOCOMPLETE_GENERATION = ENABLE_AUTOCOMPLETE_GENERATION
 app.state.config.ENABLE_TAGS_GENERATION = ENABLE_TAGS_GENERATION
+app.state.config.ENABLE_TITLE_GENERATION = ENABLE_TITLE_GENERATION
 
 
 app.state.config.TITLE_GENERATION_PROMPT_TEMPLATE = TITLE_GENERATION_PROMPT_TEMPLATE
@@ -707,6 +893,10 @@ async def commit_session_after_request(request: Request, call_next):
 @app.middleware("http")
 async def check_url(request: Request, call_next):
     start_time = int(time.time())
+    request.state.token = get_http_authorization_cred(
+        request.headers.get("Authorization")
+    )
+
     request.state.enable_api_key = app.state.config.ENABLE_API_KEY
     response = await call_next(request)
     process_time = int(time.time()) - start_time
@@ -751,6 +941,7 @@ app.include_router(openai.router, prefix="/openai", tags=["openai"])
 app.include_router(pipelines.router, prefix="/api/v1/pipelines", tags=["pipelines"])
 app.include_router(tasks.router, prefix="/api/v1/tasks", tags=["tasks"])
 app.include_router(images.router, prefix="/api/v1/images", tags=["images"])
+
 app.include_router(audio.router, prefix="/api/v1/audio", tags=["audio"])
 app.include_router(retrieval.router, prefix="/api/v1/retrieval", tags=["retrieval"])
 
@@ -779,6 +970,19 @@ app.include_router(
 app.include_router(utils.router, prefix="/api/v1/utils", tags=["utils"])
 
 
+try:
+    audit_level = AuditLevel(AUDIT_LOG_LEVEL)
+except ValueError as e:
+    logger.error(f"Invalid audit level: {AUDIT_LOG_LEVEL}. Error: {e}")
+    audit_level = AuditLevel.NONE
+
+if audit_level != AuditLevel.NONE:
+    app.add_middleware(
+        AuditLoggingMiddleware,
+        audit_level=audit_level,
+        excluded_paths=AUDIT_EXCLUDED_PATHS,
+        max_body_size=MAX_BODY_LOG_SIZE,
+    )
 ##################################
 #
 # Chat Endpoints
@@ -811,14 +1015,24 @@ async def get_models(request: Request, user=Depends(get_verified_user)):
 
         return filtered_models
 
-    models = await get_all_models(request)
+    all_models = await get_all_models(request, user=user)
 
-    # Filter out filter pipelines
-    models = [
-        model
-        for model in models
-        if "pipeline" not in model or model["pipeline"].get("type", None) != "filter"
-    ]
+    models = []
+    for model in all_models:
+        # Filter out filter pipelines
+        if "pipeline" in model and model["pipeline"].get("type", None) == "filter":
+            continue
+
+        model_tags = [
+            tag.get("name")
+            for tag in model.get("info", {}).get("meta", {}).get("tags", [])
+        ]
+        tags = [tag.get("name") for tag in model.get("tags", [])]
+
+        tags = list(set(model_tags + tags))
+        model["tags"] = [{"name": tag} for tag in tags]
+
+        models.append(model)
 
     model_order_list = request.app.state.config.MODEL_ORDER_LIST
     if model_order_list:
@@ -840,7 +1054,7 @@ async def get_models(request: Request, user=Depends(get_verified_user)):
 
 @app.get("/api/models/base")
 async def get_base_models(request: Request, user=Depends(get_admin_user)):
-    models = await get_all_base_models(request)
+    models = await get_all_base_models(request, user=user)
     return {"data": models}
 
 
@@ -851,21 +1065,33 @@ async def chat_completion(
     user=Depends(get_verified_user),
 ):
     if not request.app.state.MODELS:
-        await get_all_models(request)
+        await get_all_models(request, user=user)
 
+    model_item = form_data.pop("model_item", {})
     tasks = form_data.pop("background_tasks", None)
-    try:
-        model_id = form_data.get("model", None)
-        if model_id not in request.app.state.MODELS:
-            raise Exception("Model not found")
-        model = request.app.state.MODELS[model_id]
 
-        # Check if user has access to the model
-        if not BYPASS_MODEL_ACCESS_CONTROL and user.role == "user":
-            try:
-                check_model_access(user, model)
-            except Exception as e:
-                raise e
+    metadata = {}
+    try:
+        if not model_item.get("direct", False):
+            model_id = form_data.get("model", None)
+            if model_id not in request.app.state.MODELS:
+                raise Exception("Model not found")
+
+            model = request.app.state.MODELS[model_id]
+            model_info = Models.get_model_by_id(model_id)
+
+            # Check if user has access to the model
+            if not BYPASS_MODEL_ACCESS_CONTROL and user.role == "user":
+                try:
+                    check_model_access(user, model)
+                except Exception as e:
+                    raise e
+        else:
+            model = model_item
+            model_info = None
+
+            request.state.direct = True
+            request.state.model = model
 
         metadata = {
             "user_id": user.id,
@@ -873,15 +1099,43 @@ async def chat_completion(
             "message_id": form_data.pop("id", None),
             "session_id": form_data.pop("session_id", None),
             "tool_ids": form_data.get("tool_ids", None),
+            "tool_servers": form_data.pop("tool_servers", None),
             "files": form_data.get("files", None),
             "features": form_data.get("features", None),
+            "variables": form_data.get("variables", None),
+            "model": model,
+            "direct": model_item.get("direct", False),
+            **(
+                {"function_calling": "native"}
+                if form_data.get("params", {}).get("function_calling") == "native"
+                or (
+                    model_info
+                    and model_info.params.model_dump().get("function_calling")
+                    == "native"
+                )
+                else {}
+            ),
         }
+
+        request.state.metadata = metadata
         form_data["metadata"] = metadata
 
-        form_data, events = await process_chat_payload(
-            request, form_data, metadata, user, model
+        form_data, metadata, events = await process_chat_payload(
+            request, form_data, user, metadata, model
         )
+
     except Exception as e:
+        log.debug(f"Error processing chat payload: {e}")
+        if metadata.get("chat_id") and metadata.get("message_id"):
+            # Update the chat message with the error
+            Chats.upsert_message_to_chat_by_id_and_message_id(
+                metadata["chat_id"],
+                metadata["message_id"],
+                {
+                    "error": {"content": str(e)},
+                },
+            )
+
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=str(e),
@@ -889,8 +1143,9 @@ async def chat_completion(
 
     try:
         response = await chat_completion_handler(request, form_data, user)
+
         return await process_chat_response(
-            request, response, form_data, user, events, metadata, tasks
+            request, response, form_data, user, metadata, model, events, tasks
         )
     except Exception as e:
         raise HTTPException(
@@ -909,6 +1164,12 @@ async def chat_completed(
     request: Request, form_data: dict, user=Depends(get_verified_user)
 ):
     try:
+        model_item = form_data.pop("model_item", {})
+
+        if model_item.get("direct", False):
+            request.state.direct = True
+            request.state.model = model_item
+
         return await chat_completed_handler(request, form_data, user)
     except Exception as e:
         raise HTTPException(
@@ -922,6 +1183,12 @@ async def chat_action(
     request: Request, action_id: str, form_data: dict, user=Depends(get_verified_user)
 ):
     try:
+        model_item = form_data.pop("model_item", {})
+
+        if model_item.get("direct", False):
+            request.state.direct = True
+            request.state.model = model_item
+
         return await chat_action_handler(request, action_id, form_data, user)
     except Exception as e:
         raise HTTPException(
@@ -933,7 +1200,7 @@ async def chat_action(
 @app.post("/api/tasks/stop/{task_id}")
 async def stop_task_endpoint(task_id: str, user=Depends(get_verified_user)):
     try:
-        result = await stop_task(task_id)  # Use the function from tasks.py
+        result = await stop_task(task_id)
         return result
     except ValueError as e:
         raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail=str(e))
@@ -941,7 +1208,19 @@ async def stop_task_endpoint(task_id: str, user=Depends(get_verified_user)):
 
 @app.get("/api/tasks")
 async def list_tasks_endpoint(user=Depends(get_verified_user)):
-    return {"tasks": list_tasks()}  # Use the function from tasks.py
+    return {"tasks": list_tasks()}
+
+
+@app.get("/api/tasks/chat/{chat_id}")
+async def list_tasks_by_chat_id_endpoint(chat_id: str, user=Depends(get_verified_user)):
+    chat = Chats.get_chat_by_id(chat_id)
+    if chat is None or chat.user_id != user.id:
+        return {"task_ids": []}
+
+    task_ids = list_task_ids_by_chat_id(chat_id)
+
+    print(f"Task IDs for chat {chat_id}: {task_ids}")
+    return {"task_ids": task_ids}
 
 
 ##################################
@@ -967,15 +1246,16 @@ async def get_app_config(request: Request):
         if data is not None and "id" in data:
             user = Users.get_user_by_id(data["id"])
 
+    user_count = Users.get_num_users()
     onboarding = False
+
     if user is None:
-        user_count = Users.get_num_users()
         onboarding = user_count == 0
 
     return {
         **({"onboarding": True} if onboarding else {}),
         "status": True,
-        "name": WEBUI_NAME,
+        "name": app.state.WEBUI_NAME,
         "version": VERSION,
         "default_locale": str(DEFAULT_LOCALE),
         "oauth": {
@@ -994,27 +1274,33 @@ async def get_app_config(request: Request):
             "enable_websocket": ENABLE_WEBSOCKET_SUPPORT,
             **(
                 {
+                    "enable_direct_connections": app.state.config.ENABLE_DIRECT_CONNECTIONS,
                     "enable_channels": app.state.config.ENABLE_CHANNELS,
-                    "enable_web_search": app.state.config.ENABLE_RAG_WEB_SEARCH,
-                    "enable_google_drive_integration": app.state.config.ENABLE_GOOGLE_DRIVE_INTEGRATION,
+                    "enable_web_search": app.state.config.ENABLE_WEB_SEARCH,
+                    "enable_code_execution": app.state.config.ENABLE_CODE_EXECUTION,
+                    "enable_code_interpreter": app.state.config.ENABLE_CODE_INTERPRETER,
                     "enable_image_generation": app.state.config.ENABLE_IMAGE_GENERATION,
+                    "enable_autocomplete_generation": app.state.config.ENABLE_AUTOCOMPLETE_GENERATION,
                     "enable_community_sharing": app.state.config.ENABLE_COMMUNITY_SHARING,
                     "enable_message_rating": app.state.config.ENABLE_MESSAGE_RATING,
+                    "enable_user_webhooks": app.state.config.ENABLE_USER_WEBHOOKS,
                     "enable_admin_export": ENABLE_ADMIN_EXPORT,
                     "enable_admin_chat_access": ENABLE_ADMIN_CHAT_ACCESS,
+                    "enable_google_drive_integration": app.state.config.ENABLE_GOOGLE_DRIVE_INTEGRATION,
+                    "enable_onedrive_integration": app.state.config.ENABLE_ONEDRIVE_INTEGRATION,
                 }
                 if user is not None
                 else {}
             ),
         },
-        "google_drive": {
-            "client_id": GOOGLE_DRIVE_CLIENT_ID.value,
-            "api_key": GOOGLE_DRIVE_API_KEY.value,
-        },
         **(
             {
                 "default_models": app.state.config.DEFAULT_MODELS,
                 "default_prompt_suggestions": app.state.config.DEFAULT_PROMPT_SUGGESTIONS,
+                "user_count": user_count,
+                "code": {
+                    "engine": app.state.config.CODE_EXECUTION_ENGINE,
+                },
                 "audio": {
                     "tts": {
                         "engine": app.state.config.TTS_ENGINE,
@@ -1030,6 +1316,19 @@ async def get_app_config(request: Request):
                     "max_count": app.state.config.FILE_MAX_COUNT,
                 },
                 "permissions": {**app.state.config.USER_PERMISSIONS},
+                "google_drive": {
+                    "client_id": GOOGLE_DRIVE_CLIENT_ID.value,
+                    "api_key": GOOGLE_DRIVE_API_KEY.value,
+                },
+                "onedrive": {"client_id": ONEDRIVE_CLIENT_ID.value},
+                "license_metadata": app.state.LICENSE_METADATA,
+                **(
+                    {
+                        "active_entries": app.state.USER_COUNT,
+                    }
+                    if user.role == "admin"
+                    else {}
+                ),
             }
             if user is not None
             else {}
@@ -1063,7 +1362,7 @@ async def get_app_version():
 
 
 @app.get("/api/version/updates")
-async def get_app_latest_release_version():
+async def get_app_latest_release_version(user=Depends(get_verified_user)):
     if OFFLINE_MODE:
         log.debug(
             f"Offline mode is enabled, returning current version as latest version"
@@ -1107,7 +1406,7 @@ if len(OAUTH_PROVIDERS) > 0:
 
 @app.get("/oauth/{provider}/login")
 async def oauth_login(provider: str, request: Request):
-    return await oauth_manager.handle_login(provider, request)
+    return await oauth_manager.handle_login(request, provider)
 
 
 # OAuth login logic is as follows:
@@ -1118,42 +1417,45 @@ async def oauth_login(provider: str, request: Request):
 #    - Email addresses are considered unique, so we fail registration if the email address is already taken
 @app.get("/oauth/{provider}/callback")
 async def oauth_callback(provider: str, request: Request, response: Response):
-    return await oauth_manager.handle_callback(provider, request, response)
+    return await oauth_manager.handle_callback(request, provider, response)
 
 
 @app.get("/manifest.json")
 async def get_manifest_json():
-    return {
-        "name": WEBUI_NAME,
-        "short_name": WEBUI_NAME,
-        "description": "Open WebUI is an open, extensible, user-friendly interface for AI that adapts to your workflow.",
-        "start_url": "/",
-        "display": "standalone",
-        "background_color": "#343541",
-        "orientation": "natural",
-        "icons": [
-            {
-                "src": "/static/logo.png",
-                "type": "image/png",
-                "sizes": "500x500",
-                "purpose": "any",
-            },
-            {
-                "src": "/static/logo.png",
-                "type": "image/png",
-                "sizes": "500x500",
-                "purpose": "maskable",
-            },
-        ],
-    }
+    if app.state.EXTERNAL_PWA_MANIFEST_URL:
+        return requests.get(app.state.EXTERNAL_PWA_MANIFEST_URL).json()
+    else:
+        return {
+            "name": app.state.WEBUI_NAME,
+            "short_name": app.state.WEBUI_NAME,
+            "description": "Open WebUI is an open, extensible, user-friendly interface for AI that adapts to your workflow.",
+            "start_url": "/",
+            "display": "standalone",
+            "background_color": "#343541",
+            "orientation": "natural",
+            "icons": [
+                {
+                    "src": "/static/logo.png",
+                    "type": "image/png",
+                    "sizes": "500x500",
+                    "purpose": "any",
+                },
+                {
+                    "src": "/static/logo.png",
+                    "type": "image/png",
+                    "sizes": "500x500",
+                    "purpose": "maskable",
+                },
+            ],
+        }
 
 
 @app.get("/opensearch.xml")
 async def get_opensearch_xml():
     xml_content = rf"""
     <OpenSearchDescription xmlns="http://a9.com/-/spec/opensearch/1.1/" xmlns:moz="http://www.mozilla.org/2006/browser/search/">
-    <ShortName>{WEBUI_NAME}</ShortName>
-    <Description>Search {WEBUI_NAME}</Description>
+    <ShortName>{app.state.WEBUI_NAME}</ShortName>
+    <Description>Search {app.state.WEBUI_NAME}</Description>
     <InputEncoding>UTF-8</InputEncoding>
     <Image width="16" height="16" type="image/x-icon">{app.state.config.WEBUI_URL}/static/favicon.png</Image>
     <Url type="text/html" method="get" template="{app.state.config.WEBUI_URL}/?q={"{searchTerms}"}"/>

backend/open_webui/models/chats.py

@@ -1,3 +1,4 @@
+import logging
 import json
 import time
 import uuid
@@ -5,7 +6,7 @@ from typing import Optional
 
 from open_webui.internal.db import Base, get_db
 from open_webui.models.tags import TagModel, Tag, Tags
-
+from open_webui.env import SRC_LOG_LEVELS
 
 from pydantic import BaseModel, ConfigDict
 from sqlalchemy import BigInteger, Boolean, Column, String, Text, JSON
@@ -16,6 +17,9 @@ from sqlalchemy.sql import exists
 # Chat DB Schema
 ####################
 
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["MODELS"])
+
 
 class Chat(Base):
     __tablename__ = "chat"
@@ -470,7 +474,7 @@ class ChatTable:
         try:
             with get_db() as db:
                 # it is possible that the shared link was deleted. hence,
-                # we check if the chat is still shared by checkng if a chat with the share_id exists
+                # we check if the chat is still shared by checking if a chat with the share_id exists
                 chat = db.query(Chat).filter_by(share_id=id).first()
 
                 if chat:
@@ -670,7 +674,7 @@ class ChatTable:
             # Perform pagination at the SQL level
             all_chats = query.offset(skip).limit(limit).all()
 
-            print(len(all_chats))
+            log.info(f"The number of chats: {len(all_chats)}")
 
             # Validate and return chats
             return [ChatModel.model_validate(chat) for chat in all_chats]
@@ -731,7 +735,7 @@ class ChatTable:
             query = db.query(Chat).filter_by(user_id=user_id)
             tag_id = tag_name.replace(" ", "_").lower()
 
-            print(db.bind.dialect.name)
+            log.info(f"DB dialect name: {db.bind.dialect.name}")
             if db.bind.dialect.name == "sqlite":
                 # SQLite JSON1 querying for tags within the meta JSON field
                 query = query.filter(
@@ -752,7 +756,7 @@ class ChatTable:
                 )
 
             all_chats = query.all()
-            print("all_chats", all_chats)
+            log.debug(f"all_chats: {all_chats}")
             return [ChatModel.model_validate(chat) for chat in all_chats]
 
     def add_chat_tag_by_id_and_user_id_and_tag_name(
@@ -810,7 +814,7 @@ class ChatTable:
             count = query.count()
 
             # Debugging output for inspection
-            print(f"Count of chats for tag '{tag_name}':", count)
+            log.info(f"Count of chats for tag '{tag_name}': {count}")
 
             return count
 

backend/open_webui/models/feedbacks.py

@@ -118,7 +118,7 @@ class FeedbackTable:
                 else:
                     return None
             except Exception as e:
-                print(e)
+                log.exception(f"Error creating a new feedback: {e}")
                 return None
 
     def get_feedback_by_id(self, id: str) -> Optional[FeedbackModel]:

backend/open_webui/models/files.py

@@ -119,7 +119,7 @@ class FilesTable:
                 else:
                     return None
             except Exception as e:
-                print(f"Error creating tool: {e}")
+                log.exception(f"Error inserting a new file: {e}")
                 return None
 
     def get_file_by_id(self, id: str) -> Optional[FileModel]:

backend/open_webui/models/folders.py

@@ -9,6 +9,8 @@ from open_webui.models.chats import Chats
 from open_webui.env import SRC_LOG_LEVELS
 from pydantic import BaseModel, ConfigDict
 from sqlalchemy import BigInteger, Column, Text, JSON, Boolean
+from open_webui.utils.access_control import get_permissions
+
 
 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["MODELS"])
@@ -82,7 +84,7 @@ class FolderTable:
                 else:
                     return None
             except Exception as e:
-                print(e)
+                log.exception(f"Error inserting a new folder: {e}")
                 return None
 
     def get_folder_by_id_and_user_id(
@@ -234,15 +236,18 @@ class FolderTable:
             log.error(f"update_folder: {e}")
             return
 
-    def delete_folder_by_id_and_user_id(self, id: str, user_id: str) -> bool:
+    def delete_folder_by_id_and_user_id(
+        self, id: str, user_id: str, delete_chats=True
+    ) -> bool:
         try:
             with get_db() as db:
                 folder = db.query(Folder).filter_by(id=id, user_id=user_id).first()
                 if not folder:
                     return False
 
-                # Delete all chats in the folder
-                Chats.delete_chats_by_user_id_and_folder_id(user_id, folder.id)
+                if delete_chats:
+                    # Delete all chats in the folder
+                    Chats.delete_chats_by_user_id_and_folder_id(user_id, folder.id)
 
                 # Delete all children folders
                 def delete_children(folder):
@@ -250,9 +255,11 @@ class FolderTable:
                         folder.id, user_id
                     )
                     for folder_child in folder_children:
-                        Chats.delete_chats_by_user_id_and_folder_id(
-                            user_id, folder_child.id
-                        )
+                        if delete_chats:
+                            Chats.delete_chats_by_user_id_and_folder_id(
+                                user_id, folder_child.id
+                            )
+
                         delete_children(folder_child)
 
                         folder = db.query(Folder).filter_by(id=folder_child.id).first()

backend/open_webui/models/functions.py

@@ -105,7 +105,7 @@ class FunctionsTable:
                 else:
                     return None
         except Exception as e:
-            print(f"Error creating tool: {e}")
+            log.exception(f"Error creating a new function: {e}")
             return None
 
     def get_function_by_id(self, id: str) -> Optional[FunctionModel]:
@@ -170,7 +170,7 @@ class FunctionsTable:
                 function = db.get(Function, id)
                 return function.valves if function.valves else {}
             except Exception as e:
-                print(f"An error occurred: {e}")
+                log.exception(f"Error getting function valves by id {id}: {e}")
                 return None
 
     def update_function_valves_by_id(
@@ -202,7 +202,9 @@ class FunctionsTable:
 
             return user_settings["functions"]["valves"].get(id, {})
         except Exception as e:
-            print(f"An error occurred: {e}")
+            log.exception(
+                f"Error getting user values by id {id} and user id {user_id}: {e}"
+            )
             return None
 
     def update_user_valves_by_id_and_user_id(
@@ -225,7 +227,9 @@ class FunctionsTable:
 
             return user_settings["functions"]["valves"][id]
         except Exception as e:
-            print(f"An error occurred: {e}")
+            log.exception(
+                f"Error updating user valves by id {id} and user_id {user_id}: {e}"
+            )
             return None
 
     def update_function_by_id(self, id: str, updated: dict) -> Optional[FunctionModel]:

backend/open_webui/models/memories.py

@@ -63,14 +63,15 @@ class MemoriesTable:
             else:
                 return None
 
-    def update_memory_by_id(
+    def update_memory_by_id_and_user_id(
         self,
         id: str,
+        user_id: str,
         content: str,
     ) -> Optional[MemoryModel]:
         with get_db() as db:
             try:
-                db.query(Memory).filter_by(id=id).update(
+                db.query(Memory).filter_by(id=id, user_id=user_id).update(
                     {"content": content, "updated_at": int(time.time())}
                 )
                 db.commit()

backend/open_webui/models/models.py

@@ -166,7 +166,7 @@ class ModelsTable:
                 else:
                     return None
         except Exception as e:
-            print(e)
+            log.exception(f"Failed to insert a new model: {e}")
             return None
 
     def get_all_models(self) -> list[ModelModel]:
@@ -246,8 +246,7 @@ class ModelsTable:
                 db.refresh(model)
                 return ModelModel.model_validate(model)
         except Exception as e:
-            print(e)
-
+            log.exception(f"Failed to update the model by id {id}: {e}")
             return None
 
     def delete_model_by_id(self, id: str) -> bool:

backend/open_webui/models/tags.py

@@ -61,7 +61,7 @@ class TagTable:
                 else:
                     return None
             except Exception as e:
-                print(e)
+                log.exception(f"Error inserting a new tag: {e}")
                 return None
 
     def get_tag_by_name_and_user_id(

backend/open_webui/models/tools.py

@@ -131,7 +131,7 @@ class ToolsTable:
                 else:
                     return None
             except Exception as e:
-                print(f"Error creating tool: {e}")
+                log.exception(f"Error creating a new tool: {e}")
                 return None
 
     def get_tool_by_id(self, id: str) -> Optional[ToolModel]:
@@ -175,7 +175,7 @@ class ToolsTable:
                 tool = db.get(Tool, id)
                 return tool.valves if tool.valves else {}
         except Exception as e:
-            print(f"An error occurred: {e}")
+            log.exception(f"Error getting tool valves by id {id}: {e}")
             return None
 
     def update_tool_valves_by_id(self, id: str, valves: dict) -> Optional[ToolValves]:
@@ -204,7 +204,9 @@ class ToolsTable:
 
             return user_settings["tools"]["valves"].get(id, {})
         except Exception as e:
-            print(f"An error occurred: {e}")
+            log.exception(
+                f"Error getting user values by id {id} and user_id {user_id}: {e}"
+            )
             return None
 
     def update_user_valves_by_id_and_user_id(
@@ -227,7 +229,9 @@ class ToolsTable:
 
             return user_settings["tools"]["valves"][id]
         except Exception as e:
-            print(f"An error occurred: {e}")
+            log.exception(
+                f"Error updating user valves by id {id} and user_id {user_id}: {e}"
+            )
             return None
 
     def update_tool_by_id(self, id: str, updated: dict) -> Optional[ToolModel]:

backend/open_webui/models/users.py

@@ -271,6 +271,24 @@ class UsersTable:
         except Exception:
             return None
 
+    def update_user_settings_by_id(self, id: str, updated: dict) -> Optional[UserModel]:
+        try:
+            with get_db() as db:
+                user_settings = db.query(User).filter_by(id=id).first().settings
+
+                if user_settings is None:
+                    user_settings = {}
+
+                user_settings.update(updated)
+
+                db.query(User).filter_by(id=id).update({"settings": user_settings})
+                db.commit()
+
+                user = db.query(User).filter_by(id=id).first()
+                return UserModel.model_validate(user)
+        except Exception:
+            return None
+
     def delete_user_by_id(self, id: str) -> bool:
         try:
             # Remove User from Groups

backend/open_webui/retrieval/loaders/main.py

@@ -4,6 +4,7 @@ import ftfy
 import sys
 
 from langchain_community.document_loaders import (
+    AzureAIDocumentIntelligenceLoader,
     BSHTMLLoader,
     CSVLoader,
     Docx2txtLoader,
@@ -19,6 +20,9 @@ from langchain_community.document_loaders import (
     YoutubeLoader,
 )
 from langchain_core.documents import Document
+
+from open_webui.retrieval.loaders.mistral import MistralLoader
+
 from open_webui.env import SRC_LOG_LEVELS, GLOBAL_LOG_LEVEL
 
 logging.basicConfig(stream=sys.stdout, level=GLOBAL_LOG_LEVEL)
@@ -76,6 +80,7 @@ known_source_ext = [
     "jsx",
     "hs",
     "lhs",
+    "json",
 ]
 
 
@@ -103,7 +108,7 @@ class TikaLoader:
 
         if r.ok:
             raw_metadata = r.json()
-            text = raw_metadata.get("X-TIKA:content", "<No text content found>")
+            text = raw_metadata.get("X-TIKA:content", "<No text content found>").strip()
 
             if "Content-Type" in raw_metadata:
                 headers["Content-Type"] = raw_metadata["Content-Type"]
@@ -115,6 +120,52 @@ class TikaLoader:
             raise Exception(f"Error calling Tika: {r.reason}")
 
 
+class DoclingLoader:
+    def __init__(self, url, file_path=None, mime_type=None):
+        self.url = url.rstrip("/")
+        self.file_path = file_path
+        self.mime_type = mime_type
+
+    def load(self) -> list[Document]:
+        with open(self.file_path, "rb") as f:
+            files = {
+                "files": (
+                    self.file_path,
+                    f,
+                    self.mime_type or "application/octet-stream",
+                )
+            }
+
+            params = {
+                "image_export_mode": "placeholder",
+                "table_mode": "accurate",
+            }
+
+            endpoint = f"{self.url}/v1alpha/convert/file"
+            r = requests.post(endpoint, files=files, data=params)
+
+        if r.ok:
+            result = r.json()
+            document_data = result.get("document", {})
+            text = document_data.get("md_content", "<No text content found>")
+
+            metadata = {"Content-Type": self.mime_type} if self.mime_type else {}
+
+            log.debug("Docling extracted text: %s", text)
+
+            return [Document(page_content=text, metadata=metadata)]
+        else:
+            error_msg = f"Error calling Docling API: {r.reason}"
+            if r.text:
+                try:
+                    error_data = r.json()
+                    if "detail" in error_data:
+                        error_msg += f" - {error_data['detail']}"
+                except Exception:
+                    error_msg += f" - {r.text}"
+            raise Exception(f"Error calling Docling: {error_msg}")
+
+
 class Loader:
     def __init__(self, engine: str = "", **kwargs):
         self.engine = engine
@@ -133,13 +184,16 @@ class Loader:
             for doc in docs
         ]
 
+    def _is_text_file(self, file_ext: str, file_content_type: str) -> bool:
+        return file_ext in known_source_ext or (
+            file_content_type and file_content_type.find("text/") >= 0
+        )
+
     def _get_loader(self, filename: str, file_content_type: str, file_path: str):
         file_ext = filename.split(".")[-1].lower()
 
         if self.engine == "tika" and self.kwargs.get("TIKA_SERVER_URL"):
-            if file_ext in known_source_ext or (
-                file_content_type and file_content_type.find("text/") >= 0
-            ):
+            if self._is_text_file(file_ext, file_content_type):
                 loader = TextLoader(file_path, autodetect_encoding=True)
             else:
                 loader = TikaLoader(
@@ -147,13 +201,52 @@ class Loader:
                     file_path=file_path,
                     mime_type=file_content_type,
                 )
+        elif self.engine == "docling" and self.kwargs.get("DOCLING_SERVER_URL"):
+            if self._is_text_file(file_ext, file_content_type):
+                loader = TextLoader(file_path, autodetect_encoding=True)
+            else:
+                loader = DoclingLoader(
+                    url=self.kwargs.get("DOCLING_SERVER_URL"),
+                    file_path=file_path,
+                    mime_type=file_content_type,
+                )
+        elif (
+            self.engine == "document_intelligence"
+            and self.kwargs.get("DOCUMENT_INTELLIGENCE_ENDPOINT") != ""
+            and self.kwargs.get("DOCUMENT_INTELLIGENCE_KEY") != ""
+            and (
+                file_ext in ["pdf", "xls", "xlsx", "docx", "ppt", "pptx"]
+                or file_content_type
+                in [
+                    "application/vnd.ms-excel",
+                    "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
+                    "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
+                    "application/vnd.ms-powerpoint",
+                    "application/vnd.openxmlformats-officedocument.presentationml.presentation",
+                ]
+            )
+        ):
+            loader = AzureAIDocumentIntelligenceLoader(
+                file_path=file_path,
+                api_endpoint=self.kwargs.get("DOCUMENT_INTELLIGENCE_ENDPOINT"),
+                api_key=self.kwargs.get("DOCUMENT_INTELLIGENCE_KEY"),
+            )
+        elif (
+            self.engine == "mistral_ocr"
+            and self.kwargs.get("MISTRAL_OCR_API_KEY") != ""
+            and file_ext
+            in ["pdf"]  # Mistral OCR currently only supports PDF and images
+        ):
+            loader = MistralLoader(
+                api_key=self.kwargs.get("MISTRAL_OCR_API_KEY"), file_path=file_path
+            )
         else:
             if file_ext == "pdf":
                 loader = PyPDFLoader(
                     file_path, extract_images=self.kwargs.get("PDF_EXTRACT_IMAGES")
                 )
             elif file_ext == "csv":
-                loader = CSVLoader(file_path)
+                loader = CSVLoader(file_path, autodetect_encoding=True)
             elif file_ext == "rst":
                 loader = UnstructuredRSTLoader(file_path, mode="elements")
             elif file_ext == "xml":
@@ -182,9 +275,7 @@ class Loader:
                 loader = UnstructuredPowerPointLoader(file_path)
             elif file_ext == "msg":
                 loader = OutlookMessageLoader(file_path)
-            elif file_ext in known_source_ext or (
-                file_content_type and file_content_type.find("text/") >= 0
-            ):
+            elif self._is_text_file(file_ext, file_content_type):
                 loader = TextLoader(file_path, autodetect_encoding=True)
             else:
                 loader = TextLoader(file_path, autodetect_encoding=True)

backend/open_webui/retrieval/loaders/mistral.py

@@ -0,0 +1,225 @@
+import requests
+import logging
+import os
+import sys
+from typing import List, Dict, Any
+
+from langchain_core.documents import Document
+from open_webui.env import SRC_LOG_LEVELS, GLOBAL_LOG_LEVEL
+
+logging.basicConfig(stream=sys.stdout, level=GLOBAL_LOG_LEVEL)
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
+
+class MistralLoader:
+    """
+    Loads documents by processing them through the Mistral OCR API.
+    """
+
+    BASE_API_URL = "https://api.mistral.ai/v1"
+
+    def __init__(self, api_key: str, file_path: str):
+        """
+        Initializes the loader.
+
+        Args:
+            api_key: Your Mistral API key.
+            file_path: The local path to the PDF file to process.
+        """
+        if not api_key:
+            raise ValueError("API key cannot be empty.")
+        if not os.path.exists(file_path):
+            raise FileNotFoundError(f"File not found at {file_path}")
+
+        self.api_key = api_key
+        self.file_path = file_path
+        self.headers = {"Authorization": f"Bearer {self.api_key}"}
+
+    def _handle_response(self, response: requests.Response) -> Dict[str, Any]:
+        """Checks response status and returns JSON content."""
+        try:
+            response.raise_for_status()  # Raises HTTPError for bad responses (4xx or 5xx)
+            # Handle potential empty responses for certain successful requests (e.g., DELETE)
+            if response.status_code == 204 or not response.content:
+                return {}  # Return empty dict if no content
+            return response.json()
+        except requests.exceptions.HTTPError as http_err:
+            log.error(f"HTTP error occurred: {http_err} - Response: {response.text}")
+            raise
+        except requests.exceptions.RequestException as req_err:
+            log.error(f"Request exception occurred: {req_err}")
+            raise
+        except ValueError as json_err:  # Includes JSONDecodeError
+            log.error(f"JSON decode error: {json_err} - Response: {response.text}")
+            raise  # Re-raise after logging
+
+    def _upload_file(self) -> str:
+        """Uploads the file to Mistral for OCR processing."""
+        log.info("Uploading file to Mistral API")
+        url = f"{self.BASE_API_URL}/files"
+        file_name = os.path.basename(self.file_path)
+
+        try:
+            with open(self.file_path, "rb") as f:
+                files = {"file": (file_name, f, "application/pdf")}
+                data = {"purpose": "ocr"}
+
+                upload_headers = self.headers.copy()  # Avoid modifying self.headers
+
+                response = requests.post(
+                    url, headers=upload_headers, files=files, data=data
+                )
+
+            response_data = self._handle_response(response)
+            file_id = response_data.get("id")
+            if not file_id:
+                raise ValueError("File ID not found in upload response.")
+            log.info(f"File uploaded successfully. File ID: {file_id}")
+            return file_id
+        except Exception as e:
+            log.error(f"Failed to upload file: {e}")
+            raise
+
+    def _get_signed_url(self, file_id: str) -> str:
+        """Retrieves a temporary signed URL for the uploaded file."""
+        log.info(f"Getting signed URL for file ID: {file_id}")
+        url = f"{self.BASE_API_URL}/files/{file_id}/url"
+        params = {"expiry": 1}
+        signed_url_headers = {**self.headers, "Accept": "application/json"}
+
+        try:
+            response = requests.get(url, headers=signed_url_headers, params=params)
+            response_data = self._handle_response(response)
+            signed_url = response_data.get("url")
+            if not signed_url:
+                raise ValueError("Signed URL not found in response.")
+            log.info("Signed URL received.")
+            return signed_url
+        except Exception as e:
+            log.error(f"Failed to get signed URL: {e}")
+            raise
+
+    def _process_ocr(self, signed_url: str) -> Dict[str, Any]:
+        """Sends the signed URL to the OCR endpoint for processing."""
+        log.info("Processing OCR via Mistral API")
+        url = f"{self.BASE_API_URL}/ocr"
+        ocr_headers = {
+            **self.headers,
+            "Content-Type": "application/json",
+            "Accept": "application/json",
+        }
+        payload = {
+            "model": "mistral-ocr-latest",
+            "document": {
+                "type": "document_url",
+                "document_url": signed_url,
+            },
+            "include_image_base64": False,
+        }
+
+        try:
+            response = requests.post(url, headers=ocr_headers, json=payload)
+            ocr_response = self._handle_response(response)
+            log.info("OCR processing done.")
+            log.debug("OCR response: %s", ocr_response)
+            return ocr_response
+        except Exception as e:
+            log.error(f"Failed during OCR processing: {e}")
+            raise
+
+    def _delete_file(self, file_id: str) -> None:
+        """Deletes the file from Mistral storage."""
+        log.info(f"Deleting uploaded file ID: {file_id}")
+        url = f"{self.BASE_API_URL}/files/{file_id}"
+        # No specific Accept header needed, default or Authorization is usually sufficient
+
+        try:
+            response = requests.delete(url, headers=self.headers)
+            delete_response = self._handle_response(
+                response
+            )  # Check status, ignore response body unless needed
+            log.info(
+                f"File deleted successfully: {delete_response}"
+            )  # Log the response if available
+        except Exception as e:
+            # Log error but don't necessarily halt execution if deletion fails
+            log.error(f"Failed to delete file ID {file_id}: {e}")
+            # Depending on requirements, you might choose to raise the error here
+
+    def load(self) -> List[Document]:
+        """
+        Executes the full OCR workflow: upload, get URL, process OCR, delete file.
+
+        Returns:
+            A list of Document objects, one for each page processed.
+        """
+        file_id = None
+        try:
+            # 1. Upload file
+            file_id = self._upload_file()
+
+            # 2. Get Signed URL
+            signed_url = self._get_signed_url(file_id)
+
+            # 3. Process OCR
+            ocr_response = self._process_ocr(signed_url)
+
+            # 4. Process results
+            pages_data = ocr_response.get("pages")
+            if not pages_data:
+                log.warning("No pages found in OCR response.")
+                return [Document(page_content="No text content found", metadata={})]
+
+            documents = []
+            total_pages = len(pages_data)
+            for page_data in pages_data:
+                page_content = page_data.get("markdown")
+                page_index = page_data.get("index")  # API uses 0-based index
+
+                if page_content is not None and page_index is not None:
+                    documents.append(
+                        Document(
+                            page_content=page_content,
+                            metadata={
+                                "page": page_index,  # 0-based index from API
+                                "page_label": page_index
+                                + 1,  # 1-based label for convenience
+                                "total_pages": total_pages,
+                                # Add other relevant metadata from page_data if available/needed
+                                # e.g., page_data.get('width'), page_data.get('height')
+                            },
+                        )
+                    )
+                else:
+                    log.warning(
+                        f"Skipping page due to missing 'markdown' or 'index'. Data: {page_data}"
+                    )
+
+            if not documents:
+                # Case where pages existed but none had valid markdown/index
+                log.warning(
+                    "OCR response contained pages, but none had valid content/index."
+                )
+                return [
+                    Document(
+                        page_content="No text content found in valid pages", metadata={}
+                    )
+                ]
+
+            return documents
+
+        except Exception as e:
+            log.error(f"An error occurred during the loading process: {e}")
+            # Return an empty list or a specific error document on failure
+            return [Document(page_content=f"Error during processing: {e}", metadata={})]
+        finally:
+            # 5. Delete file (attempt even if prior steps failed after upload)
+            if file_id:
+                try:
+                    self._delete_file(file_id)
+                except Exception as del_e:
+                    # Log deletion error, but don't overwrite original error if one occurred
+                    log.error(
+                        f"Cleanup error: Could not delete file ID {file_id}. Reason: {del_e}"
+                    )

backend/open_webui/retrieval/loaders/tavily.py

@@ -0,0 +1,93 @@
+import requests
+import logging
+from typing import Iterator, List, Literal, Union
+
+from langchain_core.document_loaders import BaseLoader
+from langchain_core.documents import Document
+from open_webui.env import SRC_LOG_LEVELS
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
+
+class TavilyLoader(BaseLoader):
+    """Extract web page content from URLs using Tavily Extract API.
+
+    This is a LangChain document loader that uses Tavily's Extract API to
+    retrieve content from web pages and return it as Document objects.
+
+    Args:
+        urls: URL or list of URLs to extract content from.
+        api_key: The Tavily API key.
+        extract_depth: Depth of extraction, either "basic" or "advanced".
+        continue_on_failure: Whether to continue if extraction of a URL fails.
+    """
+
+    def __init__(
+        self,
+        urls: Union[str, List[str]],
+        api_key: str,
+        extract_depth: Literal["basic", "advanced"] = "basic",
+        continue_on_failure: bool = True,
+    ) -> None:
+        """Initialize Tavily Extract client.
+
+        Args:
+            urls: URL or list of URLs to extract content from.
+            api_key: The Tavily API key.
+            include_images: Whether to include images in the extraction.
+            extract_depth: Depth of extraction, either "basic" or "advanced".
+                advanced extraction retrieves more data, including tables and
+                embedded content, with higher success but may increase latency.
+                basic costs 1 credit per 5 successful URL extractions,
+                advanced costs 2 credits per 5 successful URL extractions.
+            continue_on_failure: Whether to continue if extraction of a URL fails.
+        """
+        if not urls:
+            raise ValueError("At least one URL must be provided.")
+
+        self.api_key = api_key
+        self.urls = urls if isinstance(urls, list) else [urls]
+        self.extract_depth = extract_depth
+        self.continue_on_failure = continue_on_failure
+        self.api_url = "https://api.tavily.com/extract"
+
+    def lazy_load(self) -> Iterator[Document]:
+        """Extract and yield documents from the URLs using Tavily Extract API."""
+        batch_size = 20
+        for i in range(0, len(self.urls), batch_size):
+            batch_urls = self.urls[i : i + batch_size]
+            try:
+                headers = {
+                    "Content-Type": "application/json",
+                    "Authorization": f"Bearer {self.api_key}",
+                }
+                # Use string for single URL, array for multiple URLs
+                urls_param = batch_urls[0] if len(batch_urls) == 1 else batch_urls
+                payload = {"urls": urls_param, "extract_depth": self.extract_depth}
+                # Make the API call
+                response = requests.post(self.api_url, headers=headers, json=payload)
+                response.raise_for_status()
+                response_data = response.json()
+                # Process successful results
+                for result in response_data.get("results", []):
+                    url = result.get("url", "")
+                    content = result.get("raw_content", "")
+                    if not content:
+                        log.warning(f"No content extracted from {url}")
+                        continue
+                    # Add URLs as metadata
+                    metadata = {"source": url}
+                    yield Document(
+                        page_content=content,
+                        metadata=metadata,
+                    )
+                for failed in response_data.get("failed_results", []):
+                    url = failed.get("url", "")
+                    error = failed.get("error", "Unknown error")
+                    log.error(f"Failed to extract content from {url}: {error}")
+            except Exception as e:
+                if self.continue_on_failure:
+                    log.error(f"Error extracting content from batch {batch_urls}: {e}")
+                else:
+                    raise e

backend/open_webui/retrieval/loaders/youtube.py

@@ -110,7 +110,7 @@ class YoutubeLoader:
 
         transcript = " ".join(
             map(
-                lambda transcript_piece: transcript_piece["text"].strip(" "),
+                lambda transcript_piece: transcript_piece.text.strip(" "),
                 transcript_pieces,
             )
         )

backend/open_webui/retrieval/models/colbert.py

@@ -1,13 +1,19 @@
 import os
+import logging
 import torch
 import numpy as np
 from colbert.infra import ColBERTConfig
 from colbert.modeling.checkpoint import Checkpoint
 
+from open_webui.env import SRC_LOG_LEVELS
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
 
 class ColBERT:
     def __init__(self, name, **kwargs) -> None:
-        print("ColBERT: Loading model", name)
+        log.info("ColBERT: Loading model", name)
         self.device = "cuda" if torch.cuda.is_available() else "cpu"
 
         DOCKER = kwargs.get("env") == "docker"

backend/open_webui/retrieval/utils.py

@@ -1,22 +1,35 @@
 import logging
 import os
-import uuid
 from typing import Optional, Union
 
-import asyncio
 import requests
+import hashlib
+from concurrent.futures import ThreadPoolExecutor
 
 from huggingface_hub import snapshot_download
 from langchain.retrievers import ContextualCompressionRetriever, EnsembleRetriever
 from langchain_community.retrievers import BM25Retriever
 from langchain_core.documents import Document
 
-
 from open_webui.config import VECTOR_DB
 from open_webui.retrieval.vector.connector import VECTOR_DB_CLIENT
-from open_webui.utils.misc import get_last_user_message
 
-from open_webui.env import SRC_LOG_LEVELS, OFFLINE_MODE
+from open_webui.models.users import UserModel
+from open_webui.models.files import Files
+
+from open_webui.retrieval.vector.main import GetResult
+
+
+from open_webui.env import (
+    SRC_LOG_LEVELS,
+    OFFLINE_MODE,
+    ENABLE_FORWARD_USER_INFO_HEADERS,
+)
+from open_webui.config import (
+    RAG_EMBEDDING_QUERY_PREFIX,
+    RAG_EMBEDDING_CONTENT_PREFIX,
+    RAG_EMBEDDING_PREFIX_FIELD_NAME,
+)
 
 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["RAG"])
@@ -41,7 +54,7 @@ class VectorSearchRetriever(BaseRetriever):
     ) -> list[Document]:
         result = VECTOR_DB_CLIENT.search(
             collection_name=self.collection_name,
-            vectors=[self.embedding_function(query)],
+            vectors=[self.embedding_function(query, RAG_EMBEDDING_QUERY_PREFIX)],
             limit=self.top_k,
         )
 
@@ -61,11 +74,10 @@ class VectorSearchRetriever(BaseRetriever):
 
 
 def query_doc(
-    collection_name: str,
-    query_embedding: list[float],
-    k: int,
+    collection_name: str, query_embedding: list[float], k: int, user: UserModel = None
 ):
     try:
+        log.debug(f"query_doc:doc {collection_name}")
         result = VECTOR_DB_CLIENT.search(
             collection_name=collection_name,
             vectors=[query_embedding],
@@ -77,24 +89,39 @@ def query_doc(
 
         return result
     except Exception as e:
-        print(e)
+        log.exception(f"Error querying doc {collection_name} with limit {k}: {e}")
+        raise e
+
+
+def get_doc(collection_name: str, user: UserModel = None):
+    try:
+        log.debug(f"get_doc:doc {collection_name}")
+        result = VECTOR_DB_CLIENT.get(collection_name=collection_name)
+
+        if result:
+            log.info(f"query_doc:result {result.ids} {result.metadatas}")
+
+        return result
+    except Exception as e:
+        log.exception(f"Error getting doc {collection_name}: {e}")
         raise e
 
 
 def query_doc_with_hybrid_search(
     collection_name: str,
+    collection_result: GetResult,
     query: str,
     embedding_function,
     k: int,
     reranking_function,
+    k_reranker: int,
     r: float,
 ) -> dict:
     try:
-        result = VECTOR_DB_CLIENT.get(collection_name=collection_name)
-
+        log.debug(f"query_doc_with_hybrid_search:doc {collection_name}")
         bm25_retriever = BM25Retriever.from_texts(
-            texts=result.documents[0],
-            metadatas=result.metadatas[0],
+            texts=collection_result.documents[0],
+            metadatas=collection_result.metadatas[0],
         )
         bm25_retriever.k = k
 
@@ -109,7 +136,7 @@ def query_doc_with_hybrid_search(
         )
         compressor = RerankCompressor(
             embedding_function=embedding_function,
-            top_n=k,
+            top_n=k_reranker,
             reranking_function=reranking_function,
             r_score=r,
         )
@@ -119,10 +146,23 @@ def query_doc_with_hybrid_search(
         )
 
         result = compression_retriever.invoke(query)
+
+        distances = [d.metadata.get("score") for d in result]
+        documents = [d.page_content for d in result]
+        metadatas = [d.metadata for d in result]
+
+        # retrieve only min(k, k_reranker) items, sort and cut by distance if k < k_reranker
+        if k < k_reranker:
+            sorted_items = sorted(
+                zip(distances, metadatas, documents), key=lambda x: x[0], reverse=True
+            )
+            sorted_items = sorted_items[:k]
+            distances, documents, metadatas = map(list, zip(*sorted_items))
+
         result = {
-            "distances": [[d.metadata.get("score") for d in result]],
-            "documents": [[d.page_content for d in result]],
-            "metadatas": [[d.metadata for d in result]],
+            "distances": [distances],
+            "documents": [documents],
+            "metadatas": [metadatas],
         }
 
         log.info(
@@ -131,52 +171,88 @@ def query_doc_with_hybrid_search(
         )
         return result
     except Exception as e:
+        log.exception(f"Error querying doc {collection_name} with hybrid search: {e}")
         raise e
 
 
-def merge_and_sort_query_results(
-    query_results: list[dict], k: int, reverse: bool = False
-) -> list[dict]:
+def merge_get_results(get_results: list[dict]) -> dict:
     # Initialize lists to store combined data
-    combined_distances = []
     combined_documents = []
     combined_metadatas = []
+    combined_ids = []
 
-    for data in query_results:
-        combined_distances.extend(data["distances"][0])
+    for data in get_results:
         combined_documents.extend(data["documents"][0])
         combined_metadatas.extend(data["metadatas"][0])
-
-    # Create a list of tuples (distance, document, metadata)
-    combined = list(zip(combined_distances, combined_documents, combined_metadatas))
-
-    # Sort the list based on distances
-    combined.sort(key=lambda x: x[0], reverse=reverse)
-
-    # We don't have anything :-(
-    if not combined:
-        sorted_distances = []
-        sorted_documents = []
-        sorted_metadatas = []
-    else:
-        # Unzip the sorted list
-        sorted_distances, sorted_documents, sorted_metadatas = zip(*combined)
-
-        # Slicing the lists to include only k elements
-        sorted_distances = list(sorted_distances)[:k]
-        sorted_documents = list(sorted_documents)[:k]
-        sorted_metadatas = list(sorted_metadatas)[:k]
+        combined_ids.extend(data["ids"][0])
 
     # Create the output dictionary
     result = {
-        "distances": [sorted_distances],
-        "documents": [sorted_documents],
-        "metadatas": [sorted_metadatas],
+        "documents": [combined_documents],
+        "metadatas": [combined_metadatas],
+        "ids": [combined_ids],
     }
 
     return result
 
 
+def merge_and_sort_query_results(query_results: list[dict], k: int) -> dict:
+    # Initialize lists to store combined data
+    combined = dict()  # To store documents with unique document hashes
+
+    for data in query_results:
+        distances = data["distances"][0]
+        documents = data["documents"][0]
+        metadatas = data["metadatas"][0]
+
+        for distance, document, metadata in zip(distances, documents, metadatas):
+            if isinstance(document, str):
+                doc_hash = hashlib.md5(
+                    document.encode()
+                ).hexdigest()  # Compute a hash for uniqueness
+
+                if doc_hash not in combined.keys():
+                    combined[doc_hash] = (distance, document, metadata)
+                    continue  # if doc is new, no further comparison is needed
+
+                # if doc is alredy in, but new distance is better, update
+                if distance > combined[doc_hash][0]:
+                    combined[doc_hash] = (distance, document, metadata)
+
+    combined = list(combined.values())
+    # Sort the list based on distances
+    combined.sort(key=lambda x: x[0], reverse=True)
+
+    # Slice to keep only the top k elements
+    sorted_distances, sorted_documents, sorted_metadatas = (
+        zip(*combined[:k]) if combined else ([], [], [])
+    )
+
+    # Create and return the output dictionary
+    return {
+        "distances": [list(sorted_distances)],
+        "documents": [list(sorted_documents)],
+        "metadatas": [list(sorted_metadatas)],
+    }
+
+
+def get_all_items_from_collections(collection_names: list[str]) -> dict:
+    results = []
+
+    for collection_name in collection_names:
+        if collection_name:
+            try:
+                result = get_doc(collection_name=collection_name)
+                if result is not None:
+                    results.append(result.model_dump())
+            except Exception as e:
+                log.exception(f"Error when querying the collection: {e}")
+        else:
+            pass
+
+    return merge_get_results(results)
+
+
 def query_collection(
     collection_names: list[str],
     queries: list[str],
@@ -185,7 +261,8 @@ def query_collection(
 ) -> dict:
     results = []
     for query in queries:
-        query_embedding = embedding_function(query)
+        log.debug(f"query_collection:query {query}")
+        query_embedding = embedding_function(query, prefix=RAG_EMBEDDING_QUERY_PREFIX)
         for collection_name in collection_names:
             if collection_name:
                 try:
@@ -201,12 +278,7 @@ def query_collection(
             else:
                 pass
 
-    if VECTOR_DB == "chroma":
-        # Chroma uses unconventional cosine similarity, so we don't need to reverse the results
-        # https://docs.trychroma.com/docs/collections/configure#configuring-chroma-collections
-        return merge_and_sort_query_results(results, k=k, reverse=False)
-    else:
-        return merge_and_sort_query_results(results, k=k, reverse=True)
+    return merge_and_sort_query_results(results, k=k)
 
 
 def query_collection_with_hybrid_search(
@@ -215,39 +287,72 @@ def query_collection_with_hybrid_search(
     embedding_function,
     k: int,
     reranking_function,
+    k_reranker: int,
     r: float,
 ) -> dict:
     results = []
     error = False
+    # Fetch collection data once per collection sequentially
+    # Avoid fetching the same data multiple times later
+    collection_results = {}
     for collection_name in collection_names:
         try:
-            for query in queries:
-                result = query_doc_with_hybrid_search(
-                    collection_name=collection_name,
-                    query=query,
-                    embedding_function=embedding_function,
-                    k=k,
-                    reranking_function=reranking_function,
-                    r=r,
-                )
-                results.append(result)
-        except Exception as e:
-            log.exception(
-                "Error when querying the collection with " f"hybrid_search: {e}"
+            log.debug(
+                f"query_collection_with_hybrid_search:VECTOR_DB_CLIENT.get:collection {collection_name}"
             )
-            error = True
+            collection_results[collection_name] = VECTOR_DB_CLIENT.get(
+                collection_name=collection_name
+            )
+        except Exception as e:
+            log.exception(f"Failed to fetch collection {collection_name}: {e}")
+            collection_results[collection_name] = None
 
-    if error:
+    log.info(
+        f"Starting hybrid search for {len(queries)} queries in {len(collection_names)} collections..."
+    )
+
+    def process_query(collection_name, query):
+        try:
+            result = query_doc_with_hybrid_search(
+                collection_name=collection_name,
+                collection_result=collection_results[collection_name],
+                query=query,
+                embedding_function=embedding_function,
+                k=k,
+                reranking_function=reranking_function,
+                k_reranker=k_reranker,
+                r=r,
+            )
+            return result, None
+        except Exception as e:
+            log.exception(f"Error when querying the collection with hybrid_search: {e}")
+            return None, e
+
+    # Prepare tasks for all collections and queries
+    # Avoid running any tasks for collections that failed to fetch data (have assigned None)
+    tasks = [
+        (cn, q)
+        for cn in collection_names
+        if collection_results[cn] is not None
+        for q in queries
+    ]
+
+    with ThreadPoolExecutor() as executor:
+        future_results = [executor.submit(process_query, cn, q) for cn, q in tasks]
+        task_results = [future.result() for future in future_results]
+
+    for result, err in task_results:
+        if err is not None:
+            error = True
+        elif result is not None:
+            results.append(result)
+
+    if error and not results:
         raise Exception(
-            "Hybrid search failed for all collections. Using Non hybrid search as fallback."
+            "Hybrid search failed for all collections. Using Non-hybrid search as fallback."
         )
 
-    if VECTOR_DB == "chroma":
-        # Chroma uses unconventional cosine similarity, so we don't need to reverse the results
-        # https://docs.trychroma.com/docs/collections/configure#configuring-chroma-collections
-        return merge_and_sort_query_results(results, k=k, reverse=False)
-    else:
-        return merge_and_sort_query_results(results, k=k, reverse=True)
+    return merge_and_sort_query_results(results, k=k)
 
 
 def get_embedding_function(
@@ -259,51 +364,127 @@ def get_embedding_function(
     embedding_batch_size,
 ):
     if embedding_engine == "":
-        return lambda query: embedding_function.encode(query).tolist()
+        return lambda query, prefix=None, user=None: embedding_function.encode(
+            query, **({"prompt": prefix} if prefix else {})
+        ).tolist()
     elif embedding_engine in ["ollama", "openai"]:
-        func = lambda query: generate_embeddings(
+        func = lambda query, prefix=None, user=None: generate_embeddings(
             engine=embedding_engine,
             model=embedding_model,
             text=query,
+            prefix=prefix,
             url=url,
             key=key,
+            user=user,
         )
 
-        def generate_multiple(query, func):
+        def generate_multiple(query, prefix, user, func):
             if isinstance(query, list):
                 embeddings = []
                 for i in range(0, len(query), embedding_batch_size):
-                    embeddings.extend(func(query[i : i + embedding_batch_size]))
+                    embeddings.extend(
+                        func(
+                            query[i : i + embedding_batch_size],
+                            prefix=prefix,
+                            user=user,
+                        )
+                    )
                 return embeddings
             else:
-                return func(query)
+                return func(query, prefix, user)
 
-        return lambda query: generate_multiple(query, func)
+        return lambda query, prefix=None, user=None: generate_multiple(
+            query, prefix, user, func
+        )
+    else:
+        raise ValueError(f"Unknown embedding engine: {embedding_engine}")
 
 
 def get_sources_from_files(
+    request,
     files,
     queries,
     embedding_function,
     k,
     reranking_function,
+    k_reranker,
     r,
     hybrid_search,
+    full_context=False,
 ):
-    log.debug(f"files: {files} {queries} {embedding_function} {reranking_function}")
+    log.debug(
+        f"files: {files} {queries} {embedding_function} {reranking_function} {full_context}"
+    )
 
     extracted_collections = []
     relevant_contexts = []
 
     for file in files:
-        if file.get("context") == "full":
+
+        context = None
+        if file.get("docs"):
+            # BYPASS_WEB_SEARCH_EMBEDDING_AND_RETRIEVAL
+            context = {
+                "documents": [[doc.get("content") for doc in file.get("docs")]],
+                "metadatas": [[doc.get("metadata") for doc in file.get("docs")]],
+            }
+        elif file.get("context") == "full":
+            # Manual Full Mode Toggle
             context = {
                 "documents": [[file.get("file").get("data", {}).get("content")]],
                 "metadatas": [[{"file_id": file.get("id"), "name": file.get("name")}]],
             }
-        else:
-            context = None
+        elif (
+            file.get("type") != "web_search"
+            and request.app.state.config.BYPASS_EMBEDDING_AND_RETRIEVAL
+        ):
+            # BYPASS_EMBEDDING_AND_RETRIEVAL
+            if file.get("type") == "collection":
+                file_ids = file.get("data", {}).get("file_ids", [])
 
+                documents = []
+                metadatas = []
+                for file_id in file_ids:
+                    file_object = Files.get_file_by_id(file_id)
+
+                    if file_object:
+                        documents.append(file_object.data.get("content", ""))
+                        metadatas.append(
+                            {
+                                "file_id": file_id,
+                                "name": file_object.filename,
+                                "source": file_object.filename,
+                            }
+                        )
+
+                context = {
+                    "documents": [documents],
+                    "metadatas": [metadatas],
+                }
+
+            elif file.get("id"):
+                file_object = Files.get_file_by_id(file.get("id"))
+                if file_object:
+                    context = {
+                        "documents": [[file_object.data.get("content", "")]],
+                        "metadatas": [
+                            [
+                                {
+                                    "file_id": file.get("id"),
+                                    "name": file_object.filename,
+                                    "source": file_object.filename,
+                                }
+                            ]
+                        ],
+                    }
+            elif file.get("file").get("data"):
+                context = {
+                    "documents": [[file.get("file").get("data", {}).get("content")]],
+                    "metadatas": [
+                        [file.get("file").get("data", {}).get("metadata", {})]
+                    ],
+                }
+        else:
             collection_names = []
             if file.get("type") == "collection":
                 if file.get("legacy"):
@@ -323,42 +504,51 @@ def get_sources_from_files(
                 log.debug(f"skipping {file} as it has already been extracted")
                 continue
 
-            try:
-                context = None
-                if file.get("type") == "text":
-                    context = file["content"]
-                else:
-                    if hybrid_search:
-                        try:
-                            context = query_collection_with_hybrid_search(
+            if full_context:
+                try:
+                    context = get_all_items_from_collections(collection_names)
+                except Exception as e:
+                    log.exception(e)
+
+            else:
+                try:
+                    context = None
+                    if file.get("type") == "text":
+                        context = file["content"]
+                    else:
+                        if hybrid_search:
+                            try:
+                                context = query_collection_with_hybrid_search(
+                                    collection_names=collection_names,
+                                    queries=queries,
+                                    embedding_function=embedding_function,
+                                    k=k,
+                                    reranking_function=reranking_function,
+                                    k_reranker=k_reranker,
+                                    r=r,
+                                )
+                            except Exception as e:
+                                log.debug(
+                                    "Error when using hybrid search, using"
+                                    " non hybrid search as fallback."
+                                )
+
+                        if (not hybrid_search) or (context is None):
+                            context = query_collection(
                                 collection_names=collection_names,
                                 queries=queries,
                                 embedding_function=embedding_function,
                                 k=k,
-                                reranking_function=reranking_function,
-                                r=r,
                             )
-                        except Exception as e:
-                            log.debug(
-                                "Error when using hybrid search, using"
-                                " non hybrid search as fallback."
-                            )
-
-                    if (not hybrid_search) or (context is None):
-                        context = query_collection(
-                            collection_names=collection_names,
-                            queries=queries,
-                            embedding_function=embedding_function,
-                            k=k,
-                        )
-            except Exception as e:
-                log.exception(e)
+                except Exception as e:
+                    log.exception(e)
 
             extracted_collections.extend(collection_names)
 
         if context:
             if "data" in file:
                 del file["data"]
+
             relevant_contexts.append({**context, "file": file})
 
     sources = []
@@ -423,16 +613,38 @@ def get_model_path(model: str, update_model: bool = False):
 
 
 def generate_openai_batch_embeddings(
-    model: str, texts: list[str], url: str = "https://api.openai.com/v1", key: str = ""
+    model: str,
+    texts: list[str],
+    url: str = "https://api.openai.com/v1",
+    key: str = "",
+    prefix: str = None,
+    user: UserModel = None,
 ) -> Optional[list[list[float]]]:
     try:
+        log.debug(
+            f"generate_openai_batch_embeddings:model {model} batch size: {len(texts)}"
+        )
+        json_data = {"input": texts, "model": model}
+        if isinstance(RAG_EMBEDDING_PREFIX_FIELD_NAME, str) and isinstance(prefix, str):
+            json_data[RAG_EMBEDDING_PREFIX_FIELD_NAME] = prefix
+
         r = requests.post(
             f"{url}/embeddings",
             headers={
                 "Content-Type": "application/json",
                 "Authorization": f"Bearer {key}",
+                **(
+                    {
+                        "X-OpenWebUI-User-Name": user.name,
+                        "X-OpenWebUI-User-Id": user.id,
+                        "X-OpenWebUI-User-Email": user.email,
+                        "X-OpenWebUI-User-Role": user.role,
+                    }
+                    if ENABLE_FORWARD_USER_INFO_HEADERS and user
+                    else {}
+                ),
             },
-            json={"input": texts, "model": model},
+            json=json_data,
         )
         r.raise_for_status()
         data = r.json()
@@ -441,21 +653,43 @@ def generate_openai_batch_embeddings(
         else:
             raise "Something went wrong :/"
     except Exception as e:
-        print(e)
+        log.exception(f"Error generating openai batch embeddings: {e}")
         return None
 
 
 def generate_ollama_batch_embeddings(
-    model: str, texts: list[str], url: str, key: str = ""
+    model: str,
+    texts: list[str],
+    url: str,
+    key: str = "",
+    prefix: str = None,
+    user: UserModel = None,
 ) -> Optional[list[list[float]]]:
     try:
+        log.debug(
+            f"generate_ollama_batch_embeddings:model {model} batch size: {len(texts)}"
+        )
+        json_data = {"input": texts, "model": model}
+        if isinstance(RAG_EMBEDDING_PREFIX_FIELD_NAME, str) and isinstance(prefix, str):
+            json_data[RAG_EMBEDDING_PREFIX_FIELD_NAME] = prefix
+
         r = requests.post(
             f"{url}/api/embed",
             headers={
                 "Content-Type": "application/json",
                 "Authorization": f"Bearer {key}",
+                **(
+                    {
+                        "X-OpenWebUI-User-Name": user.name,
+                        "X-OpenWebUI-User-Id": user.id,
+                        "X-OpenWebUI-User-Email": user.email,
+                        "X-OpenWebUI-User-Role": user.role,
+                    }
+                    if ENABLE_FORWARD_USER_INFO_HEADERS
+                    else {}
+                ),
             },
-            json={"input": texts, "model": model},
+            json=json_data,
         )
         r.raise_for_status()
         data = r.json()
@@ -465,30 +699,60 @@ def generate_ollama_batch_embeddings(
         else:
             raise "Something went wrong :/"
     except Exception as e:
-        print(e)
+        log.exception(f"Error generating ollama batch embeddings: {e}")
         return None
 
 
-def generate_embeddings(engine: str, model: str, text: Union[str, list[str]], **kwargs):
+def generate_embeddings(
+    engine: str,
+    model: str,
+    text: Union[str, list[str]],
+    prefix: Union[str, None] = None,
+    **kwargs,
+):
     url = kwargs.get("url", "")
     key = kwargs.get("key", "")
+    user = kwargs.get("user")
+
+    if prefix is not None and RAG_EMBEDDING_PREFIX_FIELD_NAME is None:
+        if isinstance(text, list):
+            text = [f"{prefix}{text_element}" for text_element in text]
+        else:
+            text = f"{prefix}{text}"
 
     if engine == "ollama":
         if isinstance(text, list):
             embeddings = generate_ollama_batch_embeddings(
-                **{"model": model, "texts": text, "url": url, "key": key}
+                **{
+                    "model": model,
+                    "texts": text,
+                    "url": url,
+                    "key": key,
+                    "prefix": prefix,
+                    "user": user,
+                }
             )
         else:
             embeddings = generate_ollama_batch_embeddings(
-                **{"model": model, "texts": [text], "url": url, "key": key}
+                **{
+                    "model": model,
+                    "texts": [text],
+                    "url": url,
+                    "key": key,
+                    "prefix": prefix,
+                    "user": user,
+                }
             )
         return embeddings[0] if isinstance(text, str) else embeddings
     elif engine == "openai":
         if isinstance(text, list):
-            embeddings = generate_openai_batch_embeddings(model, text, url, key)
+            embeddings = generate_openai_batch_embeddings(
+                model, text, url, key, prefix, user
+            )
         else:
-            embeddings = generate_openai_batch_embeddings(model, [text], url, key)
-
+            embeddings = generate_openai_batch_embeddings(
+                model, [text], url, key, prefix, user
+            )
         return embeddings[0] if isinstance(text, str) else embeddings
 
 
@@ -524,9 +788,9 @@ class RerankCompressor(BaseDocumentCompressor):
         else:
             from sentence_transformers import util
 
-            query_embedding = self.embedding_function(query)
+            query_embedding = self.embedding_function(query, RAG_EMBEDDING_QUERY_PREFIX)
             document_embedding = self.embedding_function(
-                [doc.page_content for doc in documents]
+                [doc.page_content for doc in documents], RAG_EMBEDDING_CONTENT_PREFIX
             )
             scores = util.cos_sim(query_embedding, document_embedding)[0]
 

backend/open_webui/retrieval/vector/connector.py

@@ -16,6 +16,10 @@ elif VECTOR_DB == "pgvector":
     from open_webui.retrieval.vector.dbs.pgvector import PgvectorClient
 
     VECTOR_DB_CLIENT = PgvectorClient()
+elif VECTOR_DB == "elasticsearch":
+    from open_webui.retrieval.vector.dbs.elasticsearch import ElasticsearchClient
+
+    VECTOR_DB_CLIENT = ElasticsearchClient()
 else:
     from open_webui.retrieval.vector.dbs.chroma import ChromaClient
 

backend/open_webui/retrieval/vector/dbs/chroma.py

@@ -1,4 +1,5 @@
 import chromadb
+import logging
 from chromadb import Settings
 from chromadb.utils.batch_utils import create_batches
 
@@ -16,6 +17,10 @@ from open_webui.config import (
     CHROMA_CLIENT_AUTH_PROVIDER,
     CHROMA_CLIENT_AUTH_CREDENTIALS,
 )
+from open_webui.env import SRC_LOG_LEVELS
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
 
 
 class ChromaClient:
@@ -70,10 +75,16 @@ class ChromaClient:
                     n_results=limit,
                 )
 
+                # chromadb has cosine distance, 2 (worst) -> 0 (best). Re-odering to 0 -> 1
+                # https://docs.trychroma.com/docs/collections/configure cosine equation
+                distances: list = result["distances"][0]
+                distances = [2 - dist for dist in distances]
+                distances = [[dist / 2 for dist in distances]]
+
                 return SearchResult(
                     **{
                         "ids": result["ids"],
-                        "distances": result["distances"],
+                        "distances": distances,
                         "documents": result["documents"],
                         "metadatas": result["metadatas"],
                     }
@@ -102,8 +113,7 @@ class ChromaClient:
                     }
                 )
             return None
-        except Exception as e:
-            print(e)
+        except:
             return None
 
     def get(self, collection_name: str) -> Optional[GetResult]:
@@ -162,12 +172,19 @@ class ChromaClient:
         filter: Optional[dict] = None,
     ):
         # Delete the items from the collection based on the ids.
-        collection = self.client.get_collection(name=collection_name)
-        if collection:
-            if ids:
-                collection.delete(ids=ids)
-            elif filter:
-                collection.delete(where=filter)
+        try:
+            collection = self.client.get_collection(name=collection_name)
+            if collection:
+                if ids:
+                    collection.delete(ids=ids)
+                elif filter:
+                    collection.delete(where=filter)
+        except Exception as e:
+            # If collection doesn't exist, that's fine - nothing to delete
+            log.debug(
+                f"Attempted to delete from non-existent collection {collection_name}. Ignoring."
+            )
+            pass
 
     def reset(self):
         # Resets the database. This will delete all collections and item entries.

backend/open_webui/retrieval/vector/dbs/elasticsearch.py

@@ -0,0 +1,295 @@
+from elasticsearch import Elasticsearch, BadRequestError
+from typing import Optional
+import ssl
+from elasticsearch.helpers import bulk, scan
+from open_webui.retrieval.vector.main import VectorItem, SearchResult, GetResult
+from open_webui.config import (
+    ELASTICSEARCH_URL,
+    ELASTICSEARCH_CA_CERTS,
+    ELASTICSEARCH_API_KEY,
+    ELASTICSEARCH_USERNAME,
+    ELASTICSEARCH_PASSWORD,
+    ELASTICSEARCH_CLOUD_ID,
+    ELASTICSEARCH_INDEX_PREFIX,
+    SSL_ASSERT_FINGERPRINT,
+)
+
+
+class ElasticsearchClient:
+    """
+    Important:
+    in order to reduce the number of indexes and since the embedding vector length is fixed, we avoid creating
+    an index for each file but store it as a text field, while seperating to different index
+    baesd on the embedding length.
+    """
+
+    def __init__(self):
+        self.index_prefix = ELASTICSEARCH_INDEX_PREFIX
+        self.client = Elasticsearch(
+            hosts=[ELASTICSEARCH_URL],
+            ca_certs=ELASTICSEARCH_CA_CERTS,
+            api_key=ELASTICSEARCH_API_KEY,
+            cloud_id=ELASTICSEARCH_CLOUD_ID,
+            basic_auth=(
+                (ELASTICSEARCH_USERNAME, ELASTICSEARCH_PASSWORD)
+                if ELASTICSEARCH_USERNAME and ELASTICSEARCH_PASSWORD
+                else None
+            ),
+            ssl_assert_fingerprint=SSL_ASSERT_FINGERPRINT,
+        )
+
+    # Status: works
+    def _get_index_name(self, dimension: int) -> str:
+        return f"{self.index_prefix}_d{str(dimension)}"
+
+    # Status: works
+    def _scan_result_to_get_result(self, result) -> GetResult:
+        if not result:
+            return None
+        ids = []
+        documents = []
+        metadatas = []
+
+        for hit in result:
+            ids.append(hit["_id"])
+            documents.append(hit["_source"].get("text"))
+            metadatas.append(hit["_source"].get("metadata"))
+
+        return GetResult(ids=[ids], documents=[documents], metadatas=[metadatas])
+
+    # Status: works
+    def _result_to_get_result(self, result) -> GetResult:
+        if not result["hits"]["hits"]:
+            return None
+        ids = []
+        documents = []
+        metadatas = []
+
+        for hit in result["hits"]["hits"]:
+            ids.append(hit["_id"])
+            documents.append(hit["_source"].get("text"))
+            metadatas.append(hit["_source"].get("metadata"))
+
+        return GetResult(ids=[ids], documents=[documents], metadatas=[metadatas])
+
+    # Status: works
+    def _result_to_search_result(self, result) -> SearchResult:
+        ids = []
+        distances = []
+        documents = []
+        metadatas = []
+
+        for hit in result["hits"]["hits"]:
+            ids.append(hit["_id"])
+            distances.append(hit["_score"])
+            documents.append(hit["_source"].get("text"))
+            metadatas.append(hit["_source"].get("metadata"))
+
+        return SearchResult(
+            ids=[ids],
+            distances=[distances],
+            documents=[documents],
+            metadatas=[metadatas],
+        )
+
+    # Status: works
+    def _create_index(self, dimension: int):
+        body = {
+            "mappings": {
+                "dynamic_templates": [
+                    {
+                        "strings": {
+                            "match_mapping_type": "string",
+                            "mapping": {"type": "keyword"},
+                        }
+                    }
+                ],
+                "properties": {
+                    "collection": {"type": "keyword"},
+                    "id": {"type": "keyword"},
+                    "vector": {
+                        "type": "dense_vector",
+                        "dims": dimension,  # Adjust based on your vector dimensions
+                        "index": True,
+                        "similarity": "cosine",
+                    },
+                    "text": {"type": "text"},
+                    "metadata": {"type": "object"},
+                },
+            }
+        }
+        self.client.indices.create(index=self._get_index_name(dimension), body=body)
+
+    # Status: works
+
+    def _create_batches(self, items: list[VectorItem], batch_size=100):
+        for i in range(0, len(items), batch_size):
+            yield items[i : min(i + batch_size, len(items))]
+
+    # Status: works
+    def has_collection(self, collection_name) -> bool:
+        query_body = {"query": {"bool": {"filter": []}}}
+        query_body["query"]["bool"]["filter"].append(
+            {"term": {"collection": collection_name}}
+        )
+
+        try:
+            result = self.client.count(index=f"{self.index_prefix}*", body=query_body)
+
+            return result.body["count"] > 0
+        except Exception as e:
+            return None
+
+    def delete_collection(self, collection_name: str):
+        query = {"query": {"term": {"collection": collection_name}}}
+        self.client.delete_by_query(index=f"{self.index_prefix}*", body=query)
+
+    # Status: works
+    def search(
+        self, collection_name: str, vectors: list[list[float]], limit: int
+    ) -> Optional[SearchResult]:
+        query = {
+            "size": limit,
+            "_source": ["text", "metadata"],
+            "query": {
+                "script_score": {
+                    "query": {
+                        "bool": {"filter": [{"term": {"collection": collection_name}}]}
+                    },
+                    "script": {
+                        "source": "cosineSimilarity(params.vector, 'vector') + 1.0",
+                        "params": {
+                            "vector": vectors[0]
+                        },  # Assuming single query vector
+                    },
+                }
+            },
+        }
+
+        result = self.client.search(
+            index=self._get_index_name(len(vectors[0])), body=query
+        )
+
+        return self._result_to_search_result(result)
+
+    # Status: only tested halfwat
+    def query(
+        self, collection_name: str, filter: dict, limit: Optional[int] = None
+    ) -> Optional[GetResult]:
+        if not self.has_collection(collection_name):
+            return None
+
+        query_body = {
+            "query": {"bool": {"filter": []}},
+            "_source": ["text", "metadata"],
+        }
+
+        for field, value in filter.items():
+            query_body["query"]["bool"]["filter"].append({"term": {field: value}})
+        query_body["query"]["bool"]["filter"].append(
+            {"term": {"collection": collection_name}}
+        )
+        size = limit if limit else 10
+
+        try:
+            result = self.client.search(
+                index=f"{self.index_prefix}*",
+                body=query_body,
+                size=size,
+            )
+
+            return self._result_to_get_result(result)
+
+        except Exception as e:
+            return None
+
+    # Status: works
+    def _has_index(self, dimension: int):
+        return self.client.indices.exists(
+            index=self._get_index_name(dimension=dimension)
+        )
+
+    def get_or_create_index(self, dimension: int):
+        if not self._has_index(dimension=dimension):
+            self._create_index(dimension=dimension)
+
+    # Status: works
+    def get(self, collection_name: str) -> Optional[GetResult]:
+        # Get all the items in the collection.
+        query = {
+            "query": {"bool": {"filter": [{"term": {"collection": collection_name}}]}},
+            "_source": ["text", "metadata"],
+        }
+        results = list(scan(self.client, index=f"{self.index_prefix}*", query=query))
+
+        return self._scan_result_to_get_result(results)
+
+    # Status: works
+    def insert(self, collection_name: str, items: list[VectorItem]):
+        if not self._has_index(dimension=len(items[0]["vector"])):
+            self._create_index(dimension=len(items[0]["vector"]))
+
+        for batch in self._create_batches(items):
+            actions = [
+                {
+                    "_index": self._get_index_name(dimension=len(items[0]["vector"])),
+                    "_id": item["id"],
+                    "_source": {
+                        "collection": collection_name,
+                        "vector": item["vector"],
+                        "text": item["text"],
+                        "metadata": item["metadata"],
+                    },
+                }
+                for item in batch
+            ]
+            bulk(self.client, actions)
+
+    # Upsert documents using the update API with doc_as_upsert=True.
+    def upsert(self, collection_name: str, items: list[VectorItem]):
+        if not self._has_index(dimension=len(items[0]["vector"])):
+            self._create_index(dimension=len(items[0]["vector"]))
+        for batch in self._create_batches(items):
+            actions = [
+                {
+                    "_op_type": "update",
+                    "_index": self._get_index_name(dimension=len(item["vector"])),
+                    "_id": item["id"],
+                    "doc": {
+                        "collection": collection_name,
+                        "vector": item["vector"],
+                        "text": item["text"],
+                        "metadata": item["metadata"],
+                    },
+                    "doc_as_upsert": True,
+                }
+                for item in batch
+            ]
+            bulk(self.client, actions)
+
+    # Delete specific documents from a collection by filtering on both collection and document IDs.
+    def delete(
+        self,
+        collection_name: str,
+        ids: Optional[list[str]] = None,
+        filter: Optional[dict] = None,
+    ):
+
+        query = {
+            "query": {"bool": {"filter": [{"term": {"collection": collection_name}}]}}
+        }
+        # logic based on chromaDB
+        if ids:
+            query["query"]["bool"]["filter"].append({"terms": {"_id": ids}})
+        elif filter:
+            for field, value in filter.items():
+                query["query"]["bool"]["filter"].append(
+                    {"term": {f"metadata.{field}": value}}
+                )
+
+        self.client.delete_by_query(index=f"{self.index_prefix}*", body=query)
+
+    def reset(self):
+        indices = self.client.indices.get(index=f"{self.index_prefix}*")
+        for index in indices:
+            self.client.indices.delete(index=index)

backend/open_webui/retrieval/vector/dbs/milvus.py

@@ -1,20 +1,28 @@
 from pymilvus import MilvusClient as Client
 from pymilvus import FieldSchema, DataType
 import json
-
+import logging
 from typing import Optional
 
 from open_webui.retrieval.vector.main import VectorItem, SearchResult, GetResult
 from open_webui.config import (
     MILVUS_URI,
     MILVUS_DB,
+    MILVUS_TOKEN,
 )
+from open_webui.env import SRC_LOG_LEVELS
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
 
 
 class MilvusClient:
     def __init__(self):
         self.collection_prefix = "open_webui"
-        self.client = Client(uri=MILVUS_URI, database=MILVUS_DB)
+        if MILVUS_TOKEN is None:
+            self.client = Client(uri=MILVUS_URI, db_name=MILVUS_DB)
+        else:
+            self.client = Client(uri=MILVUS_URI, db_name=MILVUS_DB, token=MILVUS_TOKEN)
 
     def _result_to_get_result(self, result) -> GetResult:
         ids = []
@@ -56,7 +64,10 @@ class MilvusClient:
 
             for item in match:
                 _ids.append(item.get("id"))
-                _distances.append(item.get("distance"))
+                # normalize milvus score from [-1, 1] to [0, 1] range
+                # https://milvus.io/docs/de/metric.md
+                _dist = (item.get("distance") + 1.0) / 2.0
+                _distances.append(_dist)
                 _documents.append(item.get("entity", {}).get("data", {}).get("text"))
                 _metadatas.append(item.get("entity", {}).get("metadata"))
 
@@ -164,7 +175,7 @@ class MilvusClient:
         try:
             # Loop until there are no more items to fetch or the desired limit is reached
             while remaining > 0:
-                print("remaining", remaining)
+                log.info(f"remaining: {remaining}")
                 current_fetch = min(
                     max_limit, remaining
                 )  # Determine how many items to fetch in this iteration
@@ -191,10 +202,12 @@ class MilvusClient:
                 if results_count < current_fetch:
                     break
 
-            print(all_results)
+            log.debug(all_results)
             return self._result_to_get_result([all_results])
         except Exception as e:
-            print(e)
+            log.exception(
+                f"Error querying collection {collection_name} with limit {limit}: {e}"
+            )
             return None
 
     def get(self, collection_name: str) -> Optional[GetResult]:

backend/open_webui/retrieval/vector/dbs/opensearch.py

@@ -1,4 +1,5 @@
 from opensearchpy import OpenSearch
+from opensearchpy.helpers import bulk
 from typing import Optional
 
 from open_webui.retrieval.vector.main import VectorItem, SearchResult, GetResult
@@ -21,7 +22,13 @@ class OpenSearchClient:
             http_auth=(OPENSEARCH_USERNAME, OPENSEARCH_PASSWORD),
         )
 
+    def _get_index_name(self, collection_name: str) -> str:
+        return f"{self.index_prefix}_{collection_name}"
+
     def _result_to_get_result(self, result) -> GetResult:
+        if not result["hits"]["hits"]:
+            return None
+
         ids = []
         documents = []
         metadatas = []
@@ -31,9 +38,12 @@ class OpenSearchClient:
             documents.append(hit["_source"].get("text"))
             metadatas.append(hit["_source"].get("metadata"))
 
-        return GetResult(ids=ids, documents=documents, metadatas=metadatas)
+        return GetResult(ids=[ids], documents=[documents], metadatas=[metadatas])
 
     def _result_to_search_result(self, result) -> SearchResult:
+        if not result["hits"]["hits"]:
+            return None
+
         ids = []
         distances = []
         documents = []
@@ -46,131 +56,201 @@ class OpenSearchClient:
             metadatas.append(hit["_source"].get("metadata"))
 
         return SearchResult(
-            ids=ids, distances=distances, documents=documents, metadatas=metadatas
+            ids=[ids],
+            distances=[distances],
+            documents=[documents],
+            metadatas=[metadatas],
         )
 
-    def _create_index(self, index_name: str, dimension: int):
+    def _create_index(self, collection_name: str, dimension: int):
         body = {
+            "settings": {"index": {"knn": True}},
             "mappings": {
                 "properties": {
                     "id": {"type": "keyword"},
                     "vector": {
-                        "type": "dense_vector",
-                        "dims": dimension,  # Adjust based on your vector dimensions
-                        "index": true,
+                        "type": "knn_vector",
+                        "dimension": dimension,  # Adjust based on your vector dimensions
+                        "index": True,
                         "similarity": "faiss",
                         "method": {
                             "name": "hnsw",
-                            "space_type": "ip",  # Use inner product to approximate cosine similarity
+                            "space_type": "innerproduct",  # Use inner product to approximate cosine similarity
                             "engine": "faiss",
-                            "ef_construction": 128,
-                            "m": 16,
+                            "parameters": {
+                                "ef_construction": 128,
+                                "m": 16,
+                            },
                         },
                     },
                     "text": {"type": "text"},
                     "metadata": {"type": "object"},
                 }
-            }
+            },
         }
-        self.client.indices.create(index=f"{self.index_prefix}_{index_name}", body=body)
+        self.client.indices.create(
+            index=self._get_index_name(collection_name), body=body
+        )
 
     def _create_batches(self, items: list[VectorItem], batch_size=100):
         for i in range(0, len(items), batch_size):
             yield items[i : i + batch_size]
 
-    def has_collection(self, index_name: str) -> bool:
+    def has_collection(self, collection_name: str) -> bool:
         # has_collection here means has index.
         # We are simply adapting to the norms of the other DBs.
-        return self.client.indices.exists(index=f"{self.index_prefix}_{index_name}")
+        return self.client.indices.exists(index=self._get_index_name(collection_name))
 
-    def delete_colleciton(self, index_name: str):
+    def delete_collection(self, collection_name: str):
         # delete_collection here means delete index.
         # We are simply adapting to the norms of the other DBs.
-        self.client.indices.delete(index=f"{self.index_prefix}_{index_name}")
+        self.client.indices.delete(index=self._get_index_name(collection_name))
 
     def search(
-        self, index_name: str, vectors: list[list[float]], limit: int
+        self, collection_name: str, vectors: list[list[float | int]], limit: int
     ) -> Optional[SearchResult]:
-        query = {
-            "size": limit,
+        try:
+            if not self.has_collection(collection_name):
+                return None
+
+            query = {
+                "size": limit,
+                "_source": ["text", "metadata"],
+                "query": {
+                    "script_score": {
+                        "query": {"match_all": {}},
+                        "script": {
+                            "source": "(cosineSimilarity(params.query_value, doc[params.field]) + 1.0) / 2.0",
+                            "params": {
+                                "field": "vector",
+                                "query_value": vectors[0],
+                            },  # Assuming single query vector
+                        },
+                    }
+                },
+            }
+
+            result = self.client.search(
+                index=self._get_index_name(collection_name), body=query
+            )
+
+            return self._result_to_search_result(result)
+
+        except Exception as e:
+            return None
+
+    def query(
+        self, collection_name: str, filter: dict, limit: Optional[int] = None
+    ) -> Optional[GetResult]:
+        if not self.has_collection(collection_name):
+            return None
+
+        query_body = {
+            "query": {"bool": {"filter": []}},
             "_source": ["text", "metadata"],
-            "query": {
-                "script_score": {
-                    "query": {"match_all": {}},
-                    "script": {
-                        "source": "cosineSimilarity(params.vector, 'vector') + 1.0",
-                        "params": {
-                            "vector": vectors[0]
-                        },  # Assuming single query vector
-                    },
-                }
-            },
         }
 
-        result = self.client.search(
-            index=f"{self.index_prefix}_{index_name}", body=query
-        )
+        for field, value in filter.items():
+            query_body["query"]["bool"]["filter"].append(
+                {"match": {"metadata." + str(field): value}}
+            )
 
-        return self._result_to_search_result(result)
+        size = limit if limit else 10
 
-    def get_or_create_index(self, index_name: str, dimension: int):
-        if not self.has_index(index_name):
-            self._create_index(index_name, dimension)
+        try:
+            result = self.client.search(
+                index=self._get_index_name(collection_name),
+                body=query_body,
+                size=size,
+            )
 
-    def get(self, index_name: str) -> Optional[GetResult]:
+            return self._result_to_get_result(result)
+
+        except Exception as e:
+            return None
+
+    def _create_index_if_not_exists(self, collection_name: str, dimension: int):
+        if not self.has_collection(collection_name):
+            self._create_index(collection_name, dimension)
+
+    def get(self, collection_name: str) -> Optional[GetResult]:
         query = {"query": {"match_all": {}}, "_source": ["text", "metadata"]}
 
         result = self.client.search(
-            index=f"{self.index_prefix}_{index_name}", body=query
+            index=self._get_index_name(collection_name), body=query
         )
         return self._result_to_get_result(result)
 
-    def insert(self, index_name: str, items: list[VectorItem]):
-        if not self.has_index(index_name):
-            self._create_index(index_name, dimension=len(items[0]["vector"]))
+    def insert(self, collection_name: str, items: list[VectorItem]):
+        self._create_index_if_not_exists(
+            collection_name=collection_name, dimension=len(items[0]["vector"])
+        )
 
         for batch in self._create_batches(items):
             actions = [
                 {
-                    "index": {
-                        "_id": item["id"],
-                        "_source": {
-                            "vector": item["vector"],
-                            "text": item["text"],
-                            "metadata": item["metadata"],
-                        },
-                    }
+                    "_op_type": "index",
+                    "_index": self._get_index_name(collection_name),
+                    "_id": item["id"],
+                    "_source": {
+                        "vector": item["vector"],
+                        "text": item["text"],
+                        "metadata": item["metadata"],
+                    },
                 }
                 for item in batch
             ]
-            self.client.bulk(actions)
+            bulk(self.client, actions)
 
-    def upsert(self, index_name: str, items: list[VectorItem]):
-        if not self.has_index(index_name):
-            self._create_index(index_name, dimension=len(items[0]["vector"]))
+    def upsert(self, collection_name: str, items: list[VectorItem]):
+        self._create_index_if_not_exists(
+            collection_name=collection_name, dimension=len(items[0]["vector"])
+        )
 
         for batch in self._create_batches(items):
             actions = [
                 {
-                    "index": {
-                        "_id": item["id"],
-                        "_source": {
-                            "vector": item["vector"],
-                            "text": item["text"],
-                            "metadata": item["metadata"],
-                        },
-                    }
+                    "_op_type": "update",
+                    "_index": self._get_index_name(collection_name),
+                    "_id": item["id"],
+                    "doc": {
+                        "vector": item["vector"],
+                        "text": item["text"],
+                        "metadata": item["metadata"],
+                    },
+                    "doc_as_upsert": True,
                 }
                 for item in batch
             ]
-            self.client.bulk(actions)
+            bulk(self.client, actions)
 
-    def delete(self, index_name: str, ids: list[str]):
-        actions = [
-            {"delete": {"_index": f"{self.index_prefix}_{index_name}", "_id": id}}
-            for id in ids
-        ]
-        self.client.bulk(body=actions)
+    def delete(
+        self,
+        collection_name: str,
+        ids: Optional[list[str]] = None,
+        filter: Optional[dict] = None,
+    ):
+        if ids:
+            actions = [
+                {
+                    "_op_type": "delete",
+                    "_index": self._get_index_name(collection_name),
+                    "_id": id,
+                }
+                for id in ids
+            ]
+            bulk(self.client, actions)
+        elif filter:
+            query_body = {
+                "query": {"bool": {"filter": []}},
+            }
+            for field, value in filter.items():
+                query_body["query"]["bool"]["filter"].append(
+                    {"match": {"metadata." + str(field): value}}
+                )
+            self.client.delete_by_query(
+                index=self._get_index_name(collection_name), body=query_body
+            )
 
     def reset(self):
         indices = self.client.indices.get(index=f"{self.index_prefix}_*")

backend/open_webui/retrieval/vector/dbs/pgvector.py

@@ -1,4 +1,5 @@
 from typing import Optional, List, Dict, Any
+import logging
 from sqlalchemy import (
     cast,
     column,
@@ -24,9 +25,14 @@ from sqlalchemy.exc import NoSuchTableError
 from open_webui.retrieval.vector.main import VectorItem, SearchResult, GetResult
 from open_webui.config import PGVECTOR_DB_URL, PGVECTOR_INITIALIZE_MAX_VECTOR_LENGTH
 
+from open_webui.env import SRC_LOG_LEVELS
+
 VECTOR_LENGTH = PGVECTOR_INITIALIZE_MAX_VECTOR_LENGTH
 Base = declarative_base()
 
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
 
 class DocumentChunk(Base):
     __tablename__ = "document_chunk"
@@ -82,10 +88,10 @@ class PgvectorClient:
                 )
             )
             self.session.commit()
-            print("Initialization complete.")
+            log.info("Initialization complete.")
         except Exception as e:
             self.session.rollback()
-            print(f"Error during initialization: {e}")
+            log.exception(f"Error during initialization: {e}")
             raise
 
     def check_vector_length(self) -> None:
@@ -150,12 +156,12 @@ class PgvectorClient:
                 new_items.append(new_chunk)
             self.session.bulk_save_objects(new_items)
             self.session.commit()
-            print(
+            log.info(
                 f"Inserted {len(new_items)} items into collection '{collection_name}'."
             )
         except Exception as e:
             self.session.rollback()
-            print(f"Error during insert: {e}")
+            log.exception(f"Error during insert: {e}")
             raise
 
     def upsert(self, collection_name: str, items: List[VectorItem]) -> None:
@@ -184,10 +190,12 @@ class PgvectorClient:
                     )
                     self.session.add(new_chunk)
             self.session.commit()
-            print(f"Upserted {len(items)} items into collection '{collection_name}'.")
+            log.info(
+                f"Upserted {len(items)} items into collection '{collection_name}'."
+            )
         except Exception as e:
             self.session.rollback()
-            print(f"Error during upsert: {e}")
+            log.exception(f"Error during upsert: {e}")
             raise
 
     def search(
@@ -270,7 +278,9 @@ class PgvectorClient:
             for row in results:
                 qid = int(row.qid)
                 ids[qid].append(row.id)
-                distances[qid].append(row.distance)
+                # normalize and re-orders pgvec distance from [2, 0] to [0, 1] score range
+                # https://github.com/pgvector/pgvector?tab=readme-ov-file#querying
+                distances[qid].append((2.0 - row.distance) / 2.0)
                 documents[qid].append(row.text)
                 metadatas[qid].append(row.vmetadata)
 
@@ -278,7 +288,7 @@ class PgvectorClient:
                 ids=ids, distances=distances, documents=documents, metadatas=metadatas
             )
         except Exception as e:
-            print(f"Error during search: {e}")
+            log.exception(f"Error during search: {e}")
             return None
 
     def query(
@@ -310,7 +320,7 @@ class PgvectorClient:
                 metadatas=metadatas,
             )
         except Exception as e:
-            print(f"Error during query: {e}")
+            log.exception(f"Error during query: {e}")
             return None
 
     def get(
@@ -334,7 +344,7 @@ class PgvectorClient:
 
             return GetResult(ids=ids, documents=documents, metadatas=metadatas)
         except Exception as e:
-            print(f"Error during get: {e}")
+            log.exception(f"Error during get: {e}")
             return None
 
     def delete(
@@ -356,22 +366,22 @@ class PgvectorClient:
                     )
             deleted = query.delete(synchronize_session=False)
             self.session.commit()
-            print(f"Deleted {deleted} items from collection '{collection_name}'.")
+            log.info(f"Deleted {deleted} items from collection '{collection_name}'.")
         except Exception as e:
             self.session.rollback()
-            print(f"Error during delete: {e}")
+            log.exception(f"Error during delete: {e}")
             raise
 
     def reset(self) -> None:
         try:
             deleted = self.session.query(DocumentChunk).delete()
             self.session.commit()
-            print(
+            log.info(
                 f"Reset complete. Deleted {deleted} items from 'document_chunk' table."
             )
         except Exception as e:
             self.session.rollback()
-            print(f"Error during reset: {e}")
+            log.exception(f"Error during reset: {e}")
             raise
 
     def close(self) -> None:
@@ -387,9 +397,9 @@ class PgvectorClient:
             )
             return exists
         except Exception as e:
-            print(f"Error checking collection existence: {e}")
+            log.exception(f"Error checking collection existence: {e}")
             return False
 
     def delete_collection(self, collection_name: str) -> None:
         self.delete(collection_name)
-        print(f"Collection '{collection_name}' deleted.")
+        log.info(f"Collection '{collection_name}' deleted.")

backend/open_webui/retrieval/vector/dbs/qdrant.py

@@ -1,4 +1,5 @@
 from typing import Optional
+import logging
 
 from qdrant_client import QdrantClient as Qclient
 from qdrant_client.http.models import PointStruct
@@ -6,9 +7,13 @@ from qdrant_client.models import models
 
 from open_webui.retrieval.vector.main import VectorItem, SearchResult, GetResult
 from open_webui.config import QDRANT_URI, QDRANT_API_KEY
+from open_webui.env import SRC_LOG_LEVELS
 
 NO_LIMIT = 999999999
 
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
 
 class QdrantClient:
     def __init__(self):
@@ -49,7 +54,7 @@ class QdrantClient:
             ),
         )
 
-        print(f"collection {collection_name_with_prefix} successfully created!")
+        log.info(f"collection {collection_name_with_prefix} successfully created!")
 
     def _create_collection_if_not_exists(self, collection_name, dimension):
         if not self.has_collection(collection_name=collection_name):
@@ -94,7 +99,8 @@ class QdrantClient:
             ids=get_result.ids,
             documents=get_result.documents,
             metadatas=get_result.metadatas,
-            distances=[[point.score for point in query_response.points]],
+            # qdrant distance is [-1, 1], normalize to [0, 1]
+            distances=[[(point.score + 1.0) / 2.0 for point in query_response.points]],
         )
 
     def query(self, collection_name: str, filter: dict, limit: Optional[int] = None):
@@ -120,7 +126,7 @@ class QdrantClient:
             )
             return self._result_to_get_result(points.points)
         except Exception as e:
-            print(e)
+            log.exception(f"Error querying a collection '{collection_name}': {e}")
             return None
 
     def get(self, collection_name: str) -> Optional[GetResult]:

backend/open_webui/retrieval/web/bocha.py

@@ -0,0 +1,65 @@
+import logging
+from typing import Optional
+
+import requests
+import json
+from open_webui.retrieval.web.main import SearchResult, get_filtered_results
+from open_webui.env import SRC_LOG_LEVELS
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
+
+def _parse_response(response):
+    result = {}
+    if "data" in response:
+        data = response["data"]
+        if "webPages" in data:
+            webPages = data["webPages"]
+            if "value" in webPages:
+                result["webpage"] = [
+                    {
+                        "id": item.get("id", ""),
+                        "name": item.get("name", ""),
+                        "url": item.get("url", ""),
+                        "snippet": item.get("snippet", ""),
+                        "summary": item.get("summary", ""),
+                        "siteName": item.get("siteName", ""),
+                        "siteIcon": item.get("siteIcon", ""),
+                        "datePublished": item.get("datePublished", "")
+                        or item.get("dateLastCrawled", ""),
+                    }
+                    for item in webPages["value"]
+                ]
+    return result
+
+
+def search_bocha(
+    api_key: str, query: str, count: int, filter_list: Optional[list[str]] = None
+) -> list[SearchResult]:
+    """Search using Bocha's Search API and return the results as a list of SearchResult objects.
+
+    Args:
+        api_key (str): A Bocha Search API key
+        query (str): The query to search for
+    """
+    url = "https://api.bochaai.com/v1/web-search?utm_source=ollama"
+    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
+
+    payload = json.dumps(
+        {"query": query, "summary": True, "freshness": "noLimit", "count": count}
+    )
+
+    response = requests.post(url, headers=headers, data=payload, timeout=5)
+    response.raise_for_status()
+    results = _parse_response(response.json())
+    print(results)
+    if filter_list:
+        results = get_filtered_results(results, filter_list)
+
+    return [
+        SearchResult(
+            link=result["url"], title=result.get("name"), snippet=result.get("summary")
+        )
+        for result in results.get("webpage", [])[:count]
+    ]

backend/open_webui/retrieval/web/duckduckgo.py

@@ -3,6 +3,7 @@ from typing import Optional
 
 from open_webui.retrieval.web.main import SearchResult, get_filtered_results
 from duckduckgo_search import DDGS
+from duckduckgo_search.exceptions import RatelimitException
 from open_webui.env import SRC_LOG_LEVELS
 
 log = logging.getLogger(__name__)
@@ -22,29 +23,24 @@ def search_duckduckgo(
         list[SearchResult]: A list of search results
     """
     # Use the DDGS context manager to create a DDGS object
+    search_results = []
     with DDGS() as ddgs:
         # Use the ddgs.text() method to perform the search
-        ddgs_gen = ddgs.text(
-            query, safesearch="moderate", max_results=count, backend="api"
-        )
-        # Check if there are search results
-        if ddgs_gen:
-            # Convert the search results into a list
-            search_results = [r for r in ddgs_gen]
-
-    # Create an empty list to store the SearchResult objects
-    results = []
-    # Iterate over each search result
-    for result in search_results:
-        # Create a SearchResult object and append it to the results list
-        results.append(
-            SearchResult(
-                link=result["href"],
-                title=result.get("title"),
-                snippet=result.get("body"),
+        try:
+            search_results = ddgs.text(
+                query, safesearch="moderate", max_results=count, backend="lite"
             )
-        )
+        except RatelimitException as e:
+            log.error(f"RatelimitException: {e}")
     if filter_list:
-        results = get_filtered_results(results, filter_list)
+        search_results = get_filtered_results(search_results, filter_list)
+
     # Return the list of search results
-    return results
+    return [
+        SearchResult(
+            link=result["href"],
+            title=result.get("title"),
+            snippet=result.get("body"),
+        )
+        for result in search_results
+    ]

backend/open_webui/retrieval/web/exa.py

@@ -0,0 +1,76 @@
+import logging
+from dataclasses import dataclass
+from typing import Optional
+
+import requests
+from open_webui.env import SRC_LOG_LEVELS
+from open_webui.retrieval.web.main import SearchResult
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
+EXA_API_BASE = "https://api.exa.ai"
+
+
+@dataclass
+class ExaResult:
+    url: str
+    title: str
+    text: str
+
+
+def search_exa(
+    api_key: str,
+    query: str,
+    count: int,
+    filter_list: Optional[list[str]] = None,
+) -> list[SearchResult]:
+    """Search using Exa Search API and return the results as a list of SearchResult objects.
+
+    Args:
+        api_key (str): A Exa Search API key
+        query (str): The query to search for
+        count (int): Number of results to return
+        filter_list (Optional[list[str]]): List of domains to filter results by
+    """
+    log.info(f"Searching with Exa for query: {query}")
+
+    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
+
+    payload = {
+        "query": query,
+        "numResults": count or 5,
+        "includeDomains": filter_list,
+        "contents": {"text": True, "highlights": True},
+        "type": "auto",  # Use the auto search type (keyword or neural)
+    }
+
+    try:
+        response = requests.post(
+            f"{EXA_API_BASE}/search", headers=headers, json=payload
+        )
+        response.raise_for_status()
+        data = response.json()
+
+        results = []
+        for result in data["results"]:
+            results.append(
+                ExaResult(
+                    url=result["url"],
+                    title=result["title"],
+                    text=result["text"],
+                )
+            )
+
+        log.info(f"Found {len(results)} results")
+        return [
+            SearchResult(
+                link=result.url,
+                title=result.title,
+                snippet=result.text,
+            )
+            for result in results
+        ]
+    except Exception as e:
+        log.error(f"Error searching Exa: {e}")
+        return []

backend/open_webui/retrieval/web/google_pse.py

@@ -17,34 +17,53 @@ def search_google_pse(
     filter_list: Optional[list[str]] = None,
 ) -> list[SearchResult]:
     """Search using Google's Programmable Search Engine API and return the results as a list of SearchResult objects.
+    Handles pagination for counts greater than 10.
 
     Args:
         api_key (str): A Programmable Search Engine API key
         search_engine_id (str): A Programmable Search Engine ID
         query (str): The query to search for
+        count (int): The number of results to return (max 100, as PSE max results per query is 10 and max page is 10)
+        filter_list (Optional[list[str]], optional): A list of keywords to filter out from results. Defaults to None.
+
+    Returns:
+        list[SearchResult]: A list of SearchResult objects.
     """
     url = "https://www.googleapis.com/customsearch/v1"
-
     headers = {"Content-Type": "application/json"}
-    params = {
-        "cx": search_engine_id,
-        "q": query,
-        "key": api_key,
-        "num": count,
-    }
+    all_results = []
+    start_index = 1  # Google PSE start parameter is 1-based
 
-    response = requests.request("GET", url, headers=headers, params=params)
-    response.raise_for_status()
+    while count > 0:
+        num_results_this_page = min(count, 10)  # Google PSE max results per page is 10
+        params = {
+            "cx": search_engine_id,
+            "q": query,
+            "key": api_key,
+            "num": num_results_this_page,
+            "start": start_index,
+        }
+        response = requests.request("GET", url, headers=headers, params=params)
+        response.raise_for_status()
+        json_response = response.json()
+        results = json_response.get("items", [])
+        if results:  # check if results are returned. If not, no more pages to fetch.
+            all_results.extend(results)
+            count -= len(
+                results
+            )  # Decrement count by the number of results fetched in this page.
+            start_index += 10  # Increment start index for the next page
+        else:
+            break  # No more results from Google PSE, break the loop
 
-    json_response = response.json()
-    results = json_response.get("items", [])
     if filter_list:
-        results = get_filtered_results(results, filter_list)
+        all_results = get_filtered_results(all_results, filter_list)
+
     return [
         SearchResult(
             link=result["link"],
             title=result.get("title"),
             snippet=result.get("snippet"),
         )
-        for result in results
+        for result in all_results
     ]

backend/open_webui/retrieval/web/jina_search.py

@@ -20,14 +20,23 @@ def search_jina(api_key: str, query: str, count: int) -> list[SearchResult]:
         list[SearchResult]: A list of search results
     """
     jina_search_endpoint = "https://s.jina.ai/"
-    headers = {"Accept": "application/json", "Authorization": f"Bearer {api_key}"}
-    url = str(URL(jina_search_endpoint + query))
-    response = requests.get(url, headers=headers)
+
+    headers = {
+        "Accept": "application/json",
+        "Content-Type": "application/json",
+        "Authorization": api_key,
+        "X-Retain-Images": "none",
+    }
+
+    payload = {"q": query, "count": count if count <= 10 else 10}
+
+    url = str(URL(jina_search_endpoint))
+    response = requests.post(url, headers=headers, json=payload)
     response.raise_for_status()
     data = response.json()
 
     results = []
-    for result in data["data"][:count]:
+    for result in data["data"]:
         results.append(
             SearchResult(
                 link=result["url"],

backend/open_webui/retrieval/web/main.py

@@ -1,3 +1,5 @@
+import validators
+
 from typing import Optional
 from urllib.parse import urlparse
 
@@ -10,6 +12,8 @@ def get_filtered_results(results, filter_list):
     filtered_results = []
     for result in results:
         url = result.get("url") or result.get("link", "")
+        if not validators.url(url):
+            continue
         domain = urlparse(url).netloc
         if any(domain.endswith(filtered_domain) for filtered_domain in filter_list):
             filtered_results.append(result)

backend/open_webui/retrieval/web/perplexity.py

@@ -0,0 +1,87 @@
+import logging
+from typing import Optional, List
+import requests
+
+from open_webui.retrieval.web.main import SearchResult, get_filtered_results
+from open_webui.env import SRC_LOG_LEVELS
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
+
+def search_perplexity(
+    api_key: str,
+    query: str,
+    count: int,
+    filter_list: Optional[list[str]] = None,
+) -> list[SearchResult]:
+    """Search using Perplexity API and return the results as a list of SearchResult objects.
+
+    Args:
+      api_key (str): A Perplexity API key
+      query (str): The query to search for
+      count (int): Maximum number of results to return
+
+    """
+
+    # Handle PersistentConfig object
+    if hasattr(api_key, "__str__"):
+        api_key = str(api_key)
+
+    try:
+        url = "https://api.perplexity.ai/chat/completions"
+
+        # Create payload for the API call
+        payload = {
+            "model": "sonar",
+            "messages": [
+                {
+                    "role": "system",
+                    "content": "You are a search assistant. Provide factual information with citations.",
+                },
+                {"role": "user", "content": query},
+            ],
+            "temperature": 0.2,  # Lower temperature for more factual responses
+            "stream": False,
+        }
+
+        headers = {
+            "Authorization": f"Bearer {api_key}",
+            "Content-Type": "application/json",
+        }
+
+        # Make the API request
+        response = requests.request("POST", url, json=payload, headers=headers)
+
+        # Parse the JSON response
+        json_response = response.json()
+
+        # Extract citations from the response
+        citations = json_response.get("citations", [])
+
+        # Create search results from citations
+        results = []
+        for i, citation in enumerate(citations[:count]):
+            # Extract content from the response to use as snippet
+            content = ""
+            if "choices" in json_response and json_response["choices"]:
+                if i == 0:
+                    content = json_response["choices"][0]["message"]["content"]
+
+            result = {"link": citation, "title": f"Source {i+1}", "snippet": content}
+            results.append(result)
+
+        if filter_list:
+
+            results = get_filtered_results(results, filter_list)
+
+        return [
+            SearchResult(
+                link=result["link"], title=result["title"], snippet=result["snippet"]
+            )
+            for result in results[:count]
+        ]
+
+    except Exception as e:
+        log.error(f"Error searching with Perplexity API: {e}")
+        return []

backend/open_webui/retrieval/web/serpapi.py

@@ -0,0 +1,48 @@
+import logging
+from typing import Optional
+from urllib.parse import urlencode
+
+import requests
+from open_webui.retrieval.web.main import SearchResult, get_filtered_results
+from open_webui.env import SRC_LOG_LEVELS
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
+
+def search_serpapi(
+    api_key: str,
+    engine: str,
+    query: str,
+    count: int,
+    filter_list: Optional[list[str]] = None,
+) -> list[SearchResult]:
+    """Search using serpapi.com's API and return the results as a list of SearchResult objects.
+
+    Args:
+      api_key (str): A serpapi.com API key
+      query (str): The query to search for
+    """
+    url = "https://serpapi.com/search"
+
+    engine = engine or "google"
+
+    payload = {"engine": engine, "q": query, "api_key": api_key}
+
+    url = f"{url}?{urlencode(payload)}"
+    response = requests.request("GET", url)
+
+    json_response = response.json()
+    log.info(f"results from serpapi search: {json_response}")
+
+    results = sorted(
+        json_response.get("organic_results", []), key=lambda x: x.get("position", 0)
+    )
+    if filter_list:
+        results = get_filtered_results(results, filter_list)
+    return [
+        SearchResult(
+            link=result["link"], title=result["title"], snippet=result["snippet"]
+        )
+        for result in results[:count]
+    ]

backend/open_webui/retrieval/web/sougou.py

@@ -0,0 +1,60 @@
+import logging
+import json
+from typing import Optional, List
+
+
+from open_webui.retrieval.web.main import SearchResult, get_filtered_results
+from open_webui.env import SRC_LOG_LEVELS
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
+
+def search_sougou(
+    sougou_api_sid: str,
+    sougou_api_sk: str,
+    query: str,
+    count: int,
+    filter_list: Optional[List[str]] = None,
+) -> List[SearchResult]:
+    from tencentcloud.common.common_client import CommonClient
+    from tencentcloud.common import credential
+    from tencentcloud.common.exception.tencent_cloud_sdk_exception import (
+        TencentCloudSDKException,
+    )
+    from tencentcloud.common.profile.client_profile import ClientProfile
+    from tencentcloud.common.profile.http_profile import HttpProfile
+
+    try:
+        cred = credential.Credential(sougou_api_sid, sougou_api_sk)
+        http_profile = HttpProfile()
+        http_profile.endpoint = "tms.tencentcloudapi.com"
+        client_profile = ClientProfile()
+        client_profile.http_profile = http_profile
+        params = json.dumps({"Query": query, "Cnt": 20})
+        common_client = CommonClient(
+            "tms", "2020-12-29", cred, "", profile=client_profile
+        )
+        results = [
+            json.loads(page)
+            for page in common_client.call_json("SearchPro", json.loads(params))[
+                "Response"
+            ]["Pages"]
+        ]
+        sorted_results = sorted(
+            results, key=lambda x: x.get("scour", 0.0), reverse=True
+        )
+        if filter_list:
+            sorted_results = get_filtered_results(sorted_results, filter_list)
+
+        return [
+            SearchResult(
+                link=result.get("url"),
+                title=result.get("title"),
+                snippet=result.get("passage"),
+            )
+            for result in sorted_results[:count]
+        ]
+    except TencentCloudSDKException as err:
+        log.error(f"Error in Sougou search: {err}")
+        return []

backend/open_webui/retrieval/web/tavily.py

@@ -1,4 +1,5 @@
 import logging
+from typing import Optional
 
 import requests
 from open_webui.retrieval.web.main import SearchResult
@@ -8,7 +9,13 @@ log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["RAG"])
 
 
-def search_tavily(api_key: str, query: str, count: int) -> list[SearchResult]:
+def search_tavily(
+    api_key: str,
+    query: str,
+    count: int,
+    filter_list: Optional[list[str]] = None,
+    # **kwargs,
+) -> list[SearchResult]:
     """Search using Tavily's Search API and return the results as a list of SearchResult objects.
 
     Args:
@@ -20,7 +27,6 @@ def search_tavily(api_key: str, query: str, count: int) -> list[SearchResult]:
     """
     url = "https://api.tavily.com/search"
     data = {"query": query, "api_key": api_key}
-
     response = requests.post(url, json=data)
     response.raise_for_status()
 

backend/open_webui/retrieval/web/utils.py

@@ -1,19 +1,42 @@
-import socket
-import urllib.parse
-import validators
-from typing import Union, Sequence, Iterator
-
-from langchain_community.document_loaders import (
-    WebBaseLoader,
-)
-from langchain_core.documents import Document
-
-
-from open_webui.constants import ERROR_MESSAGES
-from open_webui.config import ENABLE_RAG_LOCAL_WEB_FETCH
-from open_webui.env import SRC_LOG_LEVELS
-
+import asyncio
 import logging
+import socket
+import ssl
+import urllib.parse
+import urllib.request
+from collections import defaultdict
+from datetime import datetime, time, timedelta
+from typing import (
+    Any,
+    AsyncIterator,
+    Dict,
+    Iterator,
+    List,
+    Optional,
+    Sequence,
+    Union,
+    Literal,
+)
+import aiohttp
+import certifi
+import validators
+from langchain_community.document_loaders import PlaywrightURLLoader, WebBaseLoader
+from langchain_community.document_loaders.firecrawl import FireCrawlLoader
+from langchain_community.document_loaders.base import BaseLoader
+from langchain_core.documents import Document
+from open_webui.retrieval.loaders.tavily import TavilyLoader
+from open_webui.constants import ERROR_MESSAGES
+from open_webui.config import (
+    ENABLE_RAG_LOCAL_WEB_FETCH,
+    PLAYWRIGHT_WS_URL,
+    PLAYWRIGHT_TIMEOUT,
+    WEB_LOADER_ENGINE,
+    FIRECRAWL_API_BASE_URL,
+    FIRECRAWL_API_KEY,
+    TAVILY_API_KEY,
+    TAVILY_EXTRACT_DEPTH,
+)
+from open_webui.env import SRC_LOG_LEVELS
 
 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["RAG"])
@@ -43,6 +66,17 @@ def validate_url(url: Union[str, Sequence[str]]):
         return False
 
 
+def safe_validate_urls(url: Sequence[str]) -> Sequence[str]:
+    valid_urls = []
+    for u in url:
+        try:
+            if validate_url(u):
+                valid_urls.append(u)
+        except ValueError:
+            continue
+    return valid_urls
+
+
 def resolve_hostname(hostname):
     # Get address information
     addr_info = socket.getaddrinfo(hostname, None)
@@ -54,9 +88,466 @@ def resolve_hostname(hostname):
     return ipv4_addresses, ipv6_addresses
 
 
+def extract_metadata(soup, url):
+    metadata = {"source": url}
+    if title := soup.find("title"):
+        metadata["title"] = title.get_text()
+    if description := soup.find("meta", attrs={"name": "description"}):
+        metadata["description"] = description.get("content", "No description found.")
+    if html := soup.find("html"):
+        metadata["language"] = html.get("lang", "No language found.")
+    return metadata
+
+
+def verify_ssl_cert(url: str) -> bool:
+    """Verify SSL certificate for the given URL."""
+    if not url.startswith("https://"):
+        return True
+
+    try:
+        hostname = url.split("://")[-1].split("/")[0]
+        context = ssl.create_default_context(cafile=certifi.where())
+        with context.wrap_socket(ssl.socket(), server_hostname=hostname) as s:
+            s.connect((hostname, 443))
+        return True
+    except ssl.SSLError:
+        return False
+    except Exception as e:
+        log.warning(f"SSL verification failed for {url}: {str(e)}")
+        return False
+
+
+class RateLimitMixin:
+    async def _wait_for_rate_limit(self):
+        """Wait to respect the rate limit if specified."""
+        if self.requests_per_second and self.last_request_time:
+            min_interval = timedelta(seconds=1.0 / self.requests_per_second)
+            time_since_last = datetime.now() - self.last_request_time
+            if time_since_last < min_interval:
+                await asyncio.sleep((min_interval - time_since_last).total_seconds())
+        self.last_request_time = datetime.now()
+
+    def _sync_wait_for_rate_limit(self):
+        """Synchronous version of rate limit wait."""
+        if self.requests_per_second and self.last_request_time:
+            min_interval = timedelta(seconds=1.0 / self.requests_per_second)
+            time_since_last = datetime.now() - self.last_request_time
+            if time_since_last < min_interval:
+                time.sleep((min_interval - time_since_last).total_seconds())
+        self.last_request_time = datetime.now()
+
+
+class URLProcessingMixin:
+    def _verify_ssl_cert(self, url: str) -> bool:
+        """Verify SSL certificate for a URL."""
+        return verify_ssl_cert(url)
+
+    async def _safe_process_url(self, url: str) -> bool:
+        """Perform safety checks before processing a URL."""
+        if self.verify_ssl and not self._verify_ssl_cert(url):
+            raise ValueError(f"SSL certificate verification failed for {url}")
+        await self._wait_for_rate_limit()
+        return True
+
+    def _safe_process_url_sync(self, url: str) -> bool:
+        """Synchronous version of safety checks."""
+        if self.verify_ssl and not self._verify_ssl_cert(url):
+            raise ValueError(f"SSL certificate verification failed for {url}")
+        self._sync_wait_for_rate_limit()
+        return True
+
+
+class SafeFireCrawlLoader(BaseLoader, RateLimitMixin, URLProcessingMixin):
+    def __init__(
+        self,
+        web_paths,
+        verify_ssl: bool = True,
+        trust_env: bool = False,
+        requests_per_second: Optional[float] = None,
+        continue_on_failure: bool = True,
+        api_key: Optional[str] = None,
+        api_url: Optional[str] = None,
+        mode: Literal["crawl", "scrape", "map"] = "crawl",
+        proxy: Optional[Dict[str, str]] = None,
+        params: Optional[Dict] = None,
+    ):
+        """Concurrent document loader for FireCrawl operations.
+
+        Executes multiple FireCrawlLoader instances concurrently using thread pooling
+        to improve bulk processing efficiency.
+        Args:
+            web_paths: List of URLs/paths to process.
+            verify_ssl: If True, verify SSL certificates.
+            trust_env: If True, use proxy settings from environment variables.
+            requests_per_second: Number of requests per second to limit to.
+            continue_on_failure (bool): If True, continue loading other URLs on failure.
+            api_key: API key for FireCrawl service. Defaults to None
+                (uses FIRE_CRAWL_API_KEY environment variable if not provided).
+            api_url: Base URL for FireCrawl API. Defaults to official API endpoint.
+            mode: Operation mode selection:
+                - 'crawl': Website crawling mode (default)
+                - 'scrape': Direct page scraping
+                - 'map': Site map generation
+            proxy: Proxy override settings for the FireCrawl API.
+            params: The parameters to pass to the Firecrawl API.
+                Examples include crawlerOptions.
+                For more details, visit: https://github.com/mendableai/firecrawl-py
+        """
+        proxy_server = proxy.get("server") if proxy else None
+        if trust_env and not proxy_server:
+            env_proxies = urllib.request.getproxies()
+            env_proxy_server = env_proxies.get("https") or env_proxies.get("http")
+            if env_proxy_server:
+                if proxy:
+                    proxy["server"] = env_proxy_server
+                else:
+                    proxy = {"server": env_proxy_server}
+        self.web_paths = web_paths
+        self.verify_ssl = verify_ssl
+        self.requests_per_second = requests_per_second
+        self.last_request_time = None
+        self.trust_env = trust_env
+        self.continue_on_failure = continue_on_failure
+        self.api_key = api_key
+        self.api_url = api_url
+        self.mode = mode
+        self.params = params
+
+    def lazy_load(self) -> Iterator[Document]:
+        """Load documents concurrently using FireCrawl."""
+        for url in self.web_paths:
+            try:
+                self._safe_process_url_sync(url)
+                loader = FireCrawlLoader(
+                    url=url,
+                    api_key=self.api_key,
+                    api_url=self.api_url,
+                    mode=self.mode,
+                    params=self.params,
+                )
+                yield from loader.lazy_load()
+            except Exception as e:
+                if self.continue_on_failure:
+                    log.exception(f"Error loading {url}: {e}")
+                    continue
+                raise e
+
+    async def alazy_load(self):
+        """Async version of lazy_load."""
+        for url in self.web_paths:
+            try:
+                await self._safe_process_url(url)
+                loader = FireCrawlLoader(
+                    url=url,
+                    api_key=self.api_key,
+                    api_url=self.api_url,
+                    mode=self.mode,
+                    params=self.params,
+                )
+                async for document in loader.alazy_load():
+                    yield document
+            except Exception as e:
+                if self.continue_on_failure:
+                    log.exception(f"Error loading {url}: {e}")
+                    continue
+                raise e
+
+
+class SafeTavilyLoader(BaseLoader, RateLimitMixin, URLProcessingMixin):
+    def __init__(
+        self,
+        web_paths: Union[str, List[str]],
+        api_key: str,
+        extract_depth: Literal["basic", "advanced"] = "basic",
+        continue_on_failure: bool = True,
+        requests_per_second: Optional[float] = None,
+        verify_ssl: bool = True,
+        trust_env: bool = False,
+        proxy: Optional[Dict[str, str]] = None,
+    ):
+        """Initialize SafeTavilyLoader with rate limiting and SSL verification support.
+
+        Args:
+            web_paths: List of URLs/paths to process.
+            api_key: The Tavily API key.
+            extract_depth: Depth of extraction ("basic" or "advanced").
+            continue_on_failure: Whether to continue if extraction of a URL fails.
+            requests_per_second: Number of requests per second to limit to.
+            verify_ssl: If True, verify SSL certificates.
+            trust_env: If True, use proxy settings from environment variables.
+            proxy: Optional proxy configuration.
+        """
+        # Initialize proxy configuration if using environment variables
+        proxy_server = proxy.get("server") if proxy else None
+        if trust_env and not proxy_server:
+            env_proxies = urllib.request.getproxies()
+            env_proxy_server = env_proxies.get("https") or env_proxies.get("http")
+            if env_proxy_server:
+                if proxy:
+                    proxy["server"] = env_proxy_server
+                else:
+                    proxy = {"server": env_proxy_server}
+
+        # Store parameters for creating TavilyLoader instances
+        self.web_paths = web_paths if isinstance(web_paths, list) else [web_paths]
+        self.api_key = api_key
+        self.extract_depth = extract_depth
+        self.continue_on_failure = continue_on_failure
+        self.verify_ssl = verify_ssl
+        self.trust_env = trust_env
+        self.proxy = proxy
+
+        # Add rate limiting
+        self.requests_per_second = requests_per_second
+        self.last_request_time = None
+
+    def lazy_load(self) -> Iterator[Document]:
+        """Load documents with rate limiting support, delegating to TavilyLoader."""
+        valid_urls = []
+        for url in self.web_paths:
+            try:
+                self._safe_process_url_sync(url)
+                valid_urls.append(url)
+            except Exception as e:
+                log.warning(f"SSL verification failed for {url}: {str(e)}")
+                if not self.continue_on_failure:
+                    raise e
+        if not valid_urls:
+            if self.continue_on_failure:
+                log.warning("No valid URLs to process after SSL verification")
+                return
+            raise ValueError("No valid URLs to process after SSL verification")
+        try:
+            loader = TavilyLoader(
+                urls=valid_urls,
+                api_key=self.api_key,
+                extract_depth=self.extract_depth,
+                continue_on_failure=self.continue_on_failure,
+            )
+            yield from loader.lazy_load()
+        except Exception as e:
+            if self.continue_on_failure:
+                log.exception(f"Error extracting content from URLs: {e}")
+            else:
+                raise e
+
+    async def alazy_load(self) -> AsyncIterator[Document]:
+        """Async version with rate limiting and SSL verification."""
+        valid_urls = []
+        for url in self.web_paths:
+            try:
+                await self._safe_process_url(url)
+                valid_urls.append(url)
+            except Exception as e:
+                log.warning(f"SSL verification failed for {url}: {str(e)}")
+                if not self.continue_on_failure:
+                    raise e
+
+        if not valid_urls:
+            if self.continue_on_failure:
+                log.warning("No valid URLs to process after SSL verification")
+                return
+            raise ValueError("No valid URLs to process after SSL verification")
+
+        try:
+            loader = TavilyLoader(
+                urls=valid_urls,
+                api_key=self.api_key,
+                extract_depth=self.extract_depth,
+                continue_on_failure=self.continue_on_failure,
+            )
+            async for document in loader.alazy_load():
+                yield document
+        except Exception as e:
+            if self.continue_on_failure:
+                log.exception(f"Error loading URLs: {e}")
+            else:
+                raise e
+
+
+class SafePlaywrightURLLoader(PlaywrightURLLoader, RateLimitMixin, URLProcessingMixin):
+    """Load HTML pages safely with Playwright, supporting SSL verification, rate limiting, and remote browser connection.
+
+    Attributes:
+        web_paths (List[str]): List of URLs to load.
+        verify_ssl (bool): If True, verify SSL certificates.
+        trust_env (bool): If True, use proxy settings from environment variables.
+        requests_per_second (Optional[float]): Number of requests per second to limit to.
+        continue_on_failure (bool): If True, continue loading other URLs on failure.
+        headless (bool): If True, the browser will run in headless mode.
+        proxy (dict): Proxy override settings for the Playwright session.
+        playwright_ws_url (Optional[str]): WebSocket endpoint URI for remote browser connection.
+        playwright_timeout (Optional[int]): Maximum operation time in milliseconds.
+    """
+
+    def __init__(
+        self,
+        web_paths: List[str],
+        verify_ssl: bool = True,
+        trust_env: bool = False,
+        requests_per_second: Optional[float] = None,
+        continue_on_failure: bool = True,
+        headless: bool = True,
+        remove_selectors: Optional[List[str]] = None,
+        proxy: Optional[Dict[str, str]] = None,
+        playwright_ws_url: Optional[str] = None,
+        playwright_timeout: Optional[int] = 10000,
+    ):
+        """Initialize with additional safety parameters and remote browser support."""
+
+        proxy_server = proxy.get("server") if proxy else None
+        if trust_env and not proxy_server:
+            env_proxies = urllib.request.getproxies()
+            env_proxy_server = env_proxies.get("https") or env_proxies.get("http")
+            if env_proxy_server:
+                if proxy:
+                    proxy["server"] = env_proxy_server
+                else:
+                    proxy = {"server": env_proxy_server}
+
+        # We'll set headless to False if using playwright_ws_url since it's handled by the remote browser
+        super().__init__(
+            urls=web_paths,
+            continue_on_failure=continue_on_failure,
+            headless=headless if playwright_ws_url is None else False,
+            remove_selectors=remove_selectors,
+            proxy=proxy,
+        )
+        self.verify_ssl = verify_ssl
+        self.requests_per_second = requests_per_second
+        self.last_request_time = None
+        self.playwright_ws_url = playwright_ws_url
+        self.trust_env = trust_env
+        self.playwright_timeout = playwright_timeout
+
+    def lazy_load(self) -> Iterator[Document]:
+        """Safely load URLs synchronously with support for remote browser."""
+        from playwright.sync_api import sync_playwright
+
+        with sync_playwright() as p:
+            # Use remote browser if ws_endpoint is provided, otherwise use local browser
+            if self.playwright_ws_url:
+                browser = p.chromium.connect(self.playwright_ws_url)
+            else:
+                browser = p.chromium.launch(headless=self.headless, proxy=self.proxy)
+
+            for url in self.urls:
+                try:
+                    self._safe_process_url_sync(url)
+                    page = browser.new_page()
+                    response = page.goto(url, timeout=self.playwright_timeout)
+                    if response is None:
+                        raise ValueError(f"page.goto() returned None for url {url}")
+
+                    text = self.evaluator.evaluate(page, browser, response)
+                    metadata = {"source": url}
+                    yield Document(page_content=text, metadata=metadata)
+                except Exception as e:
+                    if self.continue_on_failure:
+                        log.exception(f"Error loading {url}: {e}")
+                        continue
+                    raise e
+            browser.close()
+
+    async def alazy_load(self) -> AsyncIterator[Document]:
+        """Safely load URLs asynchronously with support for remote browser."""
+        from playwright.async_api import async_playwright
+
+        async with async_playwright() as p:
+            # Use remote browser if ws_endpoint is provided, otherwise use local browser
+            if self.playwright_ws_url:
+                browser = await p.chromium.connect(self.playwright_ws_url)
+            else:
+                browser = await p.chromium.launch(
+                    headless=self.headless, proxy=self.proxy
+                )
+
+            for url in self.urls:
+                try:
+                    await self._safe_process_url(url)
+                    page = await browser.new_page()
+                    response = await page.goto(url, timeout=self.playwright_timeout)
+                    if response is None:
+                        raise ValueError(f"page.goto() returned None for url {url}")
+
+                    text = await self.evaluator.evaluate_async(page, browser, response)
+                    metadata = {"source": url}
+                    yield Document(page_content=text, metadata=metadata)
+                except Exception as e:
+                    if self.continue_on_failure:
+                        log.exception(f"Error loading {url}: {e}")
+                        continue
+                    raise e
+            await browser.close()
+
+
 class SafeWebBaseLoader(WebBaseLoader):
     """WebBaseLoader with enhanced error handling for URLs."""
 
+    def __init__(self, trust_env: bool = False, *args, **kwargs):
+        """Initialize SafeWebBaseLoader
+        Args:
+            trust_env (bool, optional): set to True if using proxy to make web requests, for example
+                using http(s)_proxy environment variables. Defaults to False.
+        """
+        super().__init__(*args, **kwargs)
+        self.trust_env = trust_env
+
+    async def _fetch(
+        self, url: str, retries: int = 3, cooldown: int = 2, backoff: float = 1.5
+    ) -> str:
+        async with aiohttp.ClientSession(trust_env=self.trust_env) as session:
+            for i in range(retries):
+                try:
+                    kwargs: Dict = dict(
+                        headers=self.session.headers,
+                        cookies=self.session.cookies.get_dict(),
+                    )
+                    if not self.session.verify:
+                        kwargs["ssl"] = False
+
+                    async with session.get(
+                        url, **(self.requests_kwargs | kwargs)
+                    ) as response:
+                        if self.raise_for_status:
+                            response.raise_for_status()
+                        return await response.text()
+                except aiohttp.ClientConnectionError as e:
+                    if i == retries - 1:
+                        raise
+                    else:
+                        log.warning(
+                            f"Error fetching {url} with attempt "
+                            f"{i + 1}/{retries}: {e}. Retrying..."
+                        )
+                        await asyncio.sleep(cooldown * backoff**i)
+        raise ValueError("retry count exceeded")
+
+    def _unpack_fetch_results(
+        self, results: Any, urls: List[str], parser: Union[str, None] = None
+    ) -> List[Any]:
+        """Unpack fetch results into BeautifulSoup objects."""
+        from bs4 import BeautifulSoup
+
+        final_results = []
+        for i, result in enumerate(results):
+            url = urls[i]
+            if parser is None:
+                if url.endswith(".xml"):
+                    parser = "xml"
+                else:
+                    parser = self.default_parser
+                self._check_parser(parser)
+            final_results.append(BeautifulSoup(result, parser, **self.bs_kwargs))
+        return final_results
+
+    async def ascrape_all(
+        self, urls: List[str], parser: Union[str, None] = None
+    ) -> List[Any]:
+        """Async fetch all urls, then return soups for all results."""
+        results = await self.fetch_all(urls)
+        return self._unpack_fetch_results(results, urls, parser=parser)
+
     def lazy_load(self) -> Iterator[Document]:
         """Lazy load text from the url(s) in web_path with error handling."""
         for path in self.web_paths:
@@ -65,33 +556,81 @@ class SafeWebBaseLoader(WebBaseLoader):
                 text = soup.get_text(**self.bs_get_text_kwargs)
 
                 # Build metadata
-                metadata = {"source": path}
-                if title := soup.find("title"):
-                    metadata["title"] = title.get_text()
-                if description := soup.find("meta", attrs={"name": "description"}):
-                    metadata["description"] = description.get(
-                        "content", "No description found."
-                    )
-                if html := soup.find("html"):
-                    metadata["language"] = html.get("lang", "No language found.")
+                metadata = extract_metadata(soup, path)
 
                 yield Document(page_content=text, metadata=metadata)
             except Exception as e:
                 # Log the error and continue with the next URL
-                log.error(f"Error loading {path}: {e}")
+                log.exception(f"Error loading {path}: {e}")
+
+    async def alazy_load(self) -> AsyncIterator[Document]:
+        """Async lazy load text from the url(s) in web_path."""
+        results = await self.ascrape_all(self.web_paths)
+        for path, soup in zip(self.web_paths, results):
+            text = soup.get_text(**self.bs_get_text_kwargs)
+            metadata = {"source": path}
+            if title := soup.find("title"):
+                metadata["title"] = title.get_text()
+            if description := soup.find("meta", attrs={"name": "description"}):
+                metadata["description"] = description.get(
+                    "content", "No description found."
+                )
+            if html := soup.find("html"):
+                metadata["language"] = html.get("lang", "No language found.")
+            yield Document(page_content=text, metadata=metadata)
+
+    async def aload(self) -> list[Document]:
+        """Load data into Document objects."""
+        return [document async for document in self.alazy_load()]
 
 
 def get_web_loader(
     urls: Union[str, Sequence[str]],
     verify_ssl: bool = True,
     requests_per_second: int = 2,
+    trust_env: bool = False,
 ):
-    # Check if the URL is valid
-    if not validate_url(urls):
-        raise ValueError(ERROR_MESSAGES.INVALID_URL)
-    return SafeWebBaseLoader(
-        urls,
-        verify_ssl=verify_ssl,
-        requests_per_second=requests_per_second,
-        continue_on_failure=True,
-    )
+    # Check if the URLs are valid
+    safe_urls = safe_validate_urls([urls] if isinstance(urls, str) else urls)
+
+    web_loader_args = {
+        "web_paths": safe_urls,
+        "verify_ssl": verify_ssl,
+        "requests_per_second": requests_per_second,
+        "continue_on_failure": True,
+        "trust_env": trust_env,
+    }
+
+    if WEB_LOADER_ENGINE.value == "" or WEB_LOADER_ENGINE.value == "safe_web":
+        WebLoaderClass = SafeWebBaseLoader
+    if WEB_LOADER_ENGINE.value == "playwright":
+        WebLoaderClass = SafePlaywrightURLLoader
+        web_loader_args["playwright_timeout"] = PLAYWRIGHT_TIMEOUT.value * 1000
+        if PLAYWRIGHT_WS_URL.value:
+            web_loader_args["playwright_ws_url"] = PLAYWRIGHT_WS_URL.value
+
+    if WEB_LOADER_ENGINE.value == "firecrawl":
+        WebLoaderClass = SafeFireCrawlLoader
+        web_loader_args["api_key"] = FIRECRAWL_API_KEY.value
+        web_loader_args["api_url"] = FIRECRAWL_API_BASE_URL.value
+
+    if WEB_LOADER_ENGINE.value == "tavily":
+        WebLoaderClass = SafeTavilyLoader
+        web_loader_args["api_key"] = TAVILY_API_KEY.value
+        web_loader_args["extract_depth"] = TAVILY_EXTRACT_DEPTH.value
+
+    if WebLoaderClass:
+        web_loader = WebLoaderClass(**web_loader_args)
+
+        log.debug(
+            "Using WEB_LOADER_ENGINE %s for %s URLs",
+            web_loader.__class__.__name__,
+            len(safe_urls),
+        )
+
+        return web_loader
+    else:
+        raise ValueError(
+            f"Invalid WEB_LOADER_ENGINE: {WEB_LOADER_ENGINE.value}. "
+            "Please set it to 'safe_web', 'playwright', 'firecrawl', or 'tavily'."
+        )

backend/open_webui/routers/audio.py

@@ -11,6 +11,7 @@ from pydub.silence import split_on_silence
 import aiohttp
 import aiofiles
 import requests
+import mimetypes
 
 from fastapi import (
     Depends,
@@ -36,6 +37,7 @@ from open_webui.config import (
 
 from open_webui.constants import ERROR_MESSAGES
 from open_webui.env import (
+    AIOHTTP_CLIENT_TIMEOUT,
     ENV,
     SRC_LOG_LEVELS,
     DEVICE_TYPE,
@@ -48,11 +50,13 @@ router = APIRouter()
 # Constants
 MAX_FILE_SIZE_MB = 25
 MAX_FILE_SIZE = MAX_FILE_SIZE_MB * 1024 * 1024  # Convert MB to bytes
+AZURE_MAX_FILE_SIZE_MB = 200
+AZURE_MAX_FILE_SIZE = AZURE_MAX_FILE_SIZE_MB * 1024 * 1024  # Convert MB to bytes
 
 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["AUDIO"])
 
-SPEECH_CACHE_DIR = Path(CACHE_DIR).joinpath("./audio/speech/")
+SPEECH_CACHE_DIR = CACHE_DIR / "audio" / "speech"
 SPEECH_CACHE_DIR.mkdir(parents=True, exist_ok=True)
 
 
@@ -66,10 +70,10 @@ from pydub import AudioSegment
 from pydub.utils import mediainfo
 
 
-def is_mp4_audio(file_path):
-    """Check if the given file is an MP4 audio file."""
+def get_audio_format(file_path):
+    """Check if the given file needs to be converted to a different format."""
     if not os.path.isfile(file_path):
-        print(f"File not found: {file_path}")
+        log.error(f"File not found: {file_path}")
         return False
 
     info = mediainfo(file_path)
@@ -78,15 +82,19 @@ def is_mp4_audio(file_path):
         and info.get("codec_type") == "audio"
         and info.get("codec_tag_string") == "mp4a"
     ):
-        return True
-    return False
+        return "mp4"
+    elif info.get("format_name") == "ogg":
+        return "ogg"
+    elif info.get("format_name") == "matroska,webm":
+        return "webm"
+    return None
 
 
-def convert_mp4_to_wav(file_path, output_path):
-    """Convert MP4 audio file to WAV format."""
-    audio = AudioSegment.from_file(file_path, format="mp4")
+def convert_audio_to_wav(file_path, output_path, conversion_type):
+    """Convert MP4/OGG audio file to WAV format."""
+    audio = AudioSegment.from_file(file_path, format=conversion_type)
     audio.export(output_path, format="wav")
-    print(f"Converted {file_path} to {output_path}")
+    log.info(f"Converted {file_path} to {output_path}")
 
 
 def set_faster_whisper_model(model: str, auto_update: bool = False):
@@ -138,6 +146,10 @@ class STTConfigForm(BaseModel):
     ENGINE: str
     MODEL: str
     WHISPER_MODEL: str
+    DEEPGRAM_API_KEY: str
+    AZURE_API_KEY: str
+    AZURE_REGION: str
+    AZURE_LOCALES: str
 
 
 class AudioConfigUpdateForm(BaseModel):
@@ -165,6 +177,10 @@ async def get_audio_config(request: Request, user=Depends(get_admin_user)):
             "ENGINE": request.app.state.config.STT_ENGINE,
             "MODEL": request.app.state.config.STT_MODEL,
             "WHISPER_MODEL": request.app.state.config.WHISPER_MODEL,
+            "DEEPGRAM_API_KEY": request.app.state.config.DEEPGRAM_API_KEY,
+            "AZURE_API_KEY": request.app.state.config.AUDIO_STT_AZURE_API_KEY,
+            "AZURE_REGION": request.app.state.config.AUDIO_STT_AZURE_REGION,
+            "AZURE_LOCALES": request.app.state.config.AUDIO_STT_AZURE_LOCALES,
         },
     }
 
@@ -190,6 +206,10 @@ async def update_audio_config(
     request.app.state.config.STT_ENGINE = form_data.stt.ENGINE
     request.app.state.config.STT_MODEL = form_data.stt.MODEL
     request.app.state.config.WHISPER_MODEL = form_data.stt.WHISPER_MODEL
+    request.app.state.config.DEEPGRAM_API_KEY = form_data.stt.DEEPGRAM_API_KEY
+    request.app.state.config.AUDIO_STT_AZURE_API_KEY = form_data.stt.AZURE_API_KEY
+    request.app.state.config.AUDIO_STT_AZURE_REGION = form_data.stt.AZURE_REGION
+    request.app.state.config.AUDIO_STT_AZURE_LOCALES = form_data.stt.AZURE_LOCALES
 
     if request.app.state.config.STT_ENGINE == "":
         request.app.state.faster_whisper_model = set_faster_whisper_model(
@@ -214,6 +234,10 @@ async def update_audio_config(
             "ENGINE": request.app.state.config.STT_ENGINE,
             "MODEL": request.app.state.config.STT_MODEL,
             "WHISPER_MODEL": request.app.state.config.WHISPER_MODEL,
+            "DEEPGRAM_API_KEY": request.app.state.config.DEEPGRAM_API_KEY,
+            "AZURE_API_KEY": request.app.state.config.AUDIO_STT_AZURE_API_KEY,
+            "AZURE_REGION": request.app.state.config.AUDIO_STT_AZURE_REGION,
+            "AZURE_LOCALES": request.app.state.config.AUDIO_STT_AZURE_LOCALES,
         },
     }
 
@@ -260,8 +284,10 @@ async def speech(request: Request, user=Depends(get_verified_user)):
         payload["model"] = request.app.state.config.TTS_MODEL
 
         try:
-            # print(payload)
-            async with aiohttp.ClientSession() as session:
+            timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT)
+            async with aiohttp.ClientSession(
+                timeout=timeout, trust_env=True
+            ) as session:
                 async with session.post(
                     url=f"{request.app.state.config.TTS_OPENAI_API_BASE_URL}/audio/speech",
                     json=payload,
@@ -304,7 +330,7 @@ async def speech(request: Request, user=Depends(get_verified_user)):
                 detail = f"External: {e}"
 
             raise HTTPException(
-                status_code=getattr(r, "status", 500),
+                status_code=getattr(r, "status", 500) if r else 500,
                 detail=detail if detail else "Open WebUI: Server Connection Error",
             )
 
@@ -318,7 +344,10 @@ async def speech(request: Request, user=Depends(get_verified_user)):
             )
 
         try:
-            async with aiohttp.ClientSession() as session:
+            timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT)
+            async with aiohttp.ClientSession(
+                timeout=timeout, trust_env=True
+            ) as session:
                 async with session.post(
                     f"https://api.elevenlabs.io/v1/text-to-speech/{voice_id}",
                     json={
@@ -355,7 +384,7 @@ async def speech(request: Request, user=Depends(get_verified_user)):
                 detail = f"External: {e}"
 
             raise HTTPException(
-                status_code=getattr(r, "status", 500),
+                status_code=getattr(r, "status", 500) if r else 500,
                 detail=detail if detail else "Open WebUI: Server Connection Error",
             )
 
@@ -375,7 +404,10 @@ async def speech(request: Request, user=Depends(get_verified_user)):
             data = f"""<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xml:lang="{locale}">
                 <voice name="{language}">{payload["input"]}</voice>
             </speak>"""
-            async with aiohttp.ClientSession() as session:
+            timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT)
+            async with aiohttp.ClientSession(
+                timeout=timeout, trust_env=True
+            ) as session:
                 async with session.post(
                     f"https://{region}.tts.speech.microsoft.com/cognitiveservices/v1",
                     headers={
@@ -408,7 +440,7 @@ async def speech(request: Request, user=Depends(get_verified_user)):
                 detail = f"External: {e}"
 
             raise HTTPException(
-                status_code=getattr(r, "status", 500),
+                status_code=getattr(r, "status", 500) if r else 500,
                 detail=detail if detail else "Open WebUI: Server Connection Error",
             )
 
@@ -453,7 +485,7 @@ async def speech(request: Request, user=Depends(get_verified_user)):
 
 
 def transcribe(request: Request, file_path):
-    print("transcribe", file_path)
+    log.info(f"transcribe: {file_path}")
     filename = os.path.basename(file_path)
     file_dir = os.path.dirname(file_path)
     id = filename.split(".")[0]
@@ -482,10 +514,15 @@ def transcribe(request: Request, file_path):
         log.debug(data)
         return data
     elif request.app.state.config.STT_ENGINE == "openai":
-        if is_mp4_audio(file_path):
-            os.rename(file_path, file_path.replace(".wav", ".mp4"))
-            # Convert MP4 audio file to WAV format
-            convert_mp4_to_wav(file_path.replace(".wav", ".mp4"), file_path)
+        audio_format = get_audio_format(file_path)
+        if audio_format:
+            os.rename(file_path, file_path.replace(".wav", f".{audio_format}"))
+            # Convert unsupported audio file to WAV format
+            convert_audio_to_wav(
+                file_path.replace(".wav", f".{audio_format}"),
+                file_path,
+                audio_format,
+            )
 
         r = None
         try:
@@ -521,6 +558,182 @@ def transcribe(request: Request, file_path):
 
             raise Exception(detail if detail else "Open WebUI: Server Connection Error")
 
+    elif request.app.state.config.STT_ENGINE == "deepgram":
+        try:
+            # Determine the MIME type of the file
+            mime, _ = mimetypes.guess_type(file_path)
+            if not mime:
+                mime = "audio/wav"  # fallback to wav if undetectable
+
+            # Read the audio file
+            with open(file_path, "rb") as f:
+                file_data = f.read()
+
+            # Build headers and parameters
+            headers = {
+                "Authorization": f"Token {request.app.state.config.DEEPGRAM_API_KEY}",
+                "Content-Type": mime,
+            }
+
+            # Add model if specified
+            params = {}
+            if request.app.state.config.STT_MODEL:
+                params["model"] = request.app.state.config.STT_MODEL
+
+            # Make request to Deepgram API
+            r = requests.post(
+                "https://api.deepgram.com/v1/listen",
+                headers=headers,
+                params=params,
+                data=file_data,
+            )
+            r.raise_for_status()
+            response_data = r.json()
+
+            # Extract transcript from Deepgram response
+            try:
+                transcript = response_data["results"]["channels"][0]["alternatives"][
+                    0
+                ].get("transcript", "")
+            except (KeyError, IndexError) as e:
+                log.error(f"Malformed response from Deepgram: {str(e)}")
+                raise Exception(
+                    "Failed to parse Deepgram response - unexpected response format"
+                )
+            data = {"text": transcript.strip()}
+
+            # Save transcript
+            transcript_file = f"{file_dir}/{id}.json"
+            with open(transcript_file, "w") as f:
+                json.dump(data, f)
+
+            return data
+
+        except Exception as e:
+            log.exception(e)
+            detail = None
+            if r is not None:
+                try:
+                    res = r.json()
+                    if "error" in res:
+                        detail = f"External: {res['error'].get('message', '')}"
+                except Exception:
+                    detail = f"External: {e}"
+            raise Exception(detail if detail else "Open WebUI: Server Connection Error")
+
+    elif request.app.state.config.STT_ENGINE == "azure":
+        # Check file exists and size
+        if not os.path.exists(file_path):
+            raise HTTPException(status_code=400, detail="Audio file not found")
+
+        # Check file size (Azure has a larger limit of 200MB)
+        file_size = os.path.getsize(file_path)
+        if file_size > AZURE_MAX_FILE_SIZE:
+            raise HTTPException(
+                status_code=400,
+                detail=f"File size exceeds Azure's limit of {AZURE_MAX_FILE_SIZE_MB}MB",
+            )
+
+        api_key = request.app.state.config.AUDIO_STT_AZURE_API_KEY
+        region = request.app.state.config.AUDIO_STT_AZURE_REGION
+        locales = request.app.state.config.AUDIO_STT_AZURE_LOCALES
+
+        # IF NO LOCALES, USE DEFAULTS
+        if len(locales) < 2:
+            locales = [
+                "en-US",
+                "es-ES",
+                "es-MX",
+                "fr-FR",
+                "hi-IN",
+                "it-IT",
+                "de-DE",
+                "en-GB",
+                "en-IN",
+                "ja-JP",
+                "ko-KR",
+                "pt-BR",
+                "zh-CN",
+            ]
+            locales = ",".join(locales)
+
+        if not api_key or not region:
+            raise HTTPException(
+                status_code=400,
+                detail="Azure API key and region are required for Azure STT",
+            )
+
+        r = None
+        try:
+            # Prepare the request
+            data = {
+                "definition": json.dumps(
+                    {
+                        "locales": locales.split(","),
+                        "diarization": {"maxSpeakers": 3, "enabled": True},
+                    }
+                    if locales
+                    else {}
+                )
+            }
+            url = f"https://{region}.api.cognitive.microsoft.com/speechtotext/transcriptions:transcribe?api-version=2024-11-15"
+
+            # Use context manager to ensure file is properly closed
+            with open(file_path, "rb") as audio_file:
+                r = requests.post(
+                    url=url,
+                    files={"audio": audio_file},
+                    data=data,
+                    headers={
+                        "Ocp-Apim-Subscription-Key": api_key,
+                    },
+                )
+
+            r.raise_for_status()
+            response = r.json()
+
+            # Extract transcript from response
+            if not response.get("combinedPhrases"):
+                raise ValueError("No transcription found in response")
+
+            # Get the full transcript from combinedPhrases
+            transcript = response["combinedPhrases"][0].get("text", "").strip()
+            if not transcript:
+                raise ValueError("Empty transcript in response")
+
+            data = {"text": transcript}
+
+            # Save transcript to json file (consistent with other providers)
+            transcript_file = f"{file_dir}/{id}.json"
+            with open(transcript_file, "w") as f:
+                json.dump(data, f)
+
+            log.debug(data)
+            return data
+
+        except (KeyError, IndexError, ValueError) as e:
+            log.exception("Error parsing Azure response")
+            raise HTTPException(
+                status_code=500,
+                detail=f"Failed to parse Azure response: {str(e)}",
+            )
+        except requests.exceptions.RequestException as e:
+            log.exception(e)
+            detail = None
+
+            try:
+                if r is not None and r.status_code != 200:
+                    res = r.json()
+                    if "error" in res:
+                        detail = f"External: {res['error'].get('message', '')}"
+            except Exception:
+                detail = f"External: {e}"
+
+            raise HTTPException(
+                status_code=getattr(r, "status_code", 500) if r else 500,
+                detail=detail if detail else "Open WebUI: Server Connection Error",
+            )
+
 
 def compress_audio(file_path):
     if os.path.getsize(file_path) > MAX_FILE_SIZE:
@@ -548,7 +761,9 @@ def transcription(
 ):
     log.info(f"file.content_type: {file.content_type}")
 
-    if file.content_type not in ["audio/mpeg", "audio/wav", "audio/ogg", "audio/x-m4a"]:
+    supported_filetypes = ("audio/mpeg", "audio/wav", "audio/ogg", "audio/x-m4a")
+
+    if not file.content_type.startswith(supported_filetypes):
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=ERROR_MESSAGES.FILE_NOT_SUPPORTED,
@@ -602,7 +817,22 @@ def transcription(
 def get_available_models(request: Request) -> list[dict]:
     available_models = []
     if request.app.state.config.TTS_ENGINE == "openai":
-        available_models = [{"id": "tts-1"}, {"id": "tts-1-hd"}]
+        # Use custom endpoint if not using the official OpenAI API URL
+        if not request.app.state.config.TTS_OPENAI_API_BASE_URL.startswith(
+            "https://api.openai.com"
+        ):
+            try:
+                response = requests.get(
+                    f"{request.app.state.config.TTS_OPENAI_API_BASE_URL}/audio/models"
+                )
+                response.raise_for_status()
+                data = response.json()
+                available_models = data.get("models", [])
+            except Exception as e:
+                log.error(f"Error fetching models from custom endpoint: {str(e)}")
+                available_models = [{"id": "tts-1"}, {"id": "tts-1-hd"}]
+        else:
+            available_models = [{"id": "tts-1"}, {"id": "tts-1-hd"}]
     elif request.app.state.config.TTS_ENGINE == "elevenlabs":
         try:
             response = requests.get(
@@ -633,14 +863,37 @@ def get_available_voices(request) -> dict:
     """Returns {voice_id: voice_name} dict"""
     available_voices = {}
     if request.app.state.config.TTS_ENGINE == "openai":
-        available_voices = {
-            "alloy": "alloy",
-            "echo": "echo",
-            "fable": "fable",
-            "onyx": "onyx",
-            "nova": "nova",
-            "shimmer": "shimmer",
-        }
+        # Use custom endpoint if not using the official OpenAI API URL
+        if not request.app.state.config.TTS_OPENAI_API_BASE_URL.startswith(
+            "https://api.openai.com"
+        ):
+            try:
+                response = requests.get(
+                    f"{request.app.state.config.TTS_OPENAI_API_BASE_URL}/audio/voices"
+                )
+                response.raise_for_status()
+                data = response.json()
+                voices_list = data.get("voices", [])
+                available_voices = {voice["id"]: voice["name"] for voice in voices_list}
+            except Exception as e:
+                log.error(f"Error fetching voices from custom endpoint: {str(e)}")
+                available_voices = {
+                    "alloy": "alloy",
+                    "echo": "echo",
+                    "fable": "fable",
+                    "onyx": "onyx",
+                    "nova": "nova",
+                    "shimmer": "shimmer",
+                }
+        else:
+            available_voices = {
+                "alloy": "alloy",
+                "echo": "echo",
+                "fable": "fable",
+                "onyx": "onyx",
+                "nova": "nova",
+                "shimmer": "shimmer",
+            }
     elif request.app.state.config.TTS_ENGINE == "elevenlabs":
         try:
             available_voices = get_elevenlabs_voices(

backend/open_webui/routers/auths.py

@@ -25,16 +25,13 @@ from open_webui.env import (
     WEBUI_AUTH,
     WEBUI_AUTH_TRUSTED_EMAIL_HEADER,
     WEBUI_AUTH_TRUSTED_NAME_HEADER,
-    WEBUI_SESSION_COOKIE_SAME_SITE,
-    WEBUI_SESSION_COOKIE_SECURE,
+    WEBUI_AUTH_COOKIE_SAME_SITE,
+    WEBUI_AUTH_COOKIE_SECURE,
     SRC_LOG_LEVELS,
 )
 from fastapi import APIRouter, Depends, HTTPException, Request, status
 from fastapi.responses import RedirectResponse, Response
-from open_webui.config import (
-    OPENID_PROVIDER_URL,
-    ENABLE_OAUTH_SIGNUP,
-)
+from open_webui.config import OPENID_PROVIDER_URL, ENABLE_OAUTH_SIGNUP, ENABLE_LDAP
 from pydantic import BaseModel
 from open_webui.utils.misc import parse_duration, validate_email_format
 from open_webui.utils.auth import (
@@ -51,8 +48,10 @@ from open_webui.utils.access_control import get_permissions
 from typing import Optional, List
 
 from ssl import CERT_REQUIRED, PROTOCOL_TLS
-from ldap3 import Server, Connection, NONE, Tls
-from ldap3.utils.conv import escape_filter_chars
+
+if ENABLE_LDAP.value:
+    from ldap3 import Server, Connection, NONE, Tls
+    from ldap3.utils.conv import escape_filter_chars
 
 router = APIRouter()
 
@@ -95,8 +94,8 @@ async def get_session_user(
         value=token,
         expires=datetime_expires_at,
         httponly=True,  # Ensures the cookie is not accessible via JavaScript
-        samesite=WEBUI_SESSION_COOKIE_SAME_SITE,
-        secure=WEBUI_SESSION_COOKIE_SECURE,
+        samesite=WEBUI_AUTH_COOKIE_SAME_SITE,
+        secure=WEBUI_AUTH_COOKIE_SECURE,
     )
 
     user_permissions = get_permissions(
@@ -164,7 +163,7 @@ async def update_password(
 ############################
 # LDAP Authentication
 ############################
-@router.post("/ldap", response_model=SigninResponse)
+@router.post("/ldap", response_model=SessionUserResponse)
 async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
     ENABLE_LDAP = request.app.state.config.ENABLE_LDAP
     LDAP_SERVER_LABEL = request.app.state.config.LDAP_SERVER_LABEL
@@ -195,8 +194,8 @@ async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
             ciphers=LDAP_CIPHERS,
         )
     except Exception as e:
-        log.error(f"An error occurred on TLS: {str(e)}")
-        raise HTTPException(400, detail=str(e))
+        log.error(f"TLS configuration error: {str(e)}")
+        raise HTTPException(400, detail="Failed to configure TLS for LDAP connection.")
 
     try:
         server = Server(
@@ -211,7 +210,7 @@ async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
             LDAP_APP_DN,
             LDAP_APP_PASSWORD,
             auto_bind="NONE",
-            authentication="SIMPLE",
+            authentication="SIMPLE" if LDAP_APP_DN else "ANONYMOUS",
         )
         if not connection_app.bind():
             raise HTTPException(400, detail="Application account bind failed")
@@ -231,9 +230,14 @@ async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
 
         entry = connection_app.entries[0]
         username = str(entry[f"{LDAP_ATTRIBUTE_FOR_USERNAME}"]).lower()
-        mail = str(entry[f"{LDAP_ATTRIBUTE_FOR_MAIL}"])
-        if not mail or mail == "" or mail == "[]":
-            raise HTTPException(400, f"User {form_data.user} does not have mail.")
+        email = entry[f"{LDAP_ATTRIBUTE_FOR_MAIL}"]
+        if not email:
+            raise HTTPException(400, "User does not have a valid email address.")
+        elif isinstance(email, str):
+            email = email.lower()
+        elif isinstance(email, list):
+            email = email[0].lower()
+
         cn = str(entry["cn"])
         user_dn = entry.entry_dn
 
@@ -246,19 +250,24 @@ async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
                 authentication="SIMPLE",
             )
             if not connection_user.bind():
-                raise HTTPException(400, f"Authentication failed for {form_data.user}")
+                raise HTTPException(400, "Authentication failed.")
 
-            user = Users.get_user_by_email(mail)
+            user = Users.get_user_by_email(email)
             if not user:
                 try:
+                    user_count = Users.get_num_users()
+
                     role = (
                         "admin"
-                        if Users.get_num_users() == 0
+                        if user_count == 0
                         else request.app.state.config.DEFAULT_USER_ROLE
                     )
 
                     user = Auths.insert_new_auth(
-                        email=mail, password=str(uuid.uuid4()), name=cn, role=role
+                        email=email,
+                        password=str(uuid.uuid4()),
+                        name=cn,
+                        role=role,
                     )
 
                     if not user:
@@ -269,9 +278,12 @@ async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
                 except HTTPException:
                     raise
                 except Exception as err:
-                    raise HTTPException(500, detail=ERROR_MESSAGES.DEFAULT(err))
+                    log.error(f"LDAP user creation error: {str(err)}")
+                    raise HTTPException(
+                        500, detail="Internal error occurred during LDAP user creation."
+                    )
 
-            user = Auths.authenticate_user_by_trusted_header(mail)
+            user = Auths.authenticate_user_by_trusted_header(email)
 
             if user:
                 token = create_token(
@@ -288,6 +300,10 @@ async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
                     httponly=True,  # Ensures the cookie is not accessible via JavaScript
                 )
 
+                user_permissions = get_permissions(
+                    user.id, request.app.state.config.USER_PERMISSIONS
+                )
+
                 return {
                     "token": token,
                     "token_type": "Bearer",
@@ -296,16 +312,15 @@ async def ldap_auth(request: Request, response: Response, form_data: LdapForm):
                     "name": user.name,
                     "role": user.role,
                     "profile_image_url": user.profile_image_url,
+                    "permissions": user_permissions,
                 }
             else:
                 raise HTTPException(400, detail=ERROR_MESSAGES.INVALID_CRED)
         else:
-            raise HTTPException(
-                400,
-                f"User {form_data.user} does not match the record. Search result: {str(entry[f'{LDAP_ATTRIBUTE_FOR_USERNAME}'])}",
-            )
+            raise HTTPException(400, "User record mismatch.")
     except Exception as e:
-        raise HTTPException(400, detail=str(e))
+        log.error(f"LDAP authentication error: {str(e)}")
+        raise HTTPException(400, detail="LDAP authentication failed.")
 
 
 ############################
@@ -378,8 +393,8 @@ async def signin(request: Request, response: Response, form_data: SigninForm):
             value=token,
             expires=datetime_expires_at,
             httponly=True,  # Ensures the cookie is not accessible via JavaScript
-            samesite=WEBUI_SESSION_COOKIE_SAME_SITE,
-            secure=WEBUI_SESSION_COOKIE_SECURE,
+            samesite=WEBUI_AUTH_COOKIE_SAME_SITE,
+            secure=WEBUI_AUTH_COOKIE_SECURE,
         )
 
         user_permissions = get_permissions(
@@ -408,6 +423,7 @@ async def signin(request: Request, response: Response, form_data: SigninForm):
 
 @router.post("/signup", response_model=SessionUserResponse)
 async def signup(request: Request, response: Response, form_data: SignupForm):
+
     if WEBUI_AUTH:
         if (
             not request.app.state.config.ENABLE_SIGNUP
@@ -422,6 +438,7 @@ async def signup(request: Request, response: Response, form_data: SignupForm):
                 status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.ACCESS_PROHIBITED
             )
 
+    user_count = Users.get_num_users()
     if not validate_email_format(form_data.email.lower()):
         raise HTTPException(
             status.HTTP_400_BAD_REQUEST, detail=ERROR_MESSAGES.INVALID_EMAIL_FORMAT
@@ -432,15 +449,20 @@ async def signup(request: Request, response: Response, form_data: SignupForm):
 
     try:
         role = (
-            "admin"
-            if Users.get_num_users() == 0
-            else request.app.state.config.DEFAULT_USER_ROLE
+            "admin" if user_count == 0 else request.app.state.config.DEFAULT_USER_ROLE
         )
 
-        if Users.get_num_users() == 0:
+        if user_count == 0:
             # Disable signup after the first user is created
             request.app.state.config.ENABLE_SIGNUP = False
 
+        # The password passed to bcrypt must be 72 bytes or fewer. If it is longer, it will be truncated before hashing.
+        if len(form_data.password.encode("utf-8")) > 72:
+            raise HTTPException(
+                status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.PASSWORD_TOO_LONG,
+            )
+
         hashed = get_password_hash(form_data.password)
         user = Auths.insert_new_auth(
             form_data.email.lower(),
@@ -473,12 +495,13 @@ async def signup(request: Request, response: Response, form_data: SignupForm):
                 value=token,
                 expires=datetime_expires_at,
                 httponly=True,  # Ensures the cookie is not accessible via JavaScript
-                samesite=WEBUI_SESSION_COOKIE_SAME_SITE,
-                secure=WEBUI_SESSION_COOKIE_SECURE,
+                samesite=WEBUI_AUTH_COOKIE_SAME_SITE,
+                secure=WEBUI_AUTH_COOKIE_SECURE,
             )
 
             if request.app.state.config.WEBHOOK_URL:
                 post_webhook(
+                    request.app.state.WEBUI_NAME,
                     request.app.state.config.WEBHOOK_URL,
                     WEBHOOK_MESSAGES.USER_SIGNUP(user.name),
                     {
@@ -506,7 +529,8 @@ async def signup(request: Request, response: Response, form_data: SignupForm):
         else:
             raise HTTPException(500, detail=ERROR_MESSAGES.CREATE_USER_ERROR)
     except Exception as err:
-        raise HTTPException(500, detail=ERROR_MESSAGES.DEFAULT(err))
+        log.error(f"Signup error: {str(err)}")
+        raise HTTPException(500, detail="An internal error occurred during signup.")
 
 
 @router.get("/signout")
@@ -525,7 +549,8 @@ async def signout(request: Request, response: Response):
                             if logout_url:
                                 response.delete_cookie("oauth_id_token")
                                 return RedirectResponse(
-                                    url=f"{logout_url}?id_token_hint={oauth_id_token}"
+                                    headers=response.headers,
+                                    url=f"{logout_url}?id_token_hint={oauth_id_token}",
                                 )
                         else:
                             raise HTTPException(
@@ -533,7 +558,11 @@ async def signout(request: Request, response: Response):
                                 detail="Failed to fetch OpenID configuration",
                             )
             except Exception as e:
-                raise HTTPException(status_code=500, detail=str(e))
+                log.error(f"OpenID signout error: {str(e)}")
+                raise HTTPException(
+                    status_code=500,
+                    detail="Failed to sign out from the OpenID provider.",
+                )
 
     return {"status": True}
 
@@ -577,7 +606,10 @@ async def add_user(form_data: AddUserForm, user=Depends(get_admin_user)):
         else:
             raise HTTPException(500, detail=ERROR_MESSAGES.CREATE_USER_ERROR)
     except Exception as err:
-        raise HTTPException(500, detail=ERROR_MESSAGES.DEFAULT(err))
+        log.error(f"Add user error: {str(err)}")
+        raise HTTPException(
+            500, detail="An internal error occurred while adding the user."
+        )
 
 
 ############################
@@ -591,7 +623,7 @@ async def get_admin_details(request: Request, user=Depends(get_current_user)):
         admin_email = request.app.state.config.ADMIN_EMAIL
         admin_name = None
 
-        print(admin_email, admin_name)
+        log.info(f"Admin details - Email: {admin_email}, Name: {admin_name}")
 
         if admin_email:
             admin = Users.get_user_by_email(admin_email)
@@ -625,11 +657,12 @@ async def get_admin_config(request: Request, user=Depends(get_admin_user)):
         "ENABLE_API_KEY": request.app.state.config.ENABLE_API_KEY,
         "ENABLE_API_KEY_ENDPOINT_RESTRICTIONS": request.app.state.config.ENABLE_API_KEY_ENDPOINT_RESTRICTIONS,
         "API_KEY_ALLOWED_ENDPOINTS": request.app.state.config.API_KEY_ALLOWED_ENDPOINTS,
-        "ENABLE_CHANNELS": request.app.state.config.ENABLE_CHANNELS,
         "DEFAULT_USER_ROLE": request.app.state.config.DEFAULT_USER_ROLE,
         "JWT_EXPIRES_IN": request.app.state.config.JWT_EXPIRES_IN,
         "ENABLE_COMMUNITY_SHARING": request.app.state.config.ENABLE_COMMUNITY_SHARING,
         "ENABLE_MESSAGE_RATING": request.app.state.config.ENABLE_MESSAGE_RATING,
+        "ENABLE_CHANNELS": request.app.state.config.ENABLE_CHANNELS,
+        "ENABLE_USER_WEBHOOKS": request.app.state.config.ENABLE_USER_WEBHOOKS,
     }
 
 
@@ -640,11 +673,12 @@ class AdminConfig(BaseModel):
     ENABLE_API_KEY: bool
     ENABLE_API_KEY_ENDPOINT_RESTRICTIONS: bool
     API_KEY_ALLOWED_ENDPOINTS: str
-    ENABLE_CHANNELS: bool
     DEFAULT_USER_ROLE: str
     JWT_EXPIRES_IN: str
     ENABLE_COMMUNITY_SHARING: bool
     ENABLE_MESSAGE_RATING: bool
+    ENABLE_CHANNELS: bool
+    ENABLE_USER_WEBHOOKS: bool
 
 
 @router.post("/admin/config")
@@ -679,6 +713,8 @@ async def update_admin_config(
     )
     request.app.state.config.ENABLE_MESSAGE_RATING = form_data.ENABLE_MESSAGE_RATING
 
+    request.app.state.config.ENABLE_USER_WEBHOOKS = form_data.ENABLE_USER_WEBHOOKS
+
     return {
         "SHOW_ADMIN_DETAILS": request.app.state.config.SHOW_ADMIN_DETAILS,
         "WEBUI_URL": request.app.state.config.WEBUI_URL,
@@ -691,6 +727,7 @@ async def update_admin_config(
         "JWT_EXPIRES_IN": request.app.state.config.JWT_EXPIRES_IN,
         "ENABLE_COMMUNITY_SHARING": request.app.state.config.ENABLE_COMMUNITY_SHARING,
         "ENABLE_MESSAGE_RATING": request.app.state.config.ENABLE_MESSAGE_RATING,
+        "ENABLE_USER_WEBHOOKS": request.app.state.config.ENABLE_USER_WEBHOOKS,
     }
 
 
@@ -745,11 +782,6 @@ async def update_ldap_server(
         if not value:
             raise HTTPException(400, detail=f"Required field {key} is empty")
 
-    if form_data.use_tls and not form_data.certificate_path:
-        raise HTTPException(
-            400, detail="TLS is enabled but certificate file path is missing"
-        )
-
     request.app.state.config.LDAP_SERVER_LABEL = form_data.label
     request.app.state.config.LDAP_SERVER_HOST = form_data.host
     request.app.state.config.LDAP_SERVER_PORT = form_data.port

backend/open_webui/routers/channels.py

@@ -192,7 +192,7 @@ async def get_channel_messages(
 ############################
 
 
-async def send_notification(webui_url, channel, message, active_user_ids):
+async def send_notification(name, webui_url, channel, message, active_user_ids):
     users = get_users_with_access("read", channel.access_control)
 
     for user in users:
@@ -206,6 +206,7 @@ async def send_notification(webui_url, channel, message, active_user_ids):
 
                 if webhook_url:
                     post_webhook(
+                        name,
                         webhook_url,
                         f"#{channel.name} - {webui_url}/channels/{channel.id}\n\n{message.content}",
                         {
@@ -302,6 +303,7 @@ async def post_new_message(
 
             background_tasks.add_task(
                 send_notification,
+                request.app.state.WEBUI_NAME,
                 request.app.state.config.WEBUI_URL,
                 channel,
                 message,

backend/open_webui/routers/chats.py

@@ -2,6 +2,8 @@ import json
 import logging
 from typing import Optional
 
+
+from open_webui.socket.main import get_event_emitter
 from open_webui.models.chats import (
     ChatForm,
     ChatImportForm,
@@ -372,6 +374,107 @@ async def update_chat_by_id(
         )
 
 
+############################
+# UpdateChatMessageById
+############################
+class MessageForm(BaseModel):
+    content: str
+
+
+@router.post("/{id}/messages/{message_id}", response_model=Optional[ChatResponse])
+async def update_chat_message_by_id(
+    id: str, message_id: str, form_data: MessageForm, user=Depends(get_verified_user)
+):
+    chat = Chats.get_chat_by_id(id)
+
+    if not chat:
+        raise HTTPException(
+            status_code=status.HTTP_401_UNAUTHORIZED,
+            detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
+        )
+
+    if chat.user_id != user.id and user.role != "admin":
+        raise HTTPException(
+            status_code=status.HTTP_401_UNAUTHORIZED,
+            detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
+        )
+
+    chat = Chats.upsert_message_to_chat_by_id_and_message_id(
+        id,
+        message_id,
+        {
+            "content": form_data.content,
+        },
+    )
+
+    event_emitter = get_event_emitter(
+        {
+            "user_id": user.id,
+            "chat_id": id,
+            "message_id": message_id,
+        },
+        False,
+    )
+
+    if event_emitter:
+        await event_emitter(
+            {
+                "type": "chat:message",
+                "data": {
+                    "chat_id": id,
+                    "message_id": message_id,
+                    "content": form_data.content,
+                },
+            }
+        )
+
+    return ChatResponse(**chat.model_dump())
+
+
+############################
+# SendChatMessageEventById
+############################
+class EventForm(BaseModel):
+    type: str
+    data: dict
+
+
+@router.post("/{id}/messages/{message_id}/event", response_model=Optional[bool])
+async def send_chat_message_event_by_id(
+    id: str, message_id: str, form_data: EventForm, user=Depends(get_verified_user)
+):
+    chat = Chats.get_chat_by_id(id)
+
+    if not chat:
+        raise HTTPException(
+            status_code=status.HTTP_401_UNAUTHORIZED,
+            detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
+        )
+
+    if chat.user_id != user.id and user.role != "admin":
+        raise HTTPException(
+            status_code=status.HTTP_401_UNAUTHORIZED,
+            detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
+        )
+
+    event_emitter = get_event_emitter(
+        {
+            "user_id": user.id,
+            "chat_id": id,
+            "message_id": message_id,
+        }
+    )
+
+    try:
+        if event_emitter:
+            await event_emitter(form_data.model_dump())
+        else:
+            return False
+        return True
+    except:
+        return False
+
+
 ############################
 # DeleteChatById
 ############################
@@ -444,15 +547,21 @@ async def pin_chat_by_id(id: str, user=Depends(get_verified_user)):
 ############################
 
 
+class CloneForm(BaseModel):
+    title: Optional[str] = None
+
+
 @router.post("/{id}/clone", response_model=Optional[ChatResponse])
-async def clone_chat_by_id(id: str, user=Depends(get_verified_user)):
+async def clone_chat_by_id(
+    form_data: CloneForm, id: str, user=Depends(get_verified_user)
+):
     chat = Chats.get_chat_by_id_and_user_id(id, user.id)
     if chat:
         updated_chat = {
             **chat.chat,
             "originalChatId": chat.id,
             "branchPointMessageId": chat.chat["history"]["currentId"],
-            "title": f"Clone of {chat.title}",
+            "title": form_data.title if form_data.title else f"Clone of {chat.title}",
         }
 
         chat = Chats.insert_new_chat(user.id, ChatForm(**{"chat": updated_chat}))
@@ -470,7 +579,12 @@ async def clone_chat_by_id(id: str, user=Depends(get_verified_user)):
 
 @router.post("/{id}/clone/shared", response_model=Optional[ChatResponse])
 async def clone_shared_chat_by_id(id: str, user=Depends(get_verified_user)):
-    chat = Chats.get_chat_by_share_id(id)
+
+    if user.role == "admin":
+        chat = Chats.get_chat_by_id(id)
+    else:
+        chat = Chats.get_chat_by_share_id(id)
+
     if chat:
         updated_chat = {
             **chat.chat,

backend/open_webui/routers/configs.py

@@ -1,5 +1,5 @@
-from fastapi import APIRouter, Depends, Request
-from pydantic import BaseModel
+from fastapi import APIRouter, Depends, Request, HTTPException
+from pydantic import BaseModel, ConfigDict
 
 from typing import Optional
 
@@ -7,6 +7,8 @@ from open_webui.utils.auth import get_admin_user, get_verified_user
 from open_webui.config import get_config, save_config
 from open_webui.config import BannerModel
 
+from open_webui.utils.tools import get_tool_server_data, get_tool_servers_data
+
 
 router = APIRouter()
 
@@ -36,6 +38,214 @@ async def export_config(user=Depends(get_admin_user)):
     return get_config()
 
 
+############################
+# Direct Connections Config
+############################
+
+
+class DirectConnectionsConfigForm(BaseModel):
+    ENABLE_DIRECT_CONNECTIONS: bool
+
+
+@router.get("/direct_connections", response_model=DirectConnectionsConfigForm)
+async def get_direct_connections_config(request: Request, user=Depends(get_admin_user)):
+    return {
+        "ENABLE_DIRECT_CONNECTIONS": request.app.state.config.ENABLE_DIRECT_CONNECTIONS,
+    }
+
+
+@router.post("/direct_connections", response_model=DirectConnectionsConfigForm)
+async def set_direct_connections_config(
+    request: Request,
+    form_data: DirectConnectionsConfigForm,
+    user=Depends(get_admin_user),
+):
+    request.app.state.config.ENABLE_DIRECT_CONNECTIONS = (
+        form_data.ENABLE_DIRECT_CONNECTIONS
+    )
+    return {
+        "ENABLE_DIRECT_CONNECTIONS": request.app.state.config.ENABLE_DIRECT_CONNECTIONS,
+    }
+
+
+############################
+# ToolServers Config
+############################
+
+
+class ToolServerConnection(BaseModel):
+    url: str
+    path: str
+    auth_type: Optional[str]
+    key: Optional[str]
+    config: Optional[dict]
+
+    model_config = ConfigDict(extra="allow")
+
+
+class ToolServersConfigForm(BaseModel):
+    TOOL_SERVER_CONNECTIONS: list[ToolServerConnection]
+
+
+@router.get("/tool_servers", response_model=ToolServersConfigForm)
+async def get_tool_servers_config(request: Request, user=Depends(get_admin_user)):
+    return {
+        "TOOL_SERVER_CONNECTIONS": request.app.state.config.TOOL_SERVER_CONNECTIONS,
+    }
+
+
+@router.post("/tool_servers", response_model=ToolServersConfigForm)
+async def set_tool_servers_config(
+    request: Request,
+    form_data: ToolServersConfigForm,
+    user=Depends(get_admin_user),
+):
+    request.app.state.config.TOOL_SERVER_CONNECTIONS = [
+        connection.model_dump() for connection in form_data.TOOL_SERVER_CONNECTIONS
+    ]
+
+    request.app.state.TOOL_SERVERS = await get_tool_servers_data(
+        request.app.state.config.TOOL_SERVER_CONNECTIONS
+    )
+
+    return {
+        "TOOL_SERVER_CONNECTIONS": request.app.state.config.TOOL_SERVER_CONNECTIONS,
+    }
+
+
+@router.post("/tool_servers/verify")
+async def verify_tool_servers_config(
+    request: Request, form_data: ToolServerConnection, user=Depends(get_admin_user)
+):
+    """
+    Verify the connection to the tool server.
+    """
+    try:
+
+        token = None
+        if form_data.auth_type == "bearer":
+            token = form_data.key
+        elif form_data.auth_type == "session":
+            token = request.state.token.credentials
+
+        url = f"{form_data.url}/{form_data.path}"
+        return await get_tool_server_data(token, url)
+    except Exception as e:
+        raise HTTPException(
+            status_code=400,
+            detail=f"Failed to connect to the tool server: {str(e)}",
+        )
+
+
+############################
+# CodeInterpreterConfig
+############################
+class CodeInterpreterConfigForm(BaseModel):
+    ENABLE_CODE_EXECUTION: bool
+    CODE_EXECUTION_ENGINE: str
+    CODE_EXECUTION_JUPYTER_URL: Optional[str]
+    CODE_EXECUTION_JUPYTER_AUTH: Optional[str]
+    CODE_EXECUTION_JUPYTER_AUTH_TOKEN: Optional[str]
+    CODE_EXECUTION_JUPYTER_AUTH_PASSWORD: Optional[str]
+    CODE_EXECUTION_JUPYTER_TIMEOUT: Optional[int]
+    ENABLE_CODE_INTERPRETER: bool
+    CODE_INTERPRETER_ENGINE: str
+    CODE_INTERPRETER_PROMPT_TEMPLATE: Optional[str]
+    CODE_INTERPRETER_JUPYTER_URL: Optional[str]
+    CODE_INTERPRETER_JUPYTER_AUTH: Optional[str]
+    CODE_INTERPRETER_JUPYTER_AUTH_TOKEN: Optional[str]
+    CODE_INTERPRETER_JUPYTER_AUTH_PASSWORD: Optional[str]
+    CODE_INTERPRETER_JUPYTER_TIMEOUT: Optional[int]
+
+
+@router.get("/code_execution", response_model=CodeInterpreterConfigForm)
+async def get_code_execution_config(request: Request, user=Depends(get_admin_user)):
+    return {
+        "ENABLE_CODE_EXECUTION": request.app.state.config.ENABLE_CODE_EXECUTION,
+        "CODE_EXECUTION_ENGINE": request.app.state.config.CODE_EXECUTION_ENGINE,
+        "CODE_EXECUTION_JUPYTER_URL": request.app.state.config.CODE_EXECUTION_JUPYTER_URL,
+        "CODE_EXECUTION_JUPYTER_AUTH": request.app.state.config.CODE_EXECUTION_JUPYTER_AUTH,
+        "CODE_EXECUTION_JUPYTER_AUTH_TOKEN": request.app.state.config.CODE_EXECUTION_JUPYTER_AUTH_TOKEN,
+        "CODE_EXECUTION_JUPYTER_AUTH_PASSWORD": request.app.state.config.CODE_EXECUTION_JUPYTER_AUTH_PASSWORD,
+        "CODE_EXECUTION_JUPYTER_TIMEOUT": request.app.state.config.CODE_EXECUTION_JUPYTER_TIMEOUT,
+        "ENABLE_CODE_INTERPRETER": request.app.state.config.ENABLE_CODE_INTERPRETER,
+        "CODE_INTERPRETER_ENGINE": request.app.state.config.CODE_INTERPRETER_ENGINE,
+        "CODE_INTERPRETER_PROMPT_TEMPLATE": request.app.state.config.CODE_INTERPRETER_PROMPT_TEMPLATE,
+        "CODE_INTERPRETER_JUPYTER_URL": request.app.state.config.CODE_INTERPRETER_JUPYTER_URL,
+        "CODE_INTERPRETER_JUPYTER_AUTH": request.app.state.config.CODE_INTERPRETER_JUPYTER_AUTH,
+        "CODE_INTERPRETER_JUPYTER_AUTH_TOKEN": request.app.state.config.CODE_INTERPRETER_JUPYTER_AUTH_TOKEN,
+        "CODE_INTERPRETER_JUPYTER_AUTH_PASSWORD": request.app.state.config.CODE_INTERPRETER_JUPYTER_AUTH_PASSWORD,
+        "CODE_INTERPRETER_JUPYTER_TIMEOUT": request.app.state.config.CODE_INTERPRETER_JUPYTER_TIMEOUT,
+    }
+
+
+@router.post("/code_execution", response_model=CodeInterpreterConfigForm)
+async def set_code_execution_config(
+    request: Request, form_data: CodeInterpreterConfigForm, user=Depends(get_admin_user)
+):
+
+    request.app.state.config.ENABLE_CODE_EXECUTION = form_data.ENABLE_CODE_EXECUTION
+
+    request.app.state.config.CODE_EXECUTION_ENGINE = form_data.CODE_EXECUTION_ENGINE
+    request.app.state.config.CODE_EXECUTION_JUPYTER_URL = (
+        form_data.CODE_EXECUTION_JUPYTER_URL
+    )
+    request.app.state.config.CODE_EXECUTION_JUPYTER_AUTH = (
+        form_data.CODE_EXECUTION_JUPYTER_AUTH
+    )
+    request.app.state.config.CODE_EXECUTION_JUPYTER_AUTH_TOKEN = (
+        form_data.CODE_EXECUTION_JUPYTER_AUTH_TOKEN
+    )
+    request.app.state.config.CODE_EXECUTION_JUPYTER_AUTH_PASSWORD = (
+        form_data.CODE_EXECUTION_JUPYTER_AUTH_PASSWORD
+    )
+    request.app.state.config.CODE_EXECUTION_JUPYTER_TIMEOUT = (
+        form_data.CODE_EXECUTION_JUPYTER_TIMEOUT
+    )
+
+    request.app.state.config.ENABLE_CODE_INTERPRETER = form_data.ENABLE_CODE_INTERPRETER
+    request.app.state.config.CODE_INTERPRETER_ENGINE = form_data.CODE_INTERPRETER_ENGINE
+    request.app.state.config.CODE_INTERPRETER_PROMPT_TEMPLATE = (
+        form_data.CODE_INTERPRETER_PROMPT_TEMPLATE
+    )
+
+    request.app.state.config.CODE_INTERPRETER_JUPYTER_URL = (
+        form_data.CODE_INTERPRETER_JUPYTER_URL
+    )
+
+    request.app.state.config.CODE_INTERPRETER_JUPYTER_AUTH = (
+        form_data.CODE_INTERPRETER_JUPYTER_AUTH
+    )
+
+    request.app.state.config.CODE_INTERPRETER_JUPYTER_AUTH_TOKEN = (
+        form_data.CODE_INTERPRETER_JUPYTER_AUTH_TOKEN
+    )
+    request.app.state.config.CODE_INTERPRETER_JUPYTER_AUTH_PASSWORD = (
+        form_data.CODE_INTERPRETER_JUPYTER_AUTH_PASSWORD
+    )
+    request.app.state.config.CODE_INTERPRETER_JUPYTER_TIMEOUT = (
+        form_data.CODE_INTERPRETER_JUPYTER_TIMEOUT
+    )
+
+    return {
+        "ENABLE_CODE_EXECUTION": request.app.state.config.ENABLE_CODE_EXECUTION,
+        "CODE_EXECUTION_ENGINE": request.app.state.config.CODE_EXECUTION_ENGINE,
+        "CODE_EXECUTION_JUPYTER_URL": request.app.state.config.CODE_EXECUTION_JUPYTER_URL,
+        "CODE_EXECUTION_JUPYTER_AUTH": request.app.state.config.CODE_EXECUTION_JUPYTER_AUTH,
+        "CODE_EXECUTION_JUPYTER_AUTH_TOKEN": request.app.state.config.CODE_EXECUTION_JUPYTER_AUTH_TOKEN,
+        "CODE_EXECUTION_JUPYTER_AUTH_PASSWORD": request.app.state.config.CODE_EXECUTION_JUPYTER_AUTH_PASSWORD,
+        "CODE_EXECUTION_JUPYTER_TIMEOUT": request.app.state.config.CODE_EXECUTION_JUPYTER_TIMEOUT,
+        "ENABLE_CODE_INTERPRETER": request.app.state.config.ENABLE_CODE_INTERPRETER,
+        "CODE_INTERPRETER_ENGINE": request.app.state.config.CODE_INTERPRETER_ENGINE,
+        "CODE_INTERPRETER_PROMPT_TEMPLATE": request.app.state.config.CODE_INTERPRETER_PROMPT_TEMPLATE,
+        "CODE_INTERPRETER_JUPYTER_URL": request.app.state.config.CODE_INTERPRETER_JUPYTER_URL,
+        "CODE_INTERPRETER_JUPYTER_AUTH": request.app.state.config.CODE_INTERPRETER_JUPYTER_AUTH,
+        "CODE_INTERPRETER_JUPYTER_AUTH_TOKEN": request.app.state.config.CODE_INTERPRETER_JUPYTER_AUTH_TOKEN,
+        "CODE_INTERPRETER_JUPYTER_AUTH_PASSWORD": request.app.state.config.CODE_INTERPRETER_JUPYTER_AUTH_PASSWORD,
+        "CODE_INTERPRETER_JUPYTER_TIMEOUT": request.app.state.config.CODE_INTERPRETER_JUPYTER_TIMEOUT,
+    }
+
+
 ############################
 # SetDefaultModels
 ############################

backend/open_webui/routers/evaluations.py

@@ -56,8 +56,19 @@ async def update_config(
     }
 
 
+class FeedbackUserReponse(BaseModel):
+    id: str
+    name: str
+    email: str
+    role: str = "pending"
+
+    last_active_at: int  # timestamp in epoch
+    updated_at: int  # timestamp in epoch
+    created_at: int  # timestamp in epoch
+
+
 class FeedbackUserResponse(FeedbackResponse):
-    user: Optional[UserModel] = None
+    user: Optional[FeedbackUserReponse] = None
 
 
 @router.get("/feedbacks/all", response_model=list[FeedbackUserResponse])
@@ -65,7 +76,10 @@ async def get_all_feedbacks(user=Depends(get_admin_user)):
     feedbacks = Feedbacks.get_all_feedbacks()
     return [
         FeedbackUserResponse(
-            **feedback.model_dump(), user=Users.get_user_by_id(feedback.user_id)
+            **feedback.model_dump(),
+            user=FeedbackUserReponse(
+                **Users.get_user_by_id(feedback.user_id).model_dump()
+            ),
         )
         for feedback in feedbacks
     ]

backend/open_webui/routers/files.py

@@ -1,32 +1,38 @@
 import logging
 import os
 import uuid
+from fnmatch import fnmatch
 from pathlib import Path
 from typing import Optional
-from pydantic import BaseModel
-import mimetypes
 from urllib.parse import quote
 
-from open_webui.storage.provider import Storage
-
+from fastapi import (
+    APIRouter,
+    Depends,
+    File,
+    HTTPException,
+    Request,
+    UploadFile,
+    status,
+    Query,
+)
+from fastapi.responses import FileResponse, StreamingResponse
+from open_webui.constants import ERROR_MESSAGES
+from open_webui.env import SRC_LOG_LEVELS
 from open_webui.models.files import (
     FileForm,
     FileModel,
     FileModelResponse,
     Files,
 )
-from open_webui.routers.retrieval import process_file, ProcessFileForm
-
-from open_webui.config import UPLOAD_DIR
-from open_webui.env import SRC_LOG_LEVELS
-from open_webui.constants import ERROR_MESSAGES
-
-
-from fastapi import APIRouter, Depends, File, HTTPException, UploadFile, status, Request
-from fastapi.responses import FileResponse, StreamingResponse
-
+from open_webui.models.knowledge import Knowledges
 
+from open_webui.routers.knowledge import get_knowledge, get_knowledge_list
+from open_webui.routers.retrieval import ProcessFileForm, process_file
+from open_webui.routers.audio import transcribe
+from open_webui.storage.provider import Storage
 from open_webui.utils.auth import get_admin_user, get_verified_user
+from pydantic import BaseModel
 
 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["MODELS"])
@@ -34,6 +40,39 @@ log.setLevel(SRC_LOG_LEVELS["MODELS"])
 
 router = APIRouter()
 
+
+############################
+# Check if the current user has access to a file through any knowledge bases the user may be in.
+############################
+
+
+def has_access_to_file(
+    file_id: Optional[str], access_type: str, user=Depends(get_verified_user)
+) -> bool:
+    file = Files.get_file_by_id(file_id)
+    log.debug(f"Checking if user has {access_type} access to file")
+
+    if not file:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+    has_access = False
+    knowledge_base_id = file.meta.get("collection_name") if file.meta else None
+
+    if knowledge_base_id:
+        knowledge_bases = Knowledges.get_knowledge_bases_by_user_id(
+            user.id, access_type
+        )
+        for knowledge_base in knowledge_bases:
+            if knowledge_base.id == knowledge_base_id:
+                has_access = True
+                break
+
+    return has_access
+
+
 ############################
 # Upload File
 ############################
@@ -41,7 +80,11 @@ router = APIRouter()
 
 @router.post("/", response_model=FileModelResponse)
 def upload_file(
-    request: Request, file: UploadFile = File(...), user=Depends(get_verified_user)
+    request: Request,
+    file: UploadFile = File(...),
+    user=Depends(get_verified_user),
+    file_metadata: dict = {},
+    process: bool = Query(True),
 ):
     log.info(f"file.content_type: {file.content_type}")
     try:
@@ -65,23 +108,40 @@ def upload_file(
                         "name": name,
                         "content_type": file.content_type,
                         "size": len(contents),
+                        "data": file_metadata,
                     },
                 }
             ),
         )
+        if process:
+            try:
+                if file.content_type in [
+                    "audio/mpeg",
+                    "audio/wav",
+                    "audio/ogg",
+                    "audio/x-m4a",
+                ]:
+                    file_path = Storage.get_file(file_path)
+                    result = transcribe(request, file_path)
 
-        try:
-            process_file(request, ProcessFileForm(file_id=id))
-            file_item = Files.get_file_by_id(id=id)
-        except Exception as e:
-            log.exception(e)
-            log.error(f"Error processing file: {file_item.id}")
-            file_item = FileModelResponse(
-                **{
-                    **file_item.model_dump(),
-                    "error": str(e.detail) if hasattr(e, "detail") else str(e),
-                }
-            )
+                    process_file(
+                        request,
+                        ProcessFileForm(file_id=id, content=result.get("text", "")),
+                        user=user,
+                    )
+                elif file.content_type not in ["image/png", "image/jpeg", "image/gif"]:
+                    process_file(request, ProcessFileForm(file_id=id), user=user)
+
+                file_item = Files.get_file_by_id(id=id)
+            except Exception as e:
+                log.exception(e)
+                log.error(f"Error processing file: {file_item.id}")
+                file_item = FileModelResponse(
+                    **{
+                        **file_item.model_dump(),
+                        "error": str(e.detail) if hasattr(e, "detail") else str(e),
+                    }
+                )
 
         if file_item:
             return file_item
@@ -105,14 +165,60 @@ def upload_file(
 
 
 @router.get("/", response_model=list[FileModelResponse])
-async def list_files(user=Depends(get_verified_user)):
+async def list_files(user=Depends(get_verified_user), content: bool = Query(True)):
     if user.role == "admin":
         files = Files.get_files()
     else:
         files = Files.get_files_by_user_id(user.id)
+
+    if not content:
+        for file in files:
+            del file.data["content"]
+
     return files
 
 
+############################
+# Search Files
+############################
+
+
+@router.get("/search", response_model=list[FileModelResponse])
+async def search_files(
+    filename: str = Query(
+        ...,
+        description="Filename pattern to search for. Supports wildcards such as '*.txt'",
+    ),
+    content: bool = Query(True),
+    user=Depends(get_verified_user),
+):
+    """
+    Search for files by filename with support for wildcard patterns.
+    """
+    # Get files according to user role
+    if user.role == "admin":
+        files = Files.get_files()
+    else:
+        files = Files.get_files_by_user_id(user.id)
+
+    # Get matching files
+    matching_files = [
+        file for file in files if fnmatch(file.filename.lower(), filename.lower())
+    ]
+
+    if not matching_files:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail="No files found matching the pattern.",
+        )
+
+    if not content:
+        for file in matching_files:
+            del file.data["content"]
+
+    return matching_files
+
+
 ############################
 # Delete All Files
 ############################
@@ -126,7 +232,7 @@ async def delete_all_files(user=Depends(get_admin_user)):
             Storage.delete_all_files()
         except Exception as e:
             log.exception(e)
-            log.error(f"Error deleting files")
+            log.error("Error deleting files")
             raise HTTPException(
                 status_code=status.HTTP_400_BAD_REQUEST,
                 detail=ERROR_MESSAGES.DEFAULT("Error deleting files"),
@@ -148,7 +254,17 @@ async def delete_all_files(user=Depends(get_admin_user)):
 async def get_file_by_id(id: str, user=Depends(get_verified_user)):
     file = Files.get_file_by_id(id)
 
-    if file and (file.user_id == user.id or user.role == "admin"):
+    if not file:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+    if (
+        file.user_id == user.id
+        or user.role == "admin"
+        or has_access_to_file(id, "read", user)
+    ):
         return file
     else:
         raise HTTPException(
@@ -166,7 +282,17 @@ async def get_file_by_id(id: str, user=Depends(get_verified_user)):
 async def get_file_data_content_by_id(id: str, user=Depends(get_verified_user)):
     file = Files.get_file_by_id(id)
 
-    if file and (file.user_id == user.id or user.role == "admin"):
+    if not file:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+    if (
+        file.user_id == user.id
+        or user.role == "admin"
+        or has_access_to_file(id, "read", user)
+    ):
         return {"content": file.data.get("content", "")}
     else:
         raise HTTPException(
@@ -190,10 +316,22 @@ async def update_file_data_content_by_id(
 ):
     file = Files.get_file_by_id(id)
 
-    if file and (file.user_id == user.id or user.role == "admin"):
+    if not file:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+    if (
+        file.user_id == user.id
+        or user.role == "admin"
+        or has_access_to_file(id, "write", user)
+    ):
         try:
             process_file(
-                request, ProcessFileForm(file_id=id, content=form_data.content)
+                request,
+                ProcessFileForm(file_id=id, content=form_data.content),
+                user=user,
             )
             file = Files.get_file_by_id(id=id)
         except Exception as e:
@@ -214,9 +352,22 @@ async def update_file_data_content_by_id(
 
 
 @router.get("/{id}/content")
-async def get_file_content_by_id(id: str, user=Depends(get_verified_user)):
+async def get_file_content_by_id(
+    id: str, user=Depends(get_verified_user), attachment: bool = Query(False)
+):
     file = Files.get_file_by_id(id)
-    if file and (file.user_id == user.id or user.role == "admin"):
+
+    if not file:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+    if (
+        file.user_id == user.id
+        or user.role == "admin"
+        or has_access_to_file(id, "read", user)
+    ):
         try:
             file_path = Storage.get_file(file.path)
             file_path = Path(file_path)
@@ -227,17 +378,29 @@ async def get_file_content_by_id(id: str, user=Depends(get_verified_user)):
                 filename = file.meta.get("name", file.filename)
                 encoded_filename = quote(filename)  # RFC5987 encoding
 
+                content_type = file.meta.get("content_type")
+                filename = file.meta.get("name", file.filename)
+                encoded_filename = quote(filename)
                 headers = {}
-                if file.meta.get("content_type") not in [
-                    "application/pdf",
-                    "text/plain",
-                ]:
-                    headers = {
-                        **headers,
-                        "Content-Disposition": f"attachment; filename*=UTF-8''{encoded_filename}",
-                    }
 
-                return FileResponse(file_path, headers=headers)
+                if attachment:
+                    headers["Content-Disposition"] = (
+                        f"attachment; filename*=UTF-8''{encoded_filename}"
+                    )
+                else:
+                    if content_type == "application/pdf" or filename.lower().endswith(
+                        ".pdf"
+                    ):
+                        headers["Content-Disposition"] = (
+                            f"inline; filename*=UTF-8''{encoded_filename}"
+                        )
+                        content_type = "application/pdf"
+                    elif content_type != "text/plain":
+                        headers["Content-Disposition"] = (
+                            f"attachment; filename*=UTF-8''{encoded_filename}"
+                        )
+
+                return FileResponse(file_path, headers=headers, media_type=content_type)
 
             else:
                 raise HTTPException(
@@ -246,7 +409,7 @@ async def get_file_content_by_id(id: str, user=Depends(get_verified_user)):
                 )
         except Exception as e:
             log.exception(e)
-            log.error(f"Error getting file content")
+            log.error("Error getting file content")
             raise HTTPException(
                 status_code=status.HTTP_400_BAD_REQUEST,
                 detail=ERROR_MESSAGES.DEFAULT("Error getting file content"),
@@ -261,14 +424,25 @@ async def get_file_content_by_id(id: str, user=Depends(get_verified_user)):
 @router.get("/{id}/content/html")
 async def get_html_file_content_by_id(id: str, user=Depends(get_verified_user)):
     file = Files.get_file_by_id(id)
-    if file and (file.user_id == user.id or user.role == "admin"):
+
+    if not file:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+    if (
+        file.user_id == user.id
+        or user.role == "admin"
+        or has_access_to_file(id, "read", user)
+    ):
         try:
             file_path = Storage.get_file(file.path)
             file_path = Path(file_path)
 
             # Check if the file already exists in the cache
             if file_path.is_file():
-                print(f"file_path: {file_path}")
+                log.info(f"file_path: {file_path}")
                 return FileResponse(file_path)
             else:
                 raise HTTPException(
@@ -277,7 +451,7 @@ async def get_html_file_content_by_id(id: str, user=Depends(get_verified_user)):
                 )
         except Exception as e:
             log.exception(e)
-            log.error(f"Error getting file content")
+            log.error("Error getting file content")
             raise HTTPException(
                 status_code=status.HTTP_400_BAD_REQUEST,
                 detail=ERROR_MESSAGES.DEFAULT("Error getting file content"),
@@ -293,7 +467,17 @@ async def get_html_file_content_by_id(id: str, user=Depends(get_verified_user)):
 async def get_file_content_by_id(id: str, user=Depends(get_verified_user)):
     file = Files.get_file_by_id(id)
 
-    if file and (file.user_id == user.id or user.role == "admin"):
+    if not file:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+    if (
+        file.user_id == user.id
+        or user.role == "admin"
+        or has_access_to_file(id, "read", user)
+    ):
         file_path = file.path
 
         # Handle Unicode filenames
@@ -344,7 +528,18 @@ async def get_file_content_by_id(id: str, user=Depends(get_verified_user)):
 @router.delete("/{id}")
 async def delete_file_by_id(id: str, user=Depends(get_verified_user)):
     file = Files.get_file_by_id(id)
-    if file and (file.user_id == user.id or user.role == "admin"):
+
+    if not file:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+    if (
+        file.user_id == user.id
+        or user.role == "admin"
+        or has_access_to_file(id, "write", user)
+    ):
         # We should add Chroma cleanup here
 
         result = Files.delete_file_by_id(id)
@@ -353,7 +548,7 @@ async def delete_file_by_id(id: str, user=Depends(get_verified_user)):
                 Storage.delete_file(file.path)
             except Exception as e:
                 log.exception(e)
-                log.error(f"Error deleting files")
+                log.error("Error deleting files")
                 raise HTTPException(
                     status_code=status.HTTP_400_BAD_REQUEST,
                     detail=ERROR_MESSAGES.DEFAULT("Error deleting files"),

backend/open_webui/routers/folders.py

@@ -20,11 +20,13 @@ from open_webui.env import SRC_LOG_LEVELS
 from open_webui.constants import ERROR_MESSAGES
 
 
-from fastapi import APIRouter, Depends, File, HTTPException, UploadFile, status
+from fastapi import APIRouter, Depends, File, HTTPException, UploadFile, status, Request
 from fastapi.responses import FileResponse, StreamingResponse
 
 
 from open_webui.utils.auth import get_admin_user, get_verified_user
+from open_webui.utils.access_control import has_permission
+
 
 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["MODELS"])
@@ -228,7 +230,19 @@ async def update_folder_is_expanded_by_id(
 
 
 @router.delete("/{id}")
-async def delete_folder_by_id(id: str, user=Depends(get_verified_user)):
+async def delete_folder_by_id(
+    request: Request, id: str, user=Depends(get_verified_user)
+):
+    chat_delete_permission = has_permission(
+        user.id, "chat.delete", request.app.state.config.USER_PERMISSIONS
+    )
+
+    if user.role != "admin" and not chat_delete_permission:
+        raise HTTPException(
+            status_code=status.HTTP_403_FORBIDDEN,
+            detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
+        )
+
     folder = Folders.get_folder_by_id_and_user_id(id, user.id)
     if folder:
         try:

backend/open_webui/routers/functions.py

@@ -1,4 +1,5 @@
 import os
+import logging
 from pathlib import Path
 from typing import Optional
 
@@ -13,6 +14,11 @@ from open_webui.config import CACHE_DIR
 from open_webui.constants import ERROR_MESSAGES
 from fastapi import APIRouter, Depends, HTTPException, Request, status
 from open_webui.utils.auth import get_admin_user, get_verified_user
+from open_webui.env import SRC_LOG_LEVELS
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["MAIN"])
+
 
 router = APIRouter()
 
@@ -68,7 +74,7 @@ async def create_new_function(
 
             function = Functions.insert_new_function(user.id, function_type, form_data)
 
-            function_cache_dir = Path(CACHE_DIR) / "functions" / form_data.id
+            function_cache_dir = CACHE_DIR / "functions" / form_data.id
             function_cache_dir.mkdir(parents=True, exist_ok=True)
 
             if function:
@@ -79,7 +85,7 @@ async def create_new_function(
                     detail=ERROR_MESSAGES.DEFAULT("Error creating function"),
                 )
         except Exception as e:
-            print(e)
+            log.exception(f"Failed to create a new function: {e}")
             raise HTTPException(
                 status_code=status.HTTP_400_BAD_REQUEST,
                 detail=ERROR_MESSAGES.DEFAULT(e),
@@ -183,7 +189,7 @@ async def update_function_by_id(
         FUNCTIONS[id] = function_module
 
         updated = {**form_data.model_dump(exclude={"id"}), "type": function_type}
-        print(updated)
+        log.debug(updated)
 
         function = Functions.update_function_by_id(id, updated)
 
@@ -299,7 +305,7 @@ async def update_function_valves_by_id(
                 Functions.update_function_valves_by_id(id, valves.model_dump())
                 return valves.model_dump()
             except Exception as e:
-                print(e)
+                log.exception(f"Error updating function values by id {id}: {e}")
                 raise HTTPException(
                     status_code=status.HTTP_400_BAD_REQUEST,
                     detail=ERROR_MESSAGES.DEFAULT(e),
@@ -388,7 +394,7 @@ async def update_function_user_valves_by_id(
                 )
                 return user_valves.model_dump()
             except Exception as e:
-                print(e)
+                log.exception(f"Error updating function user valves by id {id}: {e}")
                 raise HTTPException(
                     status_code=status.HTTP_400_BAD_REQUEST,
                     detail=ERROR_MESSAGES.DEFAULT(e),

backend/open_webui/routers/groups.py

@@ -1,7 +1,7 @@
 import os
 from pathlib import Path
 from typing import Optional
-
+import logging
 
 from open_webui.models.users import Users
 from open_webui.models.groups import (
@@ -14,7 +14,13 @@ from open_webui.models.groups import (
 from open_webui.config import CACHE_DIR
 from open_webui.constants import ERROR_MESSAGES
 from fastapi import APIRouter, Depends, HTTPException, Request, status
+
 from open_webui.utils.auth import get_admin_user, get_verified_user
+from open_webui.env import SRC_LOG_LEVELS
+
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["MAIN"])
 
 router = APIRouter()
 
@@ -37,7 +43,7 @@ async def get_groups(user=Depends(get_verified_user)):
 
 
 @router.post("/create", response_model=Optional[GroupResponse])
-async def create_new_function(form_data: GroupForm, user=Depends(get_admin_user)):
+async def create_new_group(form_data: GroupForm, user=Depends(get_admin_user)):
     try:
         group = Groups.insert_new_group(user.id, form_data)
         if group:
@@ -48,7 +54,7 @@ async def create_new_function(form_data: GroupForm, user=Depends(get_admin_user)
                 detail=ERROR_MESSAGES.DEFAULT("Error creating group"),
             )
     except Exception as e:
-        print(e)
+        log.exception(f"Error creating a new group: {e}")
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=ERROR_MESSAGES.DEFAULT(e),
@@ -94,7 +100,7 @@ async def update_group_by_id(
                 detail=ERROR_MESSAGES.DEFAULT("Error updating group"),
             )
     except Exception as e:
-        print(e)
+        log.exception(f"Error updating group {id}: {e}")
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=ERROR_MESSAGES.DEFAULT(e),
@@ -118,7 +124,7 @@ async def delete_group_by_id(id: str, user=Depends(get_admin_user)):
                 detail=ERROR_MESSAGES.DEFAULT("Error deleting group"),
             )
     except Exception as e:
-        print(e)
+        log.exception(f"Error deleting group {id}: {e}")
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=ERROR_MESSAGES.DEFAULT(e),

backend/open_webui/routers/images.py

@@ -1,37 +1,31 @@
 import asyncio
 import base64
+import io
 import json
 import logging
 import mimetypes
 import re
-import uuid
 from pathlib import Path
 from typing import Optional
 
 import requests
-
-
-from fastapi import Depends, FastAPI, HTTPException, Request, APIRouter
-from fastapi.middleware.cors import CORSMiddleware
-from pydantic import BaseModel
-
-
+from fastapi import APIRouter, Depends, HTTPException, Request, UploadFile
 from open_webui.config import CACHE_DIR
 from open_webui.constants import ERROR_MESSAGES
-from open_webui.env import ENV, SRC_LOG_LEVELS, ENABLE_FORWARD_USER_INFO_HEADERS
-
+from open_webui.env import ENABLE_FORWARD_USER_INFO_HEADERS, SRC_LOG_LEVELS
+from open_webui.routers.files import upload_file
 from open_webui.utils.auth import get_admin_user, get_verified_user
 from open_webui.utils.images.comfyui import (
     ComfyUIGenerateImageForm,
     ComfyUIWorkflow,
     comfyui_generate_image,
 )
-
+from pydantic import BaseModel
 
 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["IMAGES"])
 
-IMAGE_CACHE_DIR = Path(CACHE_DIR).joinpath("./image/generations/")
+IMAGE_CACHE_DIR = CACHE_DIR / "image" / "generations"
 IMAGE_CACHE_DIR.mkdir(parents=True, exist_ok=True)
 
 
@@ -61,6 +55,10 @@ async def get_config(request: Request, user=Depends(get_admin_user)):
             "COMFYUI_WORKFLOW": request.app.state.config.COMFYUI_WORKFLOW,
             "COMFYUI_WORKFLOW_NODES": request.app.state.config.COMFYUI_WORKFLOW_NODES,
         },
+        "gemini": {
+            "GEMINI_API_BASE_URL": request.app.state.config.IMAGES_GEMINI_API_BASE_URL,
+            "GEMINI_API_KEY": request.app.state.config.IMAGES_GEMINI_API_KEY,
+        },
     }
 
 
@@ -84,6 +82,11 @@ class ComfyUIConfigForm(BaseModel):
     COMFYUI_WORKFLOW_NODES: list[dict]
 
 
+class GeminiConfigForm(BaseModel):
+    GEMINI_API_BASE_URL: str
+    GEMINI_API_KEY: str
+
+
 class ConfigForm(BaseModel):
     enabled: bool
     engine: str
@@ -91,6 +94,7 @@ class ConfigForm(BaseModel):
     openai: OpenAIConfigForm
     automatic1111: Automatic1111ConfigForm
     comfyui: ComfyUIConfigForm
+    gemini: GeminiConfigForm
 
 
 @router.post("/config/update")
@@ -109,6 +113,11 @@ async def update_config(
     )
     request.app.state.config.IMAGES_OPENAI_API_KEY = form_data.openai.OPENAI_API_KEY
 
+    request.app.state.config.IMAGES_GEMINI_API_BASE_URL = (
+        form_data.gemini.GEMINI_API_BASE_URL
+    )
+    request.app.state.config.IMAGES_GEMINI_API_KEY = form_data.gemini.GEMINI_API_KEY
+
     request.app.state.config.AUTOMATIC1111_BASE_URL = (
         form_data.automatic1111.AUTOMATIC1111_BASE_URL
     )
@@ -135,6 +144,8 @@ async def update_config(
     request.app.state.config.COMFYUI_BASE_URL = (
         form_data.comfyui.COMFYUI_BASE_URL.strip("/")
     )
+    request.app.state.config.COMFYUI_API_KEY = form_data.comfyui.COMFYUI_API_KEY
+
     request.app.state.config.COMFYUI_WORKFLOW = form_data.comfyui.COMFYUI_WORKFLOW
     request.app.state.config.COMFYUI_WORKFLOW_NODES = (
         form_data.comfyui.COMFYUI_WORKFLOW_NODES
@@ -161,6 +172,10 @@ async def update_config(
             "COMFYUI_WORKFLOW": request.app.state.config.COMFYUI_WORKFLOW,
             "COMFYUI_WORKFLOW_NODES": request.app.state.config.COMFYUI_WORKFLOW_NODES,
         },
+        "gemini": {
+            "GEMINI_API_BASE_URL": request.app.state.config.IMAGES_GEMINI_API_BASE_URL,
+            "GEMINI_API_KEY": request.app.state.config.IMAGES_GEMINI_API_KEY,
+        },
     }
 
 
@@ -190,9 +205,17 @@ async def verify_url(request: Request, user=Depends(get_admin_user)):
             request.app.state.config.ENABLE_IMAGE_GENERATION = False
             raise HTTPException(status_code=400, detail=ERROR_MESSAGES.INVALID_URL)
     elif request.app.state.config.IMAGE_GENERATION_ENGINE == "comfyui":
+
+        headers = None
+        if request.app.state.config.COMFYUI_API_KEY:
+            headers = {
+                "Authorization": f"Bearer {request.app.state.config.COMFYUI_API_KEY}"
+            }
+
         try:
             r = requests.get(
-                url=f"{request.app.state.config.COMFYUI_BASE_URL}/object_info"
+                url=f"{request.app.state.config.COMFYUI_BASE_URL}/object_info",
+                headers=headers,
             )
             r.raise_for_status()
             return True
@@ -230,6 +253,12 @@ def get_image_model(request):
             if request.app.state.config.IMAGE_GENERATION_MODEL
             else "dall-e-2"
         )
+    elif request.app.state.config.IMAGE_GENERATION_ENGINE == "gemini":
+        return (
+            request.app.state.config.IMAGE_GENERATION_MODEL
+            if request.app.state.config.IMAGE_GENERATION_MODEL
+            else "imagen-3.0-generate-002"
+        )
     elif request.app.state.config.IMAGE_GENERATION_ENGINE == "comfyui":
         return (
             request.app.state.config.IMAGE_GENERATION_MODEL
@@ -271,7 +300,6 @@ async def get_image_config(request: Request, user=Depends(get_admin_user)):
 async def update_image_config(
     request: Request, form_data: ImageConfigForm, user=Depends(get_admin_user)
 ):
-
     set_image_model(request, form_data.MODEL)
 
     pattern = r"^\d+x\d+$"
@@ -306,6 +334,10 @@ def get_models(request: Request, user=Depends(get_verified_user)):
                 {"id": "dall-e-2", "name": "DALL·E 2"},
                 {"id": "dall-e-3", "name": "DALL·E 3"},
             ]
+        elif request.app.state.config.IMAGE_GENERATION_ENGINE == "gemini":
+            return [
+                {"id": "imagen-3-0-generate-002", "name": "imagen-3.0 generate-002"},
+            ]
         elif request.app.state.config.IMAGE_GENERATION_ENGINE == "comfyui":
             # TODO - get models from comfyui
             headers = {
@@ -329,7 +361,7 @@ def get_models(request: Request, user=Depends(get_verified_user)):
             if model_node_id:
                 model_list_key = None
 
-                print(workflow[model_node_id]["class_type"])
+                log.info(workflow[model_node_id]["class_type"])
                 for key in info[workflow[model_node_id]["class_type"]]["input"][
                     "required"
                 ]:
@@ -383,40 +415,22 @@ class GenerateImageForm(BaseModel):
     negative_prompt: Optional[str] = None
 
 
-def save_b64_image(b64_str):
+def load_b64_image_data(b64_str):
     try:
-        image_id = str(uuid.uuid4())
-
         if "," in b64_str:
             header, encoded = b64_str.split(",", 1)
             mime_type = header.split(";")[0]
-
             img_data = base64.b64decode(encoded)
-            image_format = mimetypes.guess_extension(mime_type)
-
-            image_filename = f"{image_id}{image_format}"
-            file_path = IMAGE_CACHE_DIR / f"{image_filename}"
-            with open(file_path, "wb") as f:
-                f.write(img_data)
-            return image_filename
         else:
-            image_filename = f"{image_id}.png"
-            file_path = IMAGE_CACHE_DIR.joinpath(image_filename)
-
+            mime_type = "image/png"
             img_data = base64.b64decode(b64_str)
-
-            # Write the image data to a file
-            with open(file_path, "wb") as f:
-                f.write(img_data)
-            return image_filename
-
+        return img_data, mime_type
     except Exception as e:
-        log.exception(f"Error saving image: {e}")
+        log.exception(f"Error loading image data: {e}")
         return None
 
 
-def save_url_image(url, headers=None):
-    image_id = str(uuid.uuid4())
+def load_url_image_data(url, headers=None):
     try:
         if headers:
             r = requests.get(url, headers=headers)
@@ -426,18 +440,7 @@ def save_url_image(url, headers=None):
         r.raise_for_status()
         if r.headers["content-type"].split("/")[0] == "image":
             mime_type = r.headers["content-type"]
-            image_format = mimetypes.guess_extension(mime_type)
-
-            if not image_format:
-                raise ValueError("Could not determine image type from MIME type")
-
-            image_filename = f"{image_id}{image_format}"
-
-            file_path = IMAGE_CACHE_DIR.joinpath(f"{image_filename}")
-            with open(file_path, "wb") as image_file:
-                for chunk in r.iter_content(chunk_size=8192):
-                    image_file.write(chunk)
-            return image_filename
+            return r.content, mime_type
         else:
             log.error("Url does not point to an image.")
             return None
@@ -447,6 +450,20 @@ def save_url_image(url, headers=None):
         return None
 
 
+def upload_image(request, image_metadata, image_data, content_type, user):
+    image_format = mimetypes.guess_extension(content_type)
+    file = UploadFile(
+        file=io.BytesIO(image_data),
+        filename=f"generated-image{image_format}",  # will be converted to a unique ID on upload_file
+        headers={
+            "content-type": content_type,
+        },
+    )
+    file_item = upload_file(request, file, user, file_metadata=image_metadata)
+    url = request.app.url_path_for("get_file_content_by_id", id=file_item.id)
+    return url
+
+
 @router.post("/generations")
 async def image_generations(
     request: Request,
@@ -500,12 +517,47 @@ async def image_generations(
             images = []
 
             for image in res["data"]:
-                image_filename = save_b64_image(image["b64_json"])
-                images.append({"url": f"/cache/image/generations/{image_filename}"})
-                file_body_path = IMAGE_CACHE_DIR.joinpath(f"{image_filename}.json")
+                if image_url := image.get("url", None):
+                    image_data, content_type = load_url_image_data(image_url, headers)
+                else:
+                    image_data, content_type = load_b64_image_data(image["b64_json"])
 
-                with open(file_body_path, "w") as f:
-                    json.dump(data, f)
+                url = upload_image(request, data, image_data, content_type, user)
+                images.append({"url": url})
+            return images
+
+        elif request.app.state.config.IMAGE_GENERATION_ENGINE == "gemini":
+            headers = {}
+            headers["Content-Type"] = "application/json"
+            headers["x-goog-api-key"] = request.app.state.config.IMAGES_GEMINI_API_KEY
+
+            model = get_image_model(request)
+            data = {
+                "instances": {"prompt": form_data.prompt},
+                "parameters": {
+                    "sampleCount": form_data.n,
+                    "outputOptions": {"mimeType": "image/png"},
+                },
+            }
+
+            # Use asyncio.to_thread for the requests.post call
+            r = await asyncio.to_thread(
+                requests.post,
+                url=f"{request.app.state.config.IMAGES_GEMINI_API_BASE_URL}/models/{model}:predict",
+                json=data,
+                headers=headers,
+            )
+
+            r.raise_for_status()
+            res = r.json()
+
+            images = []
+            for image in res["predictions"]:
+                image_data, content_type = load_b64_image_data(
+                    image["bytesBase64Encoded"]
+                )
+                url = upload_image(request, data, image_data, content_type, user)
+                images.append({"url": url})
 
             return images
 
@@ -552,14 +604,15 @@ async def image_generations(
                         "Authorization": f"Bearer {request.app.state.config.COMFYUI_API_KEY}"
                     }
 
-                image_filename = save_url_image(image["url"], headers)
-                images.append({"url": f"/cache/image/generations/{image_filename}"})
-                file_body_path = IMAGE_CACHE_DIR.joinpath(f"{image_filename}.json")
-
-                with open(file_body_path, "w") as f:
-                    json.dump(form_data.model_dump(exclude_none=True), f)
-
-            log.debug(f"images: {images}")
+                image_data, content_type = load_url_image_data(image["url"], headers)
+                url = upload_image(
+                    request,
+                    form_data.model_dump(exclude_none=True),
+                    image_data,
+                    content_type,
+                    user,
+                )
+                images.append({"url": url})
             return images
         elif (
             request.app.state.config.IMAGE_GENERATION_ENGINE == "automatic1111"
@@ -604,13 +657,15 @@ async def image_generations(
             images = []
 
             for image in res["images"]:
-                image_filename = save_b64_image(image)
-                images.append({"url": f"/cache/image/generations/{image_filename}"})
-                file_body_path = IMAGE_CACHE_DIR.joinpath(f"{image_filename}.json")
-
-                with open(file_body_path, "w") as f:
-                    json.dump({**data, "info": res["info"]}, f)
-
+                image_data, content_type = load_b64_image_data(image)
+                url = upload_image(
+                    request,
+                    {**data, "info": res["info"]},
+                    image_data,
+                    content_type,
+                    user,
+                )
+                images.append({"url": url})
             return images
     except Exception as e:
         error = e

backend/open_webui/routers/knowledge.py

@@ -161,6 +161,74 @@ async def create_new_knowledge(
         )
 
 
+############################
+# ReindexKnowledgeFiles
+############################
+
+
+@router.post("/reindex", response_model=bool)
+async def reindex_knowledge_files(request: Request, user=Depends(get_verified_user)):
+    if user.role != "admin":
+        raise HTTPException(
+            status_code=status.HTTP_401_UNAUTHORIZED,
+            detail=ERROR_MESSAGES.UNAUTHORIZED,
+        )
+
+    knowledge_bases = Knowledges.get_knowledge_bases()
+
+    log.info(f"Starting reindexing for {len(knowledge_bases)} knowledge bases")
+
+    for knowledge_base in knowledge_bases:
+        try:
+            files = Files.get_files_by_ids(knowledge_base.data.get("file_ids", []))
+
+            try:
+                if VECTOR_DB_CLIENT.has_collection(collection_name=knowledge_base.id):
+                    VECTOR_DB_CLIENT.delete_collection(
+                        collection_name=knowledge_base.id
+                    )
+            except Exception as e:
+                log.error(f"Error deleting collection {knowledge_base.id}: {str(e)}")
+                raise HTTPException(
+                    status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+                    detail=f"Error deleting vector DB collection",
+                )
+
+            failed_files = []
+            for file in files:
+                try:
+                    process_file(
+                        request,
+                        ProcessFileForm(
+                            file_id=file.id, collection_name=knowledge_base.id
+                        ),
+                        user=user,
+                    )
+                except Exception as e:
+                    log.error(
+                        f"Error processing file {file.filename} (ID: {file.id}): {str(e)}"
+                    )
+                    failed_files.append({"file_id": file.id, "error": str(e)})
+                    continue
+
+        except Exception as e:
+            log.error(f"Error processing knowledge base {knowledge_base.id}: {str(e)}")
+            raise HTTPException(
+                status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+                detail=f"Error processing knowledge base",
+            )
+
+        if failed_files:
+            log.warning(
+                f"Failed to process {len(failed_files)} files in knowledge base {knowledge_base.id}"
+            )
+            for failed in failed_files:
+                log.warning(f"File ID: {failed['file_id']}, Error: {failed['error']}")
+
+    log.info("Reindexing completed successfully")
+    return True
+
+
 ############################
 # GetKnowledgeById
 ############################
@@ -264,7 +332,11 @@ def add_file_to_knowledge_by_id(
             detail=ERROR_MESSAGES.NOT_FOUND,
         )
 
-    if knowledge.user_id != user.id and user.role != "admin":
+    if (
+        knowledge.user_id != user.id
+        and not has_access(user.id, "write", knowledge.access_control)
+        and user.role != "admin"
+    ):
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
@@ -285,7 +357,9 @@ def add_file_to_knowledge_by_id(
     # Add content to the vector database
     try:
         process_file(
-            request, ProcessFileForm(file_id=form_data.file_id, collection_name=id)
+            request,
+            ProcessFileForm(file_id=form_data.file_id, collection_name=id),
+            user=user,
         )
     except Exception as e:
         log.debug(e)
@@ -342,7 +416,12 @@ def update_file_from_knowledge_by_id(
             detail=ERROR_MESSAGES.NOT_FOUND,
         )
 
-    if knowledge.user_id != user.id and user.role != "admin":
+    if (
+        knowledge.user_id != user.id
+        and not has_access(user.id, "write", knowledge.access_control)
+        and user.role != "admin"
+    ):
+
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
@@ -363,7 +442,9 @@ def update_file_from_knowledge_by_id(
     # Add content to the vector database
     try:
         process_file(
-            request, ProcessFileForm(file_id=form_data.file_id, collection_name=id)
+            request,
+            ProcessFileForm(file_id=form_data.file_id, collection_name=id),
+            user=user,
         )
     except Exception as e:
         raise HTTPException(
@@ -406,7 +487,11 @@ def remove_file_from_knowledge_by_id(
             detail=ERROR_MESSAGES.NOT_FOUND,
         )
 
-    if knowledge.user_id != user.id and user.role != "admin":
+    if (
+        knowledge.user_id != user.id
+        and not has_access(user.id, "write", knowledge.access_control)
+        and user.role != "admin"
+    ):
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
@@ -420,18 +505,24 @@ def remove_file_from_knowledge_by_id(
         )
 
     # Remove content from the vector database
-    VECTOR_DB_CLIENT.delete(
-        collection_name=knowledge.id, filter={"file_id": form_data.file_id}
-    )
+    try:
+        VECTOR_DB_CLIENT.delete(
+            collection_name=knowledge.id, filter={"file_id": form_data.file_id}
+        )
+    except Exception as e:
+        log.debug("This was most likely caused by bypassing embedding processing")
+        log.debug(e)
+        pass
 
-    # Remove the file's collection from vector database
-    file_collection = f"file-{form_data.file_id}"
-    if VECTOR_DB_CLIENT.has_collection(collection_name=file_collection):
-        VECTOR_DB_CLIENT.delete_collection(collection_name=file_collection)
-
-    # Delete physical file
-    if file.path:
-        Storage.delete_file(file.path)
+    try:
+        # Remove the file's collection from vector database
+        file_collection = f"file-{form_data.file_id}"
+        if VECTOR_DB_CLIENT.has_collection(collection_name=file_collection):
+            VECTOR_DB_CLIENT.delete_collection(collection_name=file_collection)
+    except Exception as e:
+        log.debug("This was most likely caused by bypassing embedding processing")
+        log.debug(e)
+        pass
 
     # Delete file from database
     Files.delete_file_by_id(form_data.file_id)
@@ -484,7 +575,11 @@ async def delete_knowledge_by_id(id: str, user=Depends(get_verified_user)):
             detail=ERROR_MESSAGES.NOT_FOUND,
         )
 
-    if knowledge.user_id != user.id and user.role != "admin":
+    if (
+        knowledge.user_id != user.id
+        and not has_access(user.id, "write", knowledge.access_control)
+        and user.role != "admin"
+    ):
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
@@ -543,7 +638,11 @@ async def reset_knowledge_by_id(id: str, user=Depends(get_verified_user)):
             detail=ERROR_MESSAGES.NOT_FOUND,
         )
 
-    if knowledge.user_id != user.id and user.role != "admin":
+    if (
+        knowledge.user_id != user.id
+        and not has_access(user.id, "write", knowledge.access_control)
+        and user.role != "admin"
+    ):
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
@@ -582,14 +681,18 @@ def add_files_to_knowledge_batch(
             detail=ERROR_MESSAGES.NOT_FOUND,
         )
 
-    if knowledge.user_id != user.id and user.role != "admin":
+    if (
+        knowledge.user_id != user.id
+        and not has_access(user.id, "write", knowledge.access_control)
+        and user.role != "admin"
+    ):
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
         )
 
     # Get files content
-    print(f"files/batch/add - {len(form_data)} files")
+    log.info(f"files/batch/add - {len(form_data)} files")
     files: List[FileModel] = []
     for form in form_data:
         file = Files.get_file_by_id(form.file_id)

backend/open_webui/routers/memories.py

@@ -57,7 +57,9 @@ async def add_memory(
             {
                 "id": memory.id,
                 "text": memory.content,
-                "vector": request.app.state.EMBEDDING_FUNCTION(memory.content),
+                "vector": request.app.state.EMBEDDING_FUNCTION(
+                    memory.content, user=user
+                ),
                 "metadata": {"created_at": memory.created_at},
             }
         ],
@@ -82,7 +84,7 @@ async def query_memory(
 ):
     results = VECTOR_DB_CLIENT.search(
         collection_name=f"user-memory-{user.id}",
-        vectors=[request.app.state.EMBEDDING_FUNCTION(form_data.content)],
+        vectors=[request.app.state.EMBEDDING_FUNCTION(form_data.content, user=user)],
         limit=form_data.k,
     )
 
@@ -105,7 +107,9 @@ async def reset_memory_from_vector_db(
             {
                 "id": memory.id,
                 "text": memory.content,
-                "vector": request.app.state.EMBEDDING_FUNCTION(memory.content),
+                "vector": request.app.state.EMBEDDING_FUNCTION(
+                    memory.content, user=user
+                ),
                 "metadata": {
                     "created_at": memory.created_at,
                     "updated_at": memory.updated_at,
@@ -149,7 +153,9 @@ async def update_memory_by_id(
     form_data: MemoryUpdateModel,
     user=Depends(get_verified_user),
 ):
-    memory = Memories.update_memory_by_id(memory_id, form_data.content)
+    memory = Memories.update_memory_by_id_and_user_id(
+        memory_id, user.id, form_data.content
+    )
     if memory is None:
         raise HTTPException(status_code=404, detail="Memory not found")
 
@@ -160,7 +166,9 @@ async def update_memory_by_id(
                 {
                     "id": memory.id,
                     "text": memory.content,
-                    "vector": request.app.state.EMBEDDING_FUNCTION(memory.content),
+                    "vector": request.app.state.EMBEDDING_FUNCTION(
+                        memory.content, user=user
+                    ),
                     "metadata": {
                         "created_at": memory.created_at,
                         "updated_at": memory.updated_at,

backend/open_webui/routers/models.py

@@ -183,7 +183,11 @@ async def delete_model_by_id(id: str, user=Depends(get_verified_user)):
             detail=ERROR_MESSAGES.NOT_FOUND,
         )
 
-    if model.user_id != user.id and user.role != "admin":
+    if (
+        user.role != "admin"
+        and model.user_id != user.id
+        and not has_access(user.id, "write", model.access_control)
+    ):
         raise HTTPException(
             status_code=status.HTTP_401_UNAUTHORIZED,
             detail=ERROR_MESSAGES.UNAUTHORIZED,

backend/open_webui/routers/ollama.py

@@ -11,11 +11,14 @@ import re
 import time
 from typing import Optional, Union
 from urllib.parse import urlparse
-
 import aiohttp
 from aiocache import cached
-
 import requests
+from open_webui.models.users import UserModel
+
+from open_webui.env import (
+    ENABLE_FORWARD_USER_INFO_HEADERS,
+)
 
 from fastapi import (
     Depends,
@@ -28,7 +31,7 @@ from fastapi import (
 )
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import StreamingResponse
-from pydantic import BaseModel, ConfigDict
+from pydantic import BaseModel, ConfigDict, validator
 from starlette.background import BackgroundTask
 
 
@@ -52,7 +55,7 @@ from open_webui.env import (
     ENV,
     SRC_LOG_LEVELS,
     AIOHTTP_CLIENT_TIMEOUT,
-    AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST,
+    AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST,
     BYPASS_MODEL_ACCESS_CONTROL,
 )
 from open_webui.constants import ERROR_MESSAGES
@@ -68,12 +71,26 @@ log.setLevel(SRC_LOG_LEVELS["OLLAMA"])
 ##########################################
 
 
-async def send_get_request(url, key=None):
-    timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST)
+async def send_get_request(url, key=None, user: UserModel = None):
+    timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST)
     try:
         async with aiohttp.ClientSession(timeout=timeout, trust_env=True) as session:
             async with session.get(
-                url, headers={**({"Authorization": f"Bearer {key}"} if key else {})}
+                url,
+                headers={
+                    "Content-Type": "application/json",
+                    **({"Authorization": f"Bearer {key}"} if key else {}),
+                    **(
+                        {
+                            "X-OpenWebUI-User-Name": user.name,
+                            "X-OpenWebUI-User-Id": user.id,
+                            "X-OpenWebUI-User-Email": user.email,
+                            "X-OpenWebUI-User-Role": user.role,
+                        }
+                        if ENABLE_FORWARD_USER_INFO_HEADERS and user
+                        else {}
+                    ),
+                },
             ) as response:
                 return await response.json()
     except Exception as e:
@@ -98,6 +115,7 @@ async def send_post_request(
     stream: bool = True,
     key: Optional[str] = None,
     content_type: Optional[str] = None,
+    user: UserModel = None,
 ):
 
     r = None
@@ -112,6 +130,16 @@ async def send_post_request(
             headers={
                 "Content-Type": "application/json",
                 **({"Authorization": f"Bearer {key}"} if key else {}),
+                **(
+                    {
+                        "X-OpenWebUI-User-Name": user.name,
+                        "X-OpenWebUI-User-Id": user.id,
+                        "X-OpenWebUI-User-Email": user.email,
+                        "X-OpenWebUI-User-Role": user.role,
+                    }
+                    if ENABLE_FORWARD_USER_INFO_HEADERS and user
+                    else {}
+                ),
             },
         )
         r.raise_for_status()
@@ -188,12 +216,24 @@ async def verify_connection(
     key = form_data.key
 
     async with aiohttp.ClientSession(
-        timeout=aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST)
+        timeout=aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST)
     ) as session:
         try:
             async with session.get(
                 f"{url}/api/version",
-                headers={**({"Authorization": f"Bearer {key}"} if key else {})},
+                headers={
+                    **({"Authorization": f"Bearer {key}"} if key else {}),
+                    **(
+                        {
+                            "X-OpenWebUI-User-Name": user.name,
+                            "X-OpenWebUI-User-Id": user.id,
+                            "X-OpenWebUI-User-Email": user.email,
+                            "X-OpenWebUI-User-Role": user.role,
+                        }
+                        if ENABLE_FORWARD_USER_INFO_HEADERS and user
+                        else {}
+                    ),
+                },
             ) as r:
                 if r.status != 200:
                     detail = f"HTTP Error: {r.status}"
@@ -255,8 +295,8 @@ async def update_config(
     }
 
 
-@cached(ttl=3)
-async def get_all_models(request: Request):
+@cached(ttl=1)
+async def get_all_models(request: Request, user: UserModel = None):
     log.info("get_all_models()")
     if request.app.state.config.ENABLE_OLLAMA_API:
         request_tasks = []
@@ -264,7 +304,7 @@ async def get_all_models(request: Request):
             if (str(idx) not in request.app.state.config.OLLAMA_API_CONFIGS) and (
                 url not in request.app.state.config.OLLAMA_API_CONFIGS  # Legacy support
             ):
-                request_tasks.append(send_get_request(f"{url}/api/tags"))
+                request_tasks.append(send_get_request(f"{url}/api/tags", user=user))
             else:
                 api_config = request.app.state.config.OLLAMA_API_CONFIGS.get(
                     str(idx),
@@ -277,7 +317,9 @@ async def get_all_models(request: Request):
                 key = api_config.get("key", None)
 
                 if enable:
-                    request_tasks.append(send_get_request(f"{url}/api/tags", key))
+                    request_tasks.append(
+                        send_get_request(f"{url}/api/tags", key, user=user)
+                    )
                 else:
                     request_tasks.append(asyncio.ensure_future(asyncio.sleep(0, None)))
 
@@ -294,6 +336,7 @@ async def get_all_models(request: Request):
                 )
 
                 prefix_id = api_config.get("prefix_id", None)
+                tags = api_config.get("tags", [])
                 model_ids = api_config.get("model_ids", [])
 
                 if len(model_ids) != 0 and "models" in response:
@@ -308,6 +351,10 @@ async def get_all_models(request: Request):
                     for model in response.get("models", []):
                         model["model"] = f"{prefix_id}.{model['model']}"
 
+                if tags:
+                    for model in response.get("models", []):
+                        model["tags"] = tags
+
         def merge_models_lists(model_lists):
             merged_models = {}
 
@@ -362,7 +409,7 @@ async def get_ollama_tags(
     models = []
 
     if url_idx is None:
-        models = await get_all_models(request)
+        models = await get_all_models(request, user=user)
     else:
         url = request.app.state.config.OLLAMA_BASE_URLS[url_idx]
         key = get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS)
@@ -372,7 +419,19 @@ async def get_ollama_tags(
             r = requests.request(
                 method="GET",
                 url=f"{url}/api/tags",
-                headers={**({"Authorization": f"Bearer {key}"} if key else {})},
+                headers={
+                    **({"Authorization": f"Bearer {key}"} if key else {}),
+                    **(
+                        {
+                            "X-OpenWebUI-User-Name": user.name,
+                            "X-OpenWebUI-User-Id": user.id,
+                            "X-OpenWebUI-User-Email": user.email,
+                            "X-OpenWebUI-User-Role": user.role,
+                        }
+                        if ENABLE_FORWARD_USER_INFO_HEADERS and user
+                        else {}
+                    ),
+                },
             )
             r.raise_for_status()
 
@@ -395,7 +454,7 @@ async def get_ollama_tags(
             )
 
     if user.role == "user" and not BYPASS_MODEL_ACCESS_CONTROL:
-        models["models"] = get_filtered_models(models, user)
+        models["models"] = await get_filtered_models(models, user)
 
     return models
 
@@ -406,18 +465,27 @@ async def get_ollama_versions(request: Request, url_idx: Optional[int] = None):
     if request.app.state.config.ENABLE_OLLAMA_API:
         if url_idx is None:
             # returns lowest version
-            request_tasks = [
-                send_get_request(
-                    f"{url}/api/version",
+            request_tasks = []
+
+            for idx, url in enumerate(request.app.state.config.OLLAMA_BASE_URLS):
+                api_config = request.app.state.config.OLLAMA_API_CONFIGS.get(
+                    str(idx),
                     request.app.state.config.OLLAMA_API_CONFIGS.get(
-                        str(idx),
-                        request.app.state.config.OLLAMA_API_CONFIGS.get(
-                            url, {}
-                        ),  # Legacy support
-                    ).get("key", None),
+                        url, {}
+                    ),  # Legacy support
                 )
-                for idx, url in enumerate(request.app.state.config.OLLAMA_BASE_URLS)
-            ]
+
+                enable = api_config.get("enable", True)
+                key = api_config.get("key", None)
+
+                if enable:
+                    request_tasks.append(
+                        send_get_request(
+                            f"{url}/api/version",
+                            key,
+                        )
+                    )
+
             responses = await asyncio.gather(*request_tasks)
             responses = list(filter(lambda x: x is not None, responses))
 
@@ -479,6 +547,7 @@ async def get_ollama_loaded_models(request: Request, user=Depends(get_verified_u
                         url, {}
                     ),  # Legacy support
                 ).get("key", None),
+                user=user,
             )
             for idx, url in enumerate(request.app.state.config.OLLAMA_BASE_URLS)
         ]
@@ -511,6 +580,7 @@ async def pull_model(
         url=f"{url}/api/pull",
         payload=json.dumps(payload),
         key=get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS),
+        user=user,
     )
 
 
@@ -529,7 +599,7 @@ async def push_model(
     user=Depends(get_admin_user),
 ):
     if url_idx is None:
-        await get_all_models(request)
+        await get_all_models(request, user=user)
         models = request.app.state.OLLAMA_MODELS
 
         if form_data.name in models:
@@ -547,6 +617,7 @@ async def push_model(
         url=f"{url}/api/push",
         payload=form_data.model_dump_json(exclude_none=True).encode(),
         key=get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS),
+        user=user,
     )
 
 
@@ -573,6 +644,7 @@ async def create_model(
         url=f"{url}/api/create",
         payload=form_data.model_dump_json(exclude_none=True).encode(),
         key=get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS),
+        user=user,
     )
 
 
@@ -590,7 +662,7 @@ async def copy_model(
     user=Depends(get_admin_user),
 ):
     if url_idx is None:
-        await get_all_models(request)
+        await get_all_models(request, user=user)
         models = request.app.state.OLLAMA_MODELS
 
         if form_data.source in models:
@@ -611,6 +683,16 @@ async def copy_model(
             headers={
                 "Content-Type": "application/json",
                 **({"Authorization": f"Bearer {key}"} if key else {}),
+                **(
+                    {
+                        "X-OpenWebUI-User-Name": user.name,
+                        "X-OpenWebUI-User-Id": user.id,
+                        "X-OpenWebUI-User-Email": user.email,
+                        "X-OpenWebUI-User-Role": user.role,
+                    }
+                    if ENABLE_FORWARD_USER_INFO_HEADERS and user
+                    else {}
+                ),
             },
             data=form_data.model_dump_json(exclude_none=True).encode(),
         )
@@ -645,7 +727,7 @@ async def delete_model(
     user=Depends(get_admin_user),
 ):
     if url_idx is None:
-        await get_all_models(request)
+        await get_all_models(request, user=user)
         models = request.app.state.OLLAMA_MODELS
 
         if form_data.name in models:
@@ -667,6 +749,16 @@ async def delete_model(
             headers={
                 "Content-Type": "application/json",
                 **({"Authorization": f"Bearer {key}"} if key else {}),
+                **(
+                    {
+                        "X-OpenWebUI-User-Name": user.name,
+                        "X-OpenWebUI-User-Id": user.id,
+                        "X-OpenWebUI-User-Email": user.email,
+                        "X-OpenWebUI-User-Role": user.role,
+                    }
+                    if ENABLE_FORWARD_USER_INFO_HEADERS and user
+                    else {}
+                ),
             },
         )
         r.raise_for_status()
@@ -695,7 +787,7 @@ async def delete_model(
 async def show_model_info(
     request: Request, form_data: ModelNameForm, user=Depends(get_verified_user)
 ):
-    await get_all_models(request)
+    await get_all_models(request, user=user)
     models = request.app.state.OLLAMA_MODELS
 
     if form_data.name not in models:
@@ -716,6 +808,16 @@ async def show_model_info(
             headers={
                 "Content-Type": "application/json",
                 **({"Authorization": f"Bearer {key}"} if key else {}),
+                **(
+                    {
+                        "X-OpenWebUI-User-Name": user.name,
+                        "X-OpenWebUI-User-Id": user.id,
+                        "X-OpenWebUI-User-Email": user.email,
+                        "X-OpenWebUI-User-Role": user.role,
+                    }
+                    if ENABLE_FORWARD_USER_INFO_HEADERS and user
+                    else {}
+                ),
             },
             data=form_data.model_dump_json(exclude_none=True).encode(),
         )
@@ -759,7 +861,7 @@ async def embed(
     log.info(f"generate_ollama_batch_embeddings {form_data}")
 
     if url_idx is None:
-        await get_all_models(request)
+        await get_all_models(request, user=user)
         models = request.app.state.OLLAMA_MODELS
 
         model = form_data.model
@@ -785,6 +887,16 @@ async def embed(
             headers={
                 "Content-Type": "application/json",
                 **({"Authorization": f"Bearer {key}"} if key else {}),
+                **(
+                    {
+                        "X-OpenWebUI-User-Name": user.name,
+                        "X-OpenWebUI-User-Id": user.id,
+                        "X-OpenWebUI-User-Email": user.email,
+                        "X-OpenWebUI-User-Role": user.role,
+                    }
+                    if ENABLE_FORWARD_USER_INFO_HEADERS and user
+                    else {}
+                ),
             },
             data=form_data.model_dump_json(exclude_none=True).encode(),
         )
@@ -828,7 +940,7 @@ async def embeddings(
     log.info(f"generate_ollama_embeddings {form_data}")
 
     if url_idx is None:
-        await get_all_models(request)
+        await get_all_models(request, user=user)
         models = request.app.state.OLLAMA_MODELS
 
         model = form_data.model
@@ -854,6 +966,16 @@ async def embeddings(
             headers={
                 "Content-Type": "application/json",
                 **({"Authorization": f"Bearer {key}"} if key else {}),
+                **(
+                    {
+                        "X-OpenWebUI-User-Name": user.name,
+                        "X-OpenWebUI-User-Id": user.id,
+                        "X-OpenWebUI-User-Email": user.email,
+                        "X-OpenWebUI-User-Role": user.role,
+                    }
+                    if ENABLE_FORWARD_USER_INFO_HEADERS and user
+                    else {}
+                ),
             },
             data=form_data.model_dump_json(exclude_none=True).encode(),
         )
@@ -903,7 +1025,7 @@ async def generate_completion(
     user=Depends(get_verified_user),
 ):
     if url_idx is None:
-        await get_all_models(request)
+        await get_all_models(request, user=user)
         models = request.app.state.OLLAMA_MODELS
 
         model = form_data.model
@@ -933,23 +1055,39 @@ async def generate_completion(
         url=f"{url}/api/generate",
         payload=form_data.model_dump_json(exclude_none=True).encode(),
         key=get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS),
+        user=user,
     )
 
 
 class ChatMessage(BaseModel):
     role: str
-    content: str
+    content: Optional[str] = None
+    tool_calls: Optional[list[dict]] = None
     images: Optional[list[str]] = None
 
+    @validator("content", pre=True)
+    @classmethod
+    def check_at_least_one_field(cls, field_value, values, **kwargs):
+        # Raise an error if both 'content' and 'tool_calls' are None
+        if field_value is None and (
+            "tool_calls" not in values or values["tool_calls"] is None
+        ):
+            raise ValueError(
+                "At least one of 'content' or 'tool_calls' must be provided"
+            )
+
+        return field_value
+
 
 class GenerateChatCompletionForm(BaseModel):
     model: str
     messages: list[ChatMessage]
-    format: Optional[dict] = None
+    format: Optional[Union[dict, str]] = None
     options: Optional[dict] = None
     template: Optional[str] = None
     stream: Optional[bool] = True
     keep_alive: Optional[Union[int, str]] = None
+    tools: Optional[list[dict]] = None
 
 
 async def get_ollama_url(request: Request, model: str, url_idx: Optional[int] = None):
@@ -977,6 +1115,7 @@ async def generate_chat_completion(
     if BYPASS_MODEL_ACCESS_CONTROL:
         bypass_filter = True
 
+    metadata = form_data.pop("metadata", None)
     try:
         form_data = GenerateChatCompletionForm(**form_data)
     except Exception as e:
@@ -1006,7 +1145,7 @@ async def generate_chat_completion(
             payload["options"] = apply_model_params_to_body_ollama(
                 params, payload["options"]
             )
-            payload = apply_model_system_prompt_to_body(params, payload, user)
+            payload = apply_model_system_prompt_to_body(params, payload, metadata, user)
 
         # Check if user has access to the model
         if not bypass_filter and user.role == "user":
@@ -1039,13 +1178,14 @@ async def generate_chat_completion(
     prefix_id = api_config.get("prefix_id", None)
     if prefix_id:
         payload["model"] = payload["model"].replace(f"{prefix_id}.", "")
-
+    # payload["keep_alive"] = -1 # keep alive forever
     return await send_post_request(
         url=f"{url}/api/chat",
         payload=json.dumps(payload),
         stream=form_data.stream,
         key=get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS),
         content_type="application/x-ndjson",
+        user=user,
     )
 
 
@@ -1057,7 +1197,7 @@ class OpenAIChatMessageContent(BaseModel):
 
 class OpenAIChatMessage(BaseModel):
     role: str
-    content: Union[str, list[OpenAIChatMessageContent]]
+    content: Union[Optional[str], list[OpenAIChatMessageContent]]
 
     model_config = ConfigDict(extra="allow")
 
@@ -1148,6 +1288,7 @@ async def generate_openai_completion(
         payload=json.dumps(payload),
         stream=payload.get("stream", False),
         key=get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS),
+        user=user,
     )
 
 
@@ -1159,6 +1300,8 @@ async def generate_openai_chat_completion(
     url_idx: Optional[int] = None,
     user=Depends(get_verified_user),
 ):
+    metadata = form_data.pop("metadata", None)
+
     try:
         completion_form = OpenAIChatCompletionForm(**form_data)
     except Exception as e:
@@ -1185,7 +1328,7 @@ async def generate_openai_chat_completion(
 
         if params:
             payload = apply_model_params_to_body_openai(params, payload)
-            payload = apply_model_system_prompt_to_body(params, payload, user)
+            payload = apply_model_system_prompt_to_body(params, payload, metadata, user)
 
         # Check if user has access to the model
         if user.role == "user":
@@ -1224,6 +1367,7 @@ async def generate_openai_chat_completion(
         payload=json.dumps(payload),
         stream=payload.get("stream", False),
         key=get_api_key(url_idx, url, request.app.state.config.OLLAMA_API_CONFIGS),
+        user=user,
     )
 
 
@@ -1237,7 +1381,7 @@ async def get_openai_models(
 
     models = []
     if url_idx is None:
-        model_list = await get_all_models(request)
+        model_list = await get_all_models(request, user=user)
         models = [
             {
                 "id": model["model"],
@@ -1405,9 +1549,10 @@ async def download_model(
         return None
 
 
+# TODO: Progress bar does not reflect size & duration of upload.
 @router.post("/models/upload")
 @router.post("/models/upload/{url_idx}")
-def upload_model(
+async def upload_model(
     request: Request,
     file: UploadFile = File(...),
     url_idx: Optional[int] = None,
@@ -1416,59 +1561,85 @@ def upload_model(
     if url_idx is None:
         url_idx = 0
     ollama_url = request.app.state.config.OLLAMA_BASE_URLS[url_idx]
+    file_path = os.path.join(UPLOAD_DIR, file.filename)
+    os.makedirs(UPLOAD_DIR, exist_ok=True)
 
-    file_path = f"{UPLOAD_DIR}/{file.filename}"
+    # --- P1: save file locally ---
+    chunk_size = 1024 * 1024 * 2  # 2 MB chunks
+    with open(file_path, "wb") as out_f:
+        while True:
+            chunk = file.file.read(chunk_size)
+            # log.info(f"Chunk: {str(chunk)}") # DEBUG
+            if not chunk:
+                break
+            out_f.write(chunk)
 
-    # Save file in chunks
-    with open(file_path, "wb+") as f:
-        for chunk in file.file:
-            f.write(chunk)
-
-    def file_process_stream():
+    async def file_process_stream():
         nonlocal ollama_url
         total_size = os.path.getsize(file_path)
-        chunk_size = 1024 * 1024
+        log.info(f"Total Model Size: {str(total_size)}")  # DEBUG
+
+        # --- P2: SSE progress + calculate sha256 hash ---
+        file_hash = calculate_sha256(file_path, chunk_size)
+        log.info(f"Model Hash: {str(file_hash)}")  # DEBUG
         try:
             with open(file_path, "rb") as f:
-                total = 0
-                done = False
-
-                while not done:
-                    chunk = f.read(chunk_size)
-                    if not chunk:
-                        done = True
-                        continue
-
-                    total += len(chunk)
-                    progress = round((total / total_size) * 100, 2)
-
-                    res = {
+                bytes_read = 0
+                while chunk := f.read(chunk_size):
+                    bytes_read += len(chunk)
+                    progress = round(bytes_read / total_size * 100, 2)
+                    data_msg = {
                         "progress": progress,
                         "total": total_size,
-                        "completed": total,
+                        "completed": bytes_read,
                     }
-                    yield f"data: {json.dumps(res)}\n\n"
+                    yield f"data: {json.dumps(data_msg)}\n\n"
 
-                if done:
-                    f.seek(0)
-                    hashed = calculate_sha256(f)
-                    f.seek(0)
+            # --- P3: Upload to ollama /api/blobs ---
+            with open(file_path, "rb") as f:
+                url = f"{ollama_url}/api/blobs/sha256:{file_hash}"
+                response = requests.post(url, data=f)
 
-                    url = f"{ollama_url}/api/blobs/sha256:{hashed}"
-                    response = requests.post(url, data=f)
+            if response.ok:
+                log.info(f"Uploaded to /api/blobs")  # DEBUG
+                # Remove local file
+                os.remove(file_path)
 
-                    if response.ok:
-                        res = {
-                            "done": done,
-                            "blob": f"sha256:{hashed}",
-                            "name": file.filename,
-                        }
-                        os.remove(file_path)
-                        yield f"data: {json.dumps(res)}\n\n"
-                    else:
-                        raise Exception(
-                            "Ollama: Could not create blob, Please try again."
-                        )
+                # Create model in ollama
+                model_name, ext = os.path.splitext(file.filename)
+                log.info(f"Created Model: {model_name}")  # DEBUG
+
+                create_payload = {
+                    "model": model_name,
+                    # Reference the file by its original name => the uploaded blob's digest
+                    "files": {file.filename: f"sha256:{file_hash}"},
+                }
+                log.info(f"Model Payload: {create_payload}")  # DEBUG
+
+                # Call ollama /api/create
+                # https://github.com/ollama/ollama/blob/main/docs/api.md#create-a-model
+                create_resp = requests.post(
+                    url=f"{ollama_url}/api/create",
+                    headers={"Content-Type": "application/json"},
+                    data=json.dumps(create_payload),
+                )
+
+                if create_resp.ok:
+                    log.info(f"API SUCCESS!")  # DEBUG
+                    done_msg = {
+                        "done": True,
+                        "blob": f"sha256:{file_hash}",
+                        "name": file.filename,
+                        "model_created": model_name,
+                    }
+                    yield f"data: {json.dumps(done_msg)}\n\n"
+                else:
+                    raise Exception(
+                        f"Failed to create model in Ollama. {create_resp.text}"
+                    )
+
+            else:
+                raise Exception("Ollama: Could not create blob, Please try again.")
 
         except Exception as e:
             res = {"error": str(e)}

backend/open_webui/routers/openai.py

@@ -22,10 +22,11 @@ from open_webui.config import (
 )
 from open_webui.env import (
     AIOHTTP_CLIENT_TIMEOUT,
-    AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST,
+    AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST,
     ENABLE_FORWARD_USER_INFO_HEADERS,
     BYPASS_MODEL_ACCESS_CONTROL,
 )
+from open_webui.models.users import UserModel
 
 from open_webui.constants import ERROR_MESSAGES
 from open_webui.env import ENV, SRC_LOG_LEVELS
@@ -35,6 +36,9 @@ from open_webui.utils.payload import (
     apply_model_params_to_body_openai,
     apply_model_system_prompt_to_body,
 )
+from open_webui.utils.misc import (
+    convert_logit_bias_input_to_json,
+)
 
 from open_webui.utils.auth import get_admin_user, get_verified_user
 from open_webui.utils.access_control import has_access
@@ -51,12 +55,25 @@ log.setLevel(SRC_LOG_LEVELS["OPENAI"])
 ##########################################
 
 
-async def send_get_request(url, key=None):
-    timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST)
+async def send_get_request(url, key=None, user: UserModel = None):
+    timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST)
     try:
         async with aiohttp.ClientSession(timeout=timeout, trust_env=True) as session:
             async with session.get(
-                url, headers={**({"Authorization": f"Bearer {key}"} if key else {})}
+                url,
+                headers={
+                    **({"Authorization": f"Bearer {key}"} if key else {}),
+                    **(
+                        {
+                            "X-OpenWebUI-User-Name": user.name,
+                            "X-OpenWebUI-User-Id": user.id,
+                            "X-OpenWebUI-User-Email": user.email,
+                            "X-OpenWebUI-User-Role": user.role,
+                        }
+                        if ENABLE_FORWARD_USER_INFO_HEADERS and user
+                        else {}
+                    ),
+                },
             ) as response:
                 return await response.json()
     except Exception as e:
@@ -75,18 +92,24 @@ async def cleanup_response(
         await session.close()
 
 
-def openai_o1_handler(payload):
+def openai_o1_o3_handler(payload):
     """
-    Handle O1 specific parameters
+    Handle o1, o3 specific parameters
     """
     if "max_tokens" in payload:
         # Remove "max_tokens" from the payload
         payload["max_completion_tokens"] = payload["max_tokens"]
         del payload["max_tokens"]
 
-    # Fix: O1 does not support the "system" parameter, Modify "system" to "user"
+    # Fix: o1 and o3 do not support the "system" role directly.
+    # For older models like "o1-mini" or "o1-preview", use role "user".
+    # For newer o1/o3 models, replace "system" with "developer".
     if payload["messages"][0]["role"] == "system":
-        payload["messages"][0]["role"] = "user"
+        model_lower = payload["model"].lower()
+        if model_lower.startswith("o1-mini") or model_lower.startswith("o1-preview"):
+            payload["messages"][0]["role"] = "user"
+        else:
+            payload["messages"][0]["role"] = "developer"
 
     return payload
 
@@ -172,7 +195,7 @@ async def speech(request: Request, user=Depends(get_verified_user)):
         body = await request.body()
         name = hashlib.sha256(body).hexdigest()
 
-        SPEECH_CACHE_DIR = Path(CACHE_DIR).joinpath("./audio/speech/")
+        SPEECH_CACHE_DIR = CACHE_DIR / "audio" / "speech"
         SPEECH_CACHE_DIR.mkdir(parents=True, exist_ok=True)
         file_path = SPEECH_CACHE_DIR.joinpath(f"{name}.mp3")
         file_body_path = SPEECH_CACHE_DIR.joinpath(f"{name}.json")
@@ -247,7 +270,7 @@ async def speech(request: Request, user=Depends(get_verified_user)):
         raise HTTPException(status_code=401, detail=ERROR_MESSAGES.OPENAI_NOT_FOUND)
 
 
-async def get_all_models_responses(request: Request) -> list:
+async def get_all_models_responses(request: Request, user: UserModel) -> list:
     if not request.app.state.config.ENABLE_OPENAI_API:
         return []
 
@@ -271,7 +294,9 @@ async def get_all_models_responses(request: Request) -> list:
         ):
             request_tasks.append(
                 send_get_request(
-                    f"{url}/models", request.app.state.config.OPENAI_API_KEYS[idx]
+                    f"{url}/models",
+                    request.app.state.config.OPENAI_API_KEYS[idx],
+                    user=user,
                 )
             )
         else:
@@ -291,6 +316,7 @@ async def get_all_models_responses(request: Request) -> list:
                         send_get_request(
                             f"{url}/models",
                             request.app.state.config.OPENAI_API_KEYS[idx],
+                            user=user,
                         )
                     )
                 else:
@@ -327,6 +353,7 @@ async def get_all_models_responses(request: Request) -> list:
             )
 
             prefix_id = api_config.get("prefix_id", None)
+            tags = api_config.get("tags", [])
 
             if prefix_id:
                 for model in (
@@ -334,6 +361,12 @@ async def get_all_models_responses(request: Request) -> list:
                 ):
                     model["id"] = f"{prefix_id}.{model['id']}"
 
+            if tags:
+                for model in (
+                    response if isinstance(response, list) else response.get("data", [])
+                ):
+                    model["tags"] = tags
+
     log.debug(f"get_all_models:responses() {responses}")
     return responses
 
@@ -351,14 +384,14 @@ async def get_filtered_models(models, user):
     return filtered_models
 
 
-@cached(ttl=3)
-async def get_all_models(request: Request) -> dict[str, list]:
+@cached(ttl=1)
+async def get_all_models(request: Request, user: UserModel) -> dict[str, list]:
     log.info("get_all_models()")
 
     if not request.app.state.config.ENABLE_OPENAI_API:
         return {"data": []}
 
-    responses = await get_all_models_responses(request)
+    responses = await get_all_models_responses(request, user=user)
 
     def extract_data(response):
         if response and "data" in response:
@@ -373,6 +406,7 @@ async def get_all_models(request: Request) -> dict[str, list]:
 
         for idx, models in enumerate(model_lists):
             if models is not None and "error" not in models:
+
                 merged_list.extend(
                     [
                         {
@@ -383,18 +417,21 @@ async def get_all_models(request: Request) -> dict[str, list]:
                             "urlIdx": idx,
                         }
                         for model in models
-                        if "api.openai.com"
-                        not in request.app.state.config.OPENAI_API_BASE_URLS[idx]
-                        or not any(
-                            name in model["id"]
-                            for name in [
-                                "babbage",
-                                "dall-e",
-                                "davinci",
-                                "embedding",
-                                "tts",
-                                "whisper",
-                            ]
+                        if (model.get("id") or model.get("name"))
+                        and (
+                            "api.openai.com"
+                            not in request.app.state.config.OPENAI_API_BASE_URLS[idx]
+                            or not any(
+                                name in model["id"]
+                                for name in [
+                                    "babbage",
+                                    "dall-e",
+                                    "davinci",
+                                    "embedding",
+                                    "tts",
+                                    "whisper",
+                                ]
+                            )
                         )
                     ]
                 )
@@ -418,16 +455,14 @@ async def get_models(
     }
 
     if url_idx is None:
-        models = await get_all_models(request)
+        models = await get_all_models(request, user=user)
     else:
         url = request.app.state.config.OPENAI_API_BASE_URLS[url_idx]
         key = request.app.state.config.OPENAI_API_KEYS[url_idx]
 
         r = None
         async with aiohttp.ClientSession(
-            timeout=aiohttp.ClientTimeout(
-                total=AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST
-            )
+            timeout=aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST)
         ) as session:
             try:
                 async with session.get(
@@ -489,7 +524,7 @@ async def get_models(
                 raise HTTPException(status_code=500, detail=error_detail)
 
     if user.role == "user" and not BYPASS_MODEL_ACCESS_CONTROL:
-        models["data"] = get_filtered_models(models, user)
+        models["data"] = await get_filtered_models(models, user)
 
     return models
 
@@ -507,7 +542,7 @@ async def verify_connection(
     key = form_data.key
 
     async with aiohttp.ClientSession(
-        timeout=aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST)
+        timeout=aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST)
     ) as session:
         try:
             async with session.get(
@@ -515,6 +550,16 @@ async def verify_connection(
                 headers={
                     "Authorization": f"Bearer {key}",
                     "Content-Type": "application/json",
+                    **(
+                        {
+                            "X-OpenWebUI-User-Name": user.name,
+                            "X-OpenWebUI-User-Id": user.id,
+                            "X-OpenWebUI-User-Email": user.email,
+                            "X-OpenWebUI-User-Role": user.role,
+                        }
+                        if ENABLE_FORWARD_USER_INFO_HEADERS
+                        else {}
+                    ),
                 },
             ) as r:
                 if r.status != 200:
@@ -551,9 +596,9 @@ async def generate_chat_completion(
         bypass_filter = True
 
     idx = 0
+
     payload = {**form_data}
-    if "metadata" in payload:
-        del payload["metadata"]
+    metadata = payload.pop("metadata", None)
 
     model_id = form_data.get("model")
     model_info = Models.get_model_by_id(model_id)
@@ -566,7 +611,7 @@ async def generate_chat_completion(
 
         params = model_info.params.model_dump()
         payload = apply_model_params_to_body_openai(params, payload)
-        payload = apply_model_system_prompt_to_body(params, payload, user)
+        payload = apply_model_system_prompt_to_body(params, payload, metadata, user)
 
         # Check if user has access to the model
         if not bypass_filter and user.role == "user":
@@ -587,7 +632,7 @@ async def generate_chat_completion(
                 detail="Model not found",
             )
 
-    await get_all_models(request)
+    await get_all_models(request, user=user)
     model = request.app.state.OPENAI_MODELS.get(model_id)
     if model:
         idx = model["urlIdx"]
@@ -621,10 +666,10 @@ async def generate_chat_completion(
     url = request.app.state.config.OPENAI_API_BASE_URLS[idx]
     key = request.app.state.config.OPENAI_API_KEYS[idx]
 
-    # Fix: O1 does not support the "max_tokens" parameter, Modify "max_tokens" to "max_completion_tokens"
-    is_o1 = payload["model"].lower().startswith("o1-")
-    if is_o1:
-        payload = openai_o1_handler(payload)
+    # Fix: o1,o3 does not support the "max_tokens" parameter, Modify "max_tokens" to "max_completion_tokens"
+    is_o1_o3 = payload["model"].lower().startswith(("o1", "o3-"))
+    if is_o1_o3:
+        payload = openai_o1_o3_handler(payload)
     elif "api.openai.com" not in url:
         # Remove "max_completion_tokens" from the payload for backward compatibility
         if "max_completion_tokens" in payload:
@@ -635,6 +680,11 @@ async def generate_chat_completion(
         del payload["max_tokens"]
 
     # Convert the modified body back to JSON
+    if "logit_bias" in payload:
+        payload["logit_bias"] = json.loads(
+            convert_logit_bias_input_to_json(payload["logit_bias"])
+        )
+
     payload = json.dumps(payload)
 
     r = None
@@ -777,7 +827,7 @@ async def proxy(path: str, request: Request, user=Depends(get_verified_user)):
         if r is not None:
             try:
                 res = await r.json()
-                print(res)
+                log.error(res)
                 if "error" in res:
                     detail = f"External: {res['error']['message'] if 'message' in res['error'] else res['error']}"
             except Exception:

backend/open_webui/routers/pipelines.py

@@ -9,6 +9,7 @@ from fastapi import (
     status,
     APIRouter,
 )
+import aiohttp
 import os
 import logging
 import shutil
@@ -56,96 +57,103 @@ def get_sorted_filters(model_id, models):
     return sorted_filters
 
 
-def process_pipeline_inlet_filter(request, payload, user, models):
+async def process_pipeline_inlet_filter(request, payload, user, models):
     user = {"id": user.id, "email": user.email, "name": user.name, "role": user.role}
     model_id = payload["model"]
-
     sorted_filters = get_sorted_filters(model_id, models)
     model = models[model_id]
 
     if "pipeline" in model:
         sorted_filters.append(model)
 
-    for filter in sorted_filters:
-        r = None
-        try:
-            urlIdx = filter["urlIdx"]
+    async with aiohttp.ClientSession() as session:
+        for filter in sorted_filters:
+            urlIdx = filter.get("urlIdx")
+            if urlIdx is None:
+                continue
 
             url = request.app.state.config.OPENAI_API_BASE_URLS[urlIdx]
             key = request.app.state.config.OPENAI_API_KEYS[urlIdx]
 
-            if key == "":
+            if not key:
                 continue
 
             headers = {"Authorization": f"Bearer {key}"}
-            r = requests.post(
-                f"{url}/{filter['id']}/filter/inlet",
-                headers=headers,
-                json={
-                    "user": user,
-                    "body": payload,
-                },
-            )
+            request_data = {
+                "user": user,
+                "body": payload,
+            }
 
-            r.raise_for_status()
-            payload = r.json()
-        except Exception as e:
-            # Handle connection error here
-            print(f"Connection error: {e}")
-
-            if r is not None:
-                res = r.json()
+            try:
+                async with session.post(
+                    f"{url}/{filter['id']}/filter/inlet",
+                    headers=headers,
+                    json=request_data,
+                ) as response:
+                    payload = await response.json()
+                    response.raise_for_status()
+            except aiohttp.ClientResponseError as e:
+                res = (
+                    await response.json()
+                    if response.content_type == "application/json"
+                    else {}
+                )
                 if "detail" in res:
-                    raise Exception(r.status_code, res["detail"])
+                    raise Exception(response.status, res["detail"])
+            except Exception as e:
+                log.exception(f"Connection error: {e}")
 
     return payload
 
 
-def process_pipeline_outlet_filter(request, payload, user, models):
+async def process_pipeline_outlet_filter(request, payload, user, models):
     user = {"id": user.id, "email": user.email, "name": user.name, "role": user.role}
     model_id = payload["model"]
-
     sorted_filters = get_sorted_filters(model_id, models)
     model = models[model_id]
 
     if "pipeline" in model:
         sorted_filters = [model] + sorted_filters
 
-    for filter in sorted_filters:
-        r = None
-        try:
-            urlIdx = filter["urlIdx"]
+    async with aiohttp.ClientSession() as session:
+        for filter in sorted_filters:
+            urlIdx = filter.get("urlIdx")
+            if urlIdx is None:
+                continue
 
             url = request.app.state.config.OPENAI_API_BASE_URLS[urlIdx]
             key = request.app.state.config.OPENAI_API_KEYS[urlIdx]
 
-            if key != "":
-                r = requests.post(
+            if not key:
+                continue
+
+            headers = {"Authorization": f"Bearer {key}"}
+            request_data = {
+                "user": user,
+                "body": payload,
+            }
+
+            try:
+                async with session.post(
                     f"{url}/{filter['id']}/filter/outlet",
-                    headers={"Authorization": f"Bearer {key}"},
-                    json={
-                        "user": user,
-                        "body": payload,
-                    },
-                )
-
-                r.raise_for_status()
-                data = r.json()
-                payload = data
-        except Exception as e:
-            # Handle connection error here
-            print(f"Connection error: {e}")
-
-            if r is not None:
+                    headers=headers,
+                    json=request_data,
+                ) as response:
+                    payload = await response.json()
+                    response.raise_for_status()
+            except aiohttp.ClientResponseError as e:
                 try:
-                    res = r.json()
+                    res = (
+                        await response.json()
+                        if "application/json" in response.content_type
+                        else {}
+                    )
                     if "detail" in res:
-                        return Exception(r.status_code, res)
+                        raise Exception(response.status, res)
                 except Exception:
                     pass
-
-            else:
-                pass
+            except Exception as e:
+                log.exception(f"Connection error: {e}")
 
     return payload
 
@@ -161,7 +169,7 @@ router = APIRouter()
 
 @router.get("/list")
 async def get_pipelines_list(request: Request, user=Depends(get_admin_user)):
-    responses = await get_all_models_responses(request)
+    responses = await get_all_models_responses(request, user)
     log.debug(f"get_pipelines_list: get_openai_models_responses returned {responses}")
 
     urlIdxs = [
@@ -188,7 +196,7 @@ async def upload_pipeline(
     file: UploadFile = File(...),
     user=Depends(get_admin_user),
 ):
-    print("upload_pipeline", urlIdx, file.filename)
+    log.info(f"upload_pipeline: urlIdx={urlIdx}, filename={file.filename}")
     # Check if the uploaded file is a python file
     if not (file.filename and file.filename.endswith(".py")):
         raise HTTPException(
@@ -223,7 +231,7 @@ async def upload_pipeline(
         return {**data}
     except Exception as e:
         # Handle connection error here
-        print(f"Connection error: {e}")
+        log.exception(f"Connection error: {e}")
 
         detail = None
         status_code = status.HTTP_404_NOT_FOUND
@@ -274,7 +282,7 @@ async def add_pipeline(
         return {**data}
     except Exception as e:
         # Handle connection error here
-        print(f"Connection error: {e}")
+        log.exception(f"Connection error: {e}")
 
         detail = None
         if r is not None:
@@ -319,7 +327,7 @@ async def delete_pipeline(
         return {**data}
     except Exception as e:
         # Handle connection error here
-        print(f"Connection error: {e}")
+        log.exception(f"Connection error: {e}")
 
         detail = None
         if r is not None:
@@ -353,7 +361,7 @@ async def get_pipelines(
         return {**data}
     except Exception as e:
         # Handle connection error here
-        print(f"Connection error: {e}")
+        log.exception(f"Connection error: {e}")
 
         detail = None
         if r is not None:
@@ -392,7 +400,7 @@ async def get_pipeline_valves(
         return {**data}
     except Exception as e:
         # Handle connection error here
-        print(f"Connection error: {e}")
+        log.exception(f"Connection error: {e}")
 
         detail = None
         if r is not None:
@@ -432,7 +440,7 @@ async def get_pipeline_valves_spec(
         return {**data}
     except Exception as e:
         # Handle connection error here
-        print(f"Connection error: {e}")
+        log.exception(f"Connection error: {e}")
 
         detail = None
         if r is not None:
@@ -474,7 +482,7 @@ async def update_pipeline_valves(
         return {**data}
     except Exception as e:
         # Handle connection error here
-        print(f"Connection error: {e}")
+        log.exception(f"Connection error: {e}")
 
         detail = None
 

backend/open_webui/routers/prompts.py

@@ -147,7 +147,11 @@ async def delete_prompt_by_command(command: str, user=Depends(get_verified_user)
             detail=ERROR_MESSAGES.NOT_FOUND,
         )
 
-    if prompt.user_id != user.id and user.role != "admin":
+    if (
+        prompt.user_id != user.id
+        and not has_access(user.id, "write", prompt.access_control)
+        and user.role != "admin"
+    ):
         raise HTTPException(
             status_code=status.HTTP_401_UNAUTHORIZED,
             detail=ERROR_MESSAGES.ACCESS_PROHIBITED,

backend/open_webui/routers/tasks.py

@@ -4,6 +4,7 @@ from fastapi.responses import JSONResponse, RedirectResponse
 from pydantic import BaseModel
 from typing import Optional
 import logging
+import re
 
 from open_webui.utils.chat import generate_chat_completion
 from open_webui.utils.task import (
@@ -19,6 +20,10 @@ from open_webui.utils.auth import get_admin_user, get_verified_user
 from open_webui.constants import TASKS
 
 from open_webui.routers.pipelines import process_pipeline_inlet_filter
+from open_webui.utils.filter import (
+    get_sorted_filter_ids,
+    process_filter_functions,
+)
 from open_webui.utils.task import get_task_model_id
 
 from open_webui.config import (
@@ -57,6 +62,7 @@ async def get_task_config(request: Request, user=Depends(get_verified_user)):
         "AUTOCOMPLETE_GENERATION_INPUT_MAX_LENGTH": request.app.state.config.AUTOCOMPLETE_GENERATION_INPUT_MAX_LENGTH,
         "TAGS_GENERATION_PROMPT_TEMPLATE": request.app.state.config.TAGS_GENERATION_PROMPT_TEMPLATE,
         "ENABLE_TAGS_GENERATION": request.app.state.config.ENABLE_TAGS_GENERATION,
+        "ENABLE_TITLE_GENERATION": request.app.state.config.ENABLE_TITLE_GENERATION,
         "ENABLE_SEARCH_QUERY_GENERATION": request.app.state.config.ENABLE_SEARCH_QUERY_GENERATION,
         "ENABLE_RETRIEVAL_QUERY_GENERATION": request.app.state.config.ENABLE_RETRIEVAL_QUERY_GENERATION,
         "QUERY_GENERATION_PROMPT_TEMPLATE": request.app.state.config.QUERY_GENERATION_PROMPT_TEMPLATE,
@@ -67,6 +73,7 @@ async def get_task_config(request: Request, user=Depends(get_verified_user)):
 class TaskConfigForm(BaseModel):
     TASK_MODEL: Optional[str]
     TASK_MODEL_EXTERNAL: Optional[str]
+    ENABLE_TITLE_GENERATION: bool
     TITLE_GENERATION_PROMPT_TEMPLATE: str
     IMAGE_PROMPT_GENERATION_PROMPT_TEMPLATE: str
     ENABLE_AUTOCOMPLETE_GENERATION: bool
@@ -85,10 +92,15 @@ async def update_task_config(
 ):
     request.app.state.config.TASK_MODEL = form_data.TASK_MODEL
     request.app.state.config.TASK_MODEL_EXTERNAL = form_data.TASK_MODEL_EXTERNAL
+    request.app.state.config.ENABLE_TITLE_GENERATION = form_data.ENABLE_TITLE_GENERATION
     request.app.state.config.TITLE_GENERATION_PROMPT_TEMPLATE = (
         form_data.TITLE_GENERATION_PROMPT_TEMPLATE
     )
 
+    request.app.state.config.IMAGE_PROMPT_GENERATION_PROMPT_TEMPLATE = (
+        form_data.IMAGE_PROMPT_GENERATION_PROMPT_TEMPLATE
+    )
+
     request.app.state.config.ENABLE_AUTOCOMPLETE_GENERATION = (
         form_data.ENABLE_AUTOCOMPLETE_GENERATION
     )
@@ -117,6 +129,7 @@ async def update_task_config(
     return {
         "TASK_MODEL": request.app.state.config.TASK_MODEL,
         "TASK_MODEL_EXTERNAL": request.app.state.config.TASK_MODEL_EXTERNAL,
+        "ENABLE_TITLE_GENERATION": request.app.state.config.ENABLE_TITLE_GENERATION,
         "TITLE_GENERATION_PROMPT_TEMPLATE": request.app.state.config.TITLE_GENERATION_PROMPT_TEMPLATE,
         "IMAGE_PROMPT_GENERATION_PROMPT_TEMPLATE": request.app.state.config.IMAGE_PROMPT_GENERATION_PROMPT_TEMPLATE,
         "ENABLE_AUTOCOMPLETE_GENERATION": request.app.state.config.ENABLE_AUTOCOMPLETE_GENERATION,
@@ -134,7 +147,19 @@ async def update_task_config(
 async def generate_title(
     request: Request, form_data: dict, user=Depends(get_verified_user)
 ):
-    models = request.app.state.MODELS
+
+    if not request.app.state.config.ENABLE_TITLE_GENERATION:
+        return JSONResponse(
+            status_code=status.HTTP_200_OK,
+            content={"detail": "Title generation is disabled"},
+        )
+
+    if getattr(request.state, "direct", False) and hasattr(request.state, "model"):
+        models = {
+            request.state.model["id"]: request.state.model,
+        }
+    else:
+        models = request.app.state.MODELS
 
     model_id = form_data["model"]
     if model_id not in models:
@@ -161,9 +186,20 @@ async def generate_title(
     else:
         template = DEFAULT_TITLE_GENERATION_PROMPT_TEMPLATE
 
+    messages = form_data["messages"]
+
+    # Remove reasoning details from the messages
+    for message in messages:
+        message["content"] = re.sub(
+            r"<details\s+type=\"reasoning\"[^>]*>.*?<\/details>",
+            "",
+            message["content"],
+            flags=re.S,
+        ).strip()
+
     content = title_generation_template(
         template,
-        form_data["messages"],
+        messages,
         {
             "name": user.name,
             "location": user.info.get("location") if user.info else None,
@@ -175,19 +211,26 @@ async def generate_title(
         "messages": [{"role": "user", "content": content}],
         "stream": False,
         **(
-            {"max_tokens": 50}
-            if models[task_model_id]["owned_by"] == "ollama"
+            {"max_tokens": 1000}
+            if models[task_model_id].get("owned_by") == "ollama"
             else {
-                "max_completion_tokens": 50,
+                "max_completion_tokens": 1000,
             }
         ),
         "metadata": {
+            **(request.state.metadata if hasattr(request.state, "metadata") else {}),
             "task": str(TASKS.TITLE_GENERATION),
             "task_body": form_data,
             "chat_id": form_data.get("chat_id", None),
         },
     }
 
+    # Process the payload through the pipeline
+    try:
+        payload = await process_pipeline_inlet_filter(request, payload, user, models)
+    except Exception as e:
+        raise e
+
     try:
         return await generate_chat_completion(request, form_data=payload, user=user)
     except Exception as e:
@@ -209,7 +252,12 @@ async def generate_chat_tags(
             content={"detail": "Tags generation is disabled"},
         )
 
-    models = request.app.state.MODELS
+    if getattr(request.state, "direct", False) and hasattr(request.state, "model"):
+        models = {
+            request.state.model["id"]: request.state.model,
+        }
+    else:
+        models = request.app.state.MODELS
 
     model_id = form_data["model"]
     if model_id not in models:
@@ -245,12 +293,19 @@ async def generate_chat_tags(
         "messages": [{"role": "user", "content": content}],
         "stream": False,
         "metadata": {
+            **(request.state.metadata if hasattr(request.state, "metadata") else {}),
             "task": str(TASKS.TAGS_GENERATION),
             "task_body": form_data,
             "chat_id": form_data.get("chat_id", None),
         },
     }
 
+    # Process the payload through the pipeline
+    try:
+        payload = await process_pipeline_inlet_filter(request, payload, user, models)
+    except Exception as e:
+        raise e
+
     try:
         return await generate_chat_completion(request, form_data=payload, user=user)
     except Exception as e:
@@ -265,7 +320,12 @@ async def generate_chat_tags(
 async def generate_image_prompt(
     request: Request, form_data: dict, user=Depends(get_verified_user)
 ):
-    models = request.app.state.MODELS
+    if getattr(request.state, "direct", False) and hasattr(request.state, "model"):
+        models = {
+            request.state.model["id"]: request.state.model,
+        }
+    else:
+        models = request.app.state.MODELS
 
     model_id = form_data["model"]
     if model_id not in models:
@@ -305,12 +365,19 @@ async def generate_image_prompt(
         "messages": [{"role": "user", "content": content}],
         "stream": False,
         "metadata": {
+            **(request.state.metadata if hasattr(request.state, "metadata") else {}),
             "task": str(TASKS.IMAGE_PROMPT_GENERATION),
             "task_body": form_data,
             "chat_id": form_data.get("chat_id", None),
         },
     }
 
+    # Process the payload through the pipeline
+    try:
+        payload = await process_pipeline_inlet_filter(request, payload, user, models)
+    except Exception as e:
+        raise e
+
     try:
         return await generate_chat_completion(request, form_data=payload, user=user)
     except Exception as e:
@@ -340,7 +407,12 @@ async def generate_queries(
                 detail=f"Query generation is disabled",
             )
 
-    models = request.app.state.MODELS
+    if getattr(request.state, "direct", False) and hasattr(request.state, "model"):
+        models = {
+            request.state.model["id"]: request.state.model,
+        }
+    else:
+        models = request.app.state.MODELS
 
     model_id = form_data["model"]
     if model_id not in models:
@@ -376,12 +448,19 @@ async def generate_queries(
         "messages": [{"role": "user", "content": content}],
         "stream": False,
         "metadata": {
+            **(request.state.metadata if hasattr(request.state, "metadata") else {}),
             "task": str(TASKS.QUERY_GENERATION),
             "task_body": form_data,
             "chat_id": form_data.get("chat_id", None),
         },
     }
 
+    # Process the payload through the pipeline
+    try:
+        payload = await process_pipeline_inlet_filter(request, payload, user, models)
+    except Exception as e:
+        raise e
+
     try:
         return await generate_chat_completion(request, form_data=payload, user=user)
     except Exception as e:
@@ -415,7 +494,12 @@ async def generate_autocompletion(
                 detail=f"Input prompt exceeds maximum length of {request.app.state.config.AUTOCOMPLETE_GENERATION_INPUT_MAX_LENGTH}",
             )
 
-    models = request.app.state.MODELS
+    if getattr(request.state, "direct", False) and hasattr(request.state, "model"):
+        models = {
+            request.state.model["id"]: request.state.model,
+        }
+    else:
+        models = request.app.state.MODELS
 
     model_id = form_data["model"]
     if model_id not in models:
@@ -451,12 +535,19 @@ async def generate_autocompletion(
         "messages": [{"role": "user", "content": content}],
         "stream": False,
         "metadata": {
+            **(request.state.metadata if hasattr(request.state, "metadata") else {}),
             "task": str(TASKS.AUTOCOMPLETE_GENERATION),
             "task_body": form_data,
             "chat_id": form_data.get("chat_id", None),
         },
     }
 
+    # Process the payload through the pipeline
+    try:
+        payload = await process_pipeline_inlet_filter(request, payload, user, models)
+    except Exception as e:
+        raise e
+
     try:
         return await generate_chat_completion(request, form_data=payload, user=user)
     except Exception as e:
@@ -472,7 +563,12 @@ async def generate_emoji(
     request: Request, form_data: dict, user=Depends(get_verified_user)
 ):
 
-    models = request.app.state.MODELS
+    if getattr(request.state, "direct", False) and hasattr(request.state, "model"):
+        models = {
+            request.state.model["id"]: request.state.model,
+        }
+    else:
+        models = request.app.state.MODELS
 
     model_id = form_data["model"]
     if model_id not in models:
@@ -509,15 +605,25 @@ async def generate_emoji(
         "stream": False,
         **(
             {"max_tokens": 4}
-            if models[task_model_id]["owned_by"] == "ollama"
+            if models[task_model_id].get("owned_by") == "ollama"
             else {
                 "max_completion_tokens": 4,
             }
         ),
         "chat_id": form_data.get("chat_id", None),
-        "metadata": {"task": str(TASKS.EMOJI_GENERATION), "task_body": form_data},
+        "metadata": {
+            **(request.state.metadata if hasattr(request.state, "metadata") else {}),
+            "task": str(TASKS.EMOJI_GENERATION),
+            "task_body": form_data,
+        },
     }
 
+    # Process the payload through the pipeline
+    try:
+        payload = await process_pipeline_inlet_filter(request, payload, user, models)
+    except Exception as e:
+        raise e
+
     try:
         return await generate_chat_completion(request, form_data=payload, user=user)
     except Exception as e:
@@ -532,7 +638,13 @@ async def generate_moa_response(
     request: Request, form_data: dict, user=Depends(get_verified_user)
 ):
 
-    models = request.app.state.MODELS
+    if getattr(request.state, "direct", False) and hasattr(request.state, "model"):
+        models = {
+            request.state.model["id"]: request.state.model,
+        }
+    else:
+        models = request.app.state.MODELS
+
     model_id = form_data["model"]
 
     if model_id not in models:
@@ -541,17 +653,6 @@ async def generate_moa_response(
             detail="Model not found",
         )
 
-    # Check if the user has a custom task model
-    # If the user has a custom task model, use that model
-    task_model_id = get_task_model_id(
-        model_id,
-        request.app.state.config.TASK_MODEL,
-        request.app.state.config.TASK_MODEL_EXTERNAL,
-        models,
-    )
-
-    log.debug(f"generating MOA model {task_model_id} for user {user.email} ")
-
     template = DEFAULT_MOA_GENERATION_PROMPT_TEMPLATE
 
     content = moa_response_generation_template(
@@ -561,16 +662,23 @@ async def generate_moa_response(
     )
 
     payload = {
-        "model": task_model_id,
+        "model": model_id,
         "messages": [{"role": "user", "content": content}],
         "stream": form_data.get("stream", False),
         "metadata": {
+            **(request.state.metadata if hasattr(request.state, "metadata") else {}),
             "chat_id": form_data.get("chat_id", None),
             "task": str(TASKS.MOA_RESPONSE_GENERATION),
             "task_body": form_data,
         },
     }
 
+    # Process the payload through the pipeline
+    try:
+        payload = await process_pipeline_inlet_filter(request, payload, user, models)
+    except Exception as e:
+        raise e
+
     try:
         return await generate_chat_completion(request, form_data=payload, user=user)
     except Exception as e:

backend/open_webui/routers/tools.py

@@ -1,5 +1,7 @@
+import logging
 from pathlib import Path
 from typing import Optional
+import time
 
 from open_webui.models.tools import (
     ToolForm,
@@ -8,13 +10,19 @@ from open_webui.models.tools import (
     ToolUserResponse,
     Tools,
 )
-from open_webui.utils.plugin import load_tools_module_by_id, replace_imports
+from open_webui.utils.plugin import load_tool_module_by_id, replace_imports
 from open_webui.config import CACHE_DIR
 from open_webui.constants import ERROR_MESSAGES
 from fastapi import APIRouter, Depends, HTTPException, Request, status
-from open_webui.utils.tools import get_tools_specs
+from open_webui.utils.tools import get_tool_specs
 from open_webui.utils.auth import get_admin_user, get_verified_user
 from open_webui.utils.access_control import has_access, has_permission
+from open_webui.env import SRC_LOG_LEVELS
+
+from open_webui.utils.tools import get_tool_servers_data
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["MAIN"])
 
 
 router = APIRouter()
@@ -25,11 +33,51 @@ router = APIRouter()
 
 
 @router.get("/", response_model=list[ToolUserResponse])
-async def get_tools(user=Depends(get_verified_user)):
-    if user.role == "admin":
-        tools = Tools.get_tools()
-    else:
-        tools = Tools.get_tools_by_user_id(user.id, "read")
+async def get_tools(request: Request, user=Depends(get_verified_user)):
+
+    if not request.app.state.TOOL_SERVERS:
+        # If the tool servers are not set, we need to set them
+        # This is done only once when the server starts
+        # This is done to avoid loading the tool servers every time
+
+        request.app.state.TOOL_SERVERS = await get_tool_servers_data(
+            request.app.state.config.TOOL_SERVER_CONNECTIONS
+        )
+
+    tools = Tools.get_tools()
+    for server in request.app.state.TOOL_SERVERS:
+        tools.append(
+            ToolUserResponse(
+                **{
+                    "id": f"server:{server['idx']}",
+                    "user_id": f"server:{server['idx']}",
+                    "name": server["openapi"]
+                    .get("info", {})
+                    .get("title", "Tool Server"),
+                    "meta": {
+                        "description": server["openapi"]
+                        .get("info", {})
+                        .get("description", ""),
+                    },
+                    "access_control": request.app.state.config.TOOL_SERVER_CONNECTIONS[
+                        server["idx"]
+                    ]
+                    .get("config", {})
+                    .get("access_control", None),
+                    "updated_at": int(time.time()),
+                    "created_at": int(time.time()),
+                }
+            )
+        )
+
+    if user.role != "admin":
+        tools = [
+            tool
+            for tool in tools
+            if tool.user_id == user.id
+            or has_access(user.id, "read", tool.access_control)
+        ]
+
     return tools
 
 
@@ -89,18 +137,18 @@ async def create_new_tools(
     if tools is None:
         try:
             form_data.content = replace_imports(form_data.content)
-            tools_module, frontmatter = load_tools_module_by_id(
+            tool_module, frontmatter = load_tool_module_by_id(
                 form_data.id, content=form_data.content
             )
             form_data.meta.manifest = frontmatter
 
             TOOLS = request.app.state.TOOLS
-            TOOLS[form_data.id] = tools_module
+            TOOLS[form_data.id] = tool_module
 
-            specs = get_tools_specs(TOOLS[form_data.id])
+            specs = get_tool_specs(TOOLS[form_data.id])
             tools = Tools.insert_new_tool(user.id, form_data, specs)
 
-            tool_cache_dir = Path(CACHE_DIR) / "tools" / form_data.id
+            tool_cache_dir = CACHE_DIR / "tools" / form_data.id
             tool_cache_dir.mkdir(parents=True, exist_ok=True)
 
             if tools:
@@ -111,7 +159,7 @@ async def create_new_tools(
                     detail=ERROR_MESSAGES.DEFAULT("Error creating tools"),
                 )
         except Exception as e:
-            print(e)
+            log.exception(f"Failed to load the tool by id {form_data.id}: {e}")
             raise HTTPException(
                 status_code=status.HTTP_400_BAD_REQUEST,
                 detail=ERROR_MESSAGES.DEFAULT(str(e)),
@@ -178,22 +226,20 @@ async def update_tools_by_id(
 
     try:
         form_data.content = replace_imports(form_data.content)
-        tools_module, frontmatter = load_tools_module_by_id(
-            id, content=form_data.content
-        )
+        tool_module, frontmatter = load_tool_module_by_id(id, content=form_data.content)
         form_data.meta.manifest = frontmatter
 
         TOOLS = request.app.state.TOOLS
-        TOOLS[id] = tools_module
+        TOOLS[id] = tool_module
 
-        specs = get_tools_specs(TOOLS[id])
+        specs = get_tool_specs(TOOLS[id])
 
         updated = {
             **form_data.model_dump(exclude={"id"}),
             "specs": specs,
         }
 
-        print(updated)
+        log.debug(updated)
         tools = Tools.update_tool_by_id(id, updated)
 
         if tools:
@@ -227,7 +273,11 @@ async def delete_tools_by_id(
             detail=ERROR_MESSAGES.NOT_FOUND,
         )
 
-    if tools.user_id != user.id and user.role != "admin":
+    if (
+        tools.user_id != user.id
+        and not has_access(user.id, "write", tools.access_control)
+        and user.role != "admin"
+    ):
         raise HTTPException(
             status_code=status.HTTP_401_UNAUTHORIZED,
             detail=ERROR_MESSAGES.UNAUTHORIZED,
@@ -280,7 +330,7 @@ async def get_tools_valves_spec_by_id(
         if id in request.app.state.TOOLS:
             tools_module = request.app.state.TOOLS[id]
         else:
-            tools_module, _ = load_tools_module_by_id(id)
+            tools_module, _ = load_tool_module_by_id(id)
             request.app.state.TOOLS[id] = tools_module
 
         if hasattr(tools_module, "Valves"):
@@ -323,7 +373,7 @@ async def update_tools_valves_by_id(
     if id in request.app.state.TOOLS:
         tools_module = request.app.state.TOOLS[id]
     else:
-        tools_module, _ = load_tools_module_by_id(id)
+        tools_module, _ = load_tool_module_by_id(id)
         request.app.state.TOOLS[id] = tools_module
 
     if not hasattr(tools_module, "Valves"):
@@ -339,7 +389,7 @@ async def update_tools_valves_by_id(
         Tools.update_tool_valves_by_id(id, valves.model_dump())
         return valves.model_dump()
     except Exception as e:
-        print(e)
+        log.exception(f"Failed to update tool valves by id {id}: {e}")
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=ERROR_MESSAGES.DEFAULT(str(e)),
@@ -379,7 +429,7 @@ async def get_tools_user_valves_spec_by_id(
         if id in request.app.state.TOOLS:
             tools_module = request.app.state.TOOLS[id]
         else:
-            tools_module, _ = load_tools_module_by_id(id)
+            tools_module, _ = load_tool_module_by_id(id)
             request.app.state.TOOLS[id] = tools_module
 
         if hasattr(tools_module, "UserValves"):
@@ -403,7 +453,7 @@ async def update_tools_user_valves_by_id(
         if id in request.app.state.TOOLS:
             tools_module = request.app.state.TOOLS[id]
         else:
-            tools_module, _ = load_tools_module_by_id(id)
+            tools_module, _ = load_tool_module_by_id(id)
             request.app.state.TOOLS[id] = tools_module
 
         if hasattr(tools_module, "UserValves"):
@@ -417,7 +467,7 @@ async def update_tools_user_valves_by_id(
                 )
                 return user_valves.model_dump()
             except Exception as e:
-                print(e)
+                log.exception(f"Failed to update user valves by id {id}: {e}")
                 raise HTTPException(
                     status_code=status.HTTP_400_BAD_REQUEST,
                     detail=ERROR_MESSAGES.DEFAULT(str(e)),

backend/open_webui/routers/users.py

@@ -2,6 +2,7 @@ import logging
 from typing import Optional
 
 from open_webui.models.auths import Auths
+from open_webui.models.groups import Groups
 from open_webui.models.chats import Chats
 from open_webui.models.users import (
     UserModel,
@@ -17,7 +18,10 @@ from open_webui.constants import ERROR_MESSAGES
 from open_webui.env import SRC_LOG_LEVELS
 from fastapi import APIRouter, Depends, HTTPException, Request, status
 from pydantic import BaseModel
+
 from open_webui.utils.auth import get_admin_user, get_password_hash, get_verified_user
+from open_webui.utils.access_control import get_permissions
+
 
 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["MODELS"])
@@ -45,7 +49,7 @@ async def get_users(
 
 @router.get("/groups")
 async def get_user_groups(user=Depends(get_verified_user)):
-    return Users.get_user_groups(user.id)
+    return Groups.get_groups_by_member_id(user.id)
 
 
 ############################
@@ -54,8 +58,12 @@ async def get_user_groups(user=Depends(get_verified_user)):
 
 
 @router.get("/permissions")
-async def get_user_permissisions(user=Depends(get_verified_user)):
-    return Users.get_user_groups(user.id)
+async def get_user_permissisions(request: Request, user=Depends(get_verified_user)):
+    user_permissions = get_permissions(
+        user.id, request.app.state.config.USER_PERMISSIONS
+    )
+
+    return user_permissions
 
 
 ############################
@@ -68,31 +76,46 @@ class WorkspacePermissions(BaseModel):
     tools: bool = False
 
 
+class SharingPermissions(BaseModel):
+    public_models: bool = True
+    public_knowledge: bool = True
+    public_prompts: bool = True
+    public_tools: bool = True
+
+
 class ChatPermissions(BaseModel):
     controls: bool = True
     file_upload: bool = True
     delete: bool = True
     edit: bool = True
+    multiple_models: bool = True
     temporary: bool = True
+    temporary_enforced: bool = False
 
 
 class FeaturesPermissions(BaseModel):
+    direct_tool_servers: bool = False
     web_search: bool = True
     image_generation: bool = True
+    code_interpreter: bool = True
 
 
 class UserPermissions(BaseModel):
     workspace: WorkspacePermissions
+    sharing: SharingPermissions
     chat: ChatPermissions
     features: FeaturesPermissions
 
 
 @router.get("/default/permissions", response_model=UserPermissions)
-async def get_user_permissions(request: Request, user=Depends(get_admin_user)):
+async def get_default_user_permissions(request: Request, user=Depends(get_admin_user)):
     return {
         "workspace": WorkspacePermissions(
             **request.app.state.config.USER_PERMISSIONS.get("workspace", {})
         ),
+        "sharing": SharingPermissions(
+            **request.app.state.config.USER_PERMISSIONS.get("sharing", {})
+        ),
         "chat": ChatPermissions(
             **request.app.state.config.USER_PERMISSIONS.get("chat", {})
         ),
@@ -103,7 +126,7 @@ async def get_user_permissions(request: Request, user=Depends(get_admin_user)):
 
 
 @router.post("/default/permissions")
-async def update_user_permissions(
+async def update_default_user_permissions(
     request: Request, form_data: UserPermissions, user=Depends(get_admin_user)
 ):
     request.app.state.config.USER_PERMISSIONS = form_data.model_dump()
@@ -152,7 +175,7 @@ async def get_user_settings_by_session_user(user=Depends(get_verified_user)):
 async def update_user_settings_by_session_user(
     form_data: UserSettings, user=Depends(get_verified_user)
 ):
-    user = Users.update_user_by_id(user.id, {"settings": form_data.model_dump()})
+    user = Users.update_user_settings_by_id(user.id, form_data.model_dump())
     if user:
         return user.settings
     else:

backend/open_webui/routers/utils.py

@@ -1,48 +1,84 @@
 import black
+import logging
 import markdown
 
 from open_webui.models.chats import ChatTitleMessagesForm
 from open_webui.config import DATA_DIR, ENABLE_ADMIN_EXPORT
 from open_webui.constants import ERROR_MESSAGES
-from fastapi import APIRouter, Depends, HTTPException, Response, status
+from fastapi import APIRouter, Depends, HTTPException, Request, Response, status
 from pydantic import BaseModel
 from starlette.responses import FileResponse
+
+
 from open_webui.utils.misc import get_gravatar_url
 from open_webui.utils.pdf_generator import PDFGenerator
-from open_webui.utils.auth import get_admin_user
+from open_webui.utils.auth import get_admin_user, get_verified_user
+from open_webui.utils.code_interpreter import execute_code_jupyter
+from open_webui.env import SRC_LOG_LEVELS
+
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["MAIN"])
 
 router = APIRouter()
 
 
 @router.get("/gravatar")
-async def get_gravatar(
-    email: str,
-):
+async def get_gravatar(email: str, user=Depends(get_verified_user)):
     return get_gravatar_url(email)
 
 
-class CodeFormatRequest(BaseModel):
+class CodeForm(BaseModel):
     code: str
 
 
 @router.post("/code/format")
-async def format_code(request: CodeFormatRequest):
+async def format_code(form_data: CodeForm, user=Depends(get_verified_user)):
     try:
-        formatted_code = black.format_str(request.code, mode=black.Mode())
+        formatted_code = black.format_str(form_data.code, mode=black.Mode())
         return {"code": formatted_code}
     except black.NothingChanged:
-        return {"code": request.code}
+        return {"code": form_data.code}
     except Exception as e:
         raise HTTPException(status_code=400, detail=str(e))
 
 
+@router.post("/code/execute")
+async def execute_code(
+    request: Request, form_data: CodeForm, user=Depends(get_verified_user)
+):
+    if request.app.state.config.CODE_EXECUTION_ENGINE == "jupyter":
+        output = await execute_code_jupyter(
+            request.app.state.config.CODE_EXECUTION_JUPYTER_URL,
+            form_data.code,
+            (
+                request.app.state.config.CODE_EXECUTION_JUPYTER_AUTH_TOKEN
+                if request.app.state.config.CODE_EXECUTION_JUPYTER_AUTH == "token"
+                else None
+            ),
+            (
+                request.app.state.config.CODE_EXECUTION_JUPYTER_AUTH_PASSWORD
+                if request.app.state.config.CODE_EXECUTION_JUPYTER_AUTH == "password"
+                else None
+            ),
+            request.app.state.config.CODE_EXECUTION_JUPYTER_TIMEOUT,
+        )
+
+        return output
+    else:
+        raise HTTPException(
+            status_code=400,
+            detail="Code execution engine not supported",
+        )
+
+
 class MarkdownForm(BaseModel):
     md: str
 
 
 @router.post("/markdown")
 async def get_html_from_markdown(
-    form_data: MarkdownForm,
+    form_data: MarkdownForm, user=Depends(get_verified_user)
 ):
     return {"html": markdown.markdown(form_data.md)}
 
@@ -54,7 +90,7 @@ class ChatForm(BaseModel):
 
 @router.post("/pdf")
 async def download_chat_as_pdf(
-    form_data: ChatTitleMessagesForm,
+    form_data: ChatTitleMessagesForm, user=Depends(get_verified_user)
 ):
     try:
         pdf_bytes = PDFGenerator(form_data).generate_chat_pdf()
@@ -65,7 +101,7 @@ async def download_chat_as_pdf(
             headers={"Content-Disposition": "attachment;filename=chat.pdf"},
         )
     except Exception as e:
-        print(e)
+        log.exception(f"Error generating PDF: {e}")
         raise HTTPException(status_code=400, detail=str(e))
 
 

backend/open_webui/socket/main.py

@@ -3,15 +3,23 @@ import socketio
 import logging
 import sys
 import time
+from redis import asyncio as aioredis
 
 from open_webui.models.users import Users, UserNameResponse
 from open_webui.models.channels import Channels
 from open_webui.models.chats import Chats
+from open_webui.utils.redis import (
+    get_sentinels_from_env,
+    get_sentinel_url_from_env,
+)
 
 from open_webui.env import (
     ENABLE_WEBSOCKET_SUPPORT,
     WEBSOCKET_MANAGER,
     WEBSOCKET_REDIS_URL,
+    WEBSOCKET_REDIS_LOCK_TIMEOUT,
+    WEBSOCKET_SENTINEL_PORT,
+    WEBSOCKET_SENTINEL_HOSTS,
 )
 from open_webui.utils.auth import decode_token
 from open_webui.socket.utils import RedisDict, RedisLock
@@ -28,7 +36,14 @@ log.setLevel(SRC_LOG_LEVELS["SOCKET"])
 
 
 if WEBSOCKET_MANAGER == "redis":
-    mgr = socketio.AsyncRedisManager(WEBSOCKET_REDIS_URL)
+    if WEBSOCKET_SENTINEL_HOSTS:
+        mgr = socketio.AsyncRedisManager(
+            get_sentinel_url_from_env(
+                WEBSOCKET_REDIS_URL, WEBSOCKET_SENTINEL_HOSTS, WEBSOCKET_SENTINEL_PORT
+            )
+        )
+    else:
+        mgr = socketio.AsyncRedisManager(WEBSOCKET_REDIS_URL)
     sio = socketio.AsyncServer(
         cors_allowed_origins=[],
         async_mode="asgi",
@@ -54,14 +69,30 @@ TIMEOUT_DURATION = 3
 
 if WEBSOCKET_MANAGER == "redis":
     log.debug("Using Redis to manage websockets.")
-    SESSION_POOL = RedisDict("open-webui:session_pool", redis_url=WEBSOCKET_REDIS_URL)
-    USER_POOL = RedisDict("open-webui:user_pool", redis_url=WEBSOCKET_REDIS_URL)
-    USAGE_POOL = RedisDict("open-webui:usage_pool", redis_url=WEBSOCKET_REDIS_URL)
+    redis_sentinels = get_sentinels_from_env(
+        WEBSOCKET_SENTINEL_HOSTS, WEBSOCKET_SENTINEL_PORT
+    )
+    SESSION_POOL = RedisDict(
+        "open-webui:session_pool",
+        redis_url=WEBSOCKET_REDIS_URL,
+        redis_sentinels=redis_sentinels,
+    )
+    USER_POOL = RedisDict(
+        "open-webui:user_pool",
+        redis_url=WEBSOCKET_REDIS_URL,
+        redis_sentinels=redis_sentinels,
+    )
+    USAGE_POOL = RedisDict(
+        "open-webui:usage_pool",
+        redis_url=WEBSOCKET_REDIS_URL,
+        redis_sentinels=redis_sentinels,
+    )
 
     clean_up_lock = RedisLock(
         redis_url=WEBSOCKET_REDIS_URL,
         lock_name="usage_cleanup_lock",
-        timeout_secs=TIMEOUT_DURATION * 2,
+        timeout_secs=WEBSOCKET_REDIS_LOCK_TIMEOUT,
+        redis_sentinels=redis_sentinels,
     )
     aquire_func = clean_up_lock.aquire_lock
     renew_func = clean_up_lock.renew_lock
@@ -268,76 +299,88 @@ async def disconnect(sid):
         # print(f"Unknown session ID {sid} disconnected")
 
 
-def get_event_emitter(request_info):
+def get_event_emitter(request_info, update_db=True):
     async def __event_emitter__(event_data):
         user_id = request_info["user_id"]
+
         session_ids = list(
-            set(USER_POOL.get(user_id, []) + [request_info["session_id"]])
+            set(
+                USER_POOL.get(user_id, [])
+                + (
+                    [request_info.get("session_id")]
+                    if request_info.get("session_id")
+                    else []
+                )
+            )
         )
 
         for session_id in session_ids:
             await sio.emit(
                 "chat-events",
                 {
-                    "chat_id": request_info["chat_id"],
-                    "message_id": request_info["message_id"],
+                    "chat_id": request_info.get("chat_id", None),
+                    "message_id": request_info.get("message_id", None),
                     "data": event_data,
                 },
                 to=session_id,
             )
 
-        if "type" in event_data and event_data["type"] == "status":
-            Chats.add_message_status_to_chat_by_id_and_message_id(
-                request_info["chat_id"],
-                request_info["message_id"],
-                event_data.get("data", {}),
-            )
+        if update_db:
+            if "type" in event_data and event_data["type"] == "status":
+                Chats.add_message_status_to_chat_by_id_and_message_id(
+                    request_info["chat_id"],
+                    request_info["message_id"],
+                    event_data.get("data", {}),
+                )
 
-        if "type" in event_data and event_data["type"] == "message":
-            message = Chats.get_message_by_id_and_message_id(
-                request_info["chat_id"],
-                request_info["message_id"],
-            )
+            if "type" in event_data and event_data["type"] == "message":
+                message = Chats.get_message_by_id_and_message_id(
+                    request_info["chat_id"],
+                    request_info["message_id"],
+                )
 
-            content = message.get("content", "")
-            content += event_data.get("data", {}).get("content", "")
+                content = message.get("content", "")
+                content += event_data.get("data", {}).get("content", "")
 
-            Chats.upsert_message_to_chat_by_id_and_message_id(
-                request_info["chat_id"],
-                request_info["message_id"],
-                {
-                    "content": content,
-                },
-            )
+                Chats.upsert_message_to_chat_by_id_and_message_id(
+                    request_info["chat_id"],
+                    request_info["message_id"],
+                    {
+                        "content": content,
+                    },
+                )
 
-        if "type" in event_data and event_data["type"] == "replace":
-            content = event_data.get("data", {}).get("content", "")
+            if "type" in event_data and event_data["type"] == "replace":
+                content = event_data.get("data", {}).get("content", "")
 
-            Chats.upsert_message_to_chat_by_id_and_message_id(
-                request_info["chat_id"],
-                request_info["message_id"],
-                {
-                    "content": content,
-                },
-            )
+                Chats.upsert_message_to_chat_by_id_and_message_id(
+                    request_info["chat_id"],
+                    request_info["message_id"],
+                    {
+                        "content": content,
+                    },
+                )
 
     return __event_emitter__
 
 
 def get_event_call(request_info):
-    async def __event_call__(event_data):
+    async def __event_caller__(event_data):
         response = await sio.call(
             "chat-events",
             {
-                "chat_id": request_info["chat_id"],
-                "message_id": request_info["message_id"],
+                "chat_id": request_info.get("chat_id", None),
+                "message_id": request_info.get("message_id", None),
                 "data": event_data,
             },
             to=request_info["session_id"],
         )
         return response
 
-    return __event_call__
+    return __event_caller__
+
+
+get_event_caller = get_event_call
 
 
 def get_user_id_from_session_pool(sid):

backend/open_webui/socket/utils.py

@@ -1,15 +1,17 @@
 import json
-import redis
 import uuid
+from open_webui.utils.redis import get_redis_connection
 
 
 class RedisLock:
-    def __init__(self, redis_url, lock_name, timeout_secs):
+    def __init__(self, redis_url, lock_name, timeout_secs, redis_sentinels=[]):
         self.lock_name = lock_name
         self.lock_id = str(uuid.uuid4())
         self.timeout_secs = timeout_secs
         self.lock_obtained = False
-        self.redis = redis.Redis.from_url(redis_url, decode_responses=True)
+        self.redis = get_redis_connection(
+            redis_url, redis_sentinels, decode_responses=True
+        )
 
     def aquire_lock(self):
         # nx=True will only set this key if it _hasn't_ already been set
@@ -31,9 +33,11 @@ class RedisLock:
 
 
 class RedisDict:
-    def __init__(self, name, redis_url):
+    def __init__(self, name, redis_url, redis_sentinels=[]):
         self.name = name
-        self.redis = redis.Redis.from_url(redis_url, decode_responses=True)
+        self.redis = get_redis_connection(
+            redis_url, redis_sentinels, decode_responses=True
+        )
 
     def __setitem__(self, key, value):
         serialized_value = json.dumps(value)

backend/open_webui/static/site.webmanifest

@@ -3,13 +3,13 @@
   "short_name": "WebUI",
   "icons": [
     {
-      "src": "/favicon/web-app-manifest-192x192.png",
+      "src": "/static/web-app-manifest-192x192.png",
       "sizes": "192x192",
       "type": "image/png",
       "purpose": "maskable"
     },
     {
-      "src": "/favicon/web-app-manifest-512x512.png",
+      "src": "/static/web-app-manifest-512x512.png",
       "sizes": "512x512",
       "type": "image/png",
       "purpose": "maskable"

backend/open_webui/static/swagger-ui/swagger-ui.css

@@ -9308,5 +9308,3 @@
 	.json-schema-2020-12__title:first-of-type {
 	font-size: 16px;
 }
-
-/*# sourceMappingURL=swagger-ui.css.map*/

backend/open_webui/storage/provider.py

@@ -1,25 +1,41 @@
 import os
 import shutil
 import json
+import logging
 from abc import ABC, abstractmethod
 from typing import BinaryIO, Tuple
 
 import boto3
+from botocore.config import Config
 from botocore.exceptions import ClientError
 from open_webui.config import (
     S3_ACCESS_KEY_ID,
     S3_BUCKET_NAME,
     S3_ENDPOINT_URL,
+    S3_KEY_PREFIX,
     S3_REGION_NAME,
     S3_SECRET_ACCESS_KEY,
+    S3_USE_ACCELERATE_ENDPOINT,
+    S3_ADDRESSING_STYLE,
     GCS_BUCKET_NAME,
     GOOGLE_APPLICATION_CREDENTIALS_JSON,
+    AZURE_STORAGE_ENDPOINT,
+    AZURE_STORAGE_CONTAINER_NAME,
+    AZURE_STORAGE_KEY,
     STORAGE_PROVIDER,
     UPLOAD_DIR,
 )
 from google.cloud import storage
 from google.cloud.exceptions import GoogleCloudError, NotFound
 from open_webui.constants import ERROR_MESSAGES
+from azure.identity import DefaultAzureCredential
+from azure.storage.blob import BlobServiceClient
+from azure.core.exceptions import ResourceNotFoundError
+from open_webui.env import SRC_LOG_LEVELS
+
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["MAIN"])
 
 
 class StorageProvider(ABC):
@@ -64,7 +80,7 @@ class LocalStorageProvider(StorageProvider):
         if os.path.isfile(file_path):
             os.remove(file_path)
         else:
-            print(f"File {file_path} not found in local storage.")
+            log.warning(f"File {file_path} not found in local storage.")
 
     @staticmethod
     def delete_all_files() -> None:
@@ -78,30 +94,52 @@ class LocalStorageProvider(StorageProvider):
                     elif os.path.isdir(file_path):
                         shutil.rmtree(file_path)  # Remove the directory
                 except Exception as e:
-                    print(f"Failed to delete {file_path}. Reason: {e}")
+                    log.exception(f"Failed to delete {file_path}. Reason: {e}")
         else:
-            print(f"Directory {UPLOAD_DIR} not found in local storage.")
+            log.warning(f"Directory {UPLOAD_DIR} not found in local storage.")
 
 
 class S3StorageProvider(StorageProvider):
     def __init__(self):
-        self.s3_client = boto3.client(
-            "s3",
-            region_name=S3_REGION_NAME,
-            endpoint_url=S3_ENDPOINT_URL,
-            aws_access_key_id=S3_ACCESS_KEY_ID,
-            aws_secret_access_key=S3_SECRET_ACCESS_KEY,
+        config = Config(
+            s3={
+                "use_accelerate_endpoint": S3_USE_ACCELERATE_ENDPOINT,
+                "addressing_style": S3_ADDRESSING_STYLE,
+            },
         )
+
+        # If access key and secret are provided, use them for authentication
+        if S3_ACCESS_KEY_ID and S3_SECRET_ACCESS_KEY:
+            self.s3_client = boto3.client(
+                "s3",
+                region_name=S3_REGION_NAME,
+                endpoint_url=S3_ENDPOINT_URL,
+                aws_access_key_id=S3_ACCESS_KEY_ID,
+                aws_secret_access_key=S3_SECRET_ACCESS_KEY,
+                config=config,
+            )
+        else:
+            # If no explicit credentials are provided, fall back to default AWS credentials
+            # This supports workload identity (IAM roles for EC2, EKS, etc.)
+            self.s3_client = boto3.client(
+                "s3",
+                region_name=S3_REGION_NAME,
+                endpoint_url=S3_ENDPOINT_URL,
+                config=config,
+            )
+
         self.bucket_name = S3_BUCKET_NAME
+        self.key_prefix = S3_KEY_PREFIX if S3_KEY_PREFIX else ""
 
     def upload_file(self, file: BinaryIO, filename: str) -> Tuple[bytes, str]:
         """Handles uploading of the file to S3 storage."""
         _, file_path = LocalStorageProvider.upload_file(file, filename)
         try:
-            self.s3_client.upload_file(file_path, self.bucket_name, filename)
+            s3_key = os.path.join(self.key_prefix, filename)
+            self.s3_client.upload_file(file_path, self.bucket_name, s3_key)
             return (
                 open(file_path, "rb").read(),
-                "s3://" + self.bucket_name + "/" + filename,
+                "s3://" + self.bucket_name + "/" + s3_key,
             )
         except ClientError as e:
             raise RuntimeError(f"Error uploading file to S3: {e}")
@@ -109,18 +147,18 @@ class S3StorageProvider(StorageProvider):
     def get_file(self, file_path: str) -> str:
         """Handles downloading of the file from S3 storage."""
         try:
-            bucket_name, key = file_path.split("//")[1].split("/")
-            local_file_path = f"{UPLOAD_DIR}/{key}"
-            self.s3_client.download_file(bucket_name, key, local_file_path)
+            s3_key = self._extract_s3_key(file_path)
+            local_file_path = self._get_local_file_path(s3_key)
+            self.s3_client.download_file(self.bucket_name, s3_key, local_file_path)
             return local_file_path
         except ClientError as e:
             raise RuntimeError(f"Error downloading file from S3: {e}")
 
     def delete_file(self, file_path: str) -> None:
         """Handles deletion of the file from S3 storage."""
-        filename = file_path.split("/")[-1]
         try:
-            self.s3_client.delete_object(Bucket=self.bucket_name, Key=filename)
+            s3_key = self._extract_s3_key(file_path)
+            self.s3_client.delete_object(Bucket=self.bucket_name, Key=s3_key)
         except ClientError as e:
             raise RuntimeError(f"Error deleting file from S3: {e}")
 
@@ -133,6 +171,10 @@ class S3StorageProvider(StorageProvider):
             response = self.s3_client.list_objects_v2(Bucket=self.bucket_name)
             if "Contents" in response:
                 for content in response["Contents"]:
+                    # Skip objects that were not uploaded from open-webui in the first place
+                    if not content["Key"].startswith(self.key_prefix):
+                        continue
+
                     self.s3_client.delete_object(
                         Bucket=self.bucket_name, Key=content["Key"]
                     )
@@ -142,6 +184,13 @@ class S3StorageProvider(StorageProvider):
         # Always delete from local storage
         LocalStorageProvider.delete_all_files()
 
+    # The s3 key is the name assigned to an object. It excludes the bucket name, but includes the internal path and the file name.
+    def _extract_s3_key(self, full_file_path: str) -> str:
+        return "/".join(full_file_path.split("//")[1].split("/")[1:])
+
+    def _get_local_file_path(self, s3_key: str) -> str:
+        return f"{UPLOAD_DIR}/{s3_key.split('/')[-1]}"
+
 
 class GCSStorageProvider(StorageProvider):
     def __init__(self):
@@ -207,6 +256,74 @@ class GCSStorageProvider(StorageProvider):
         LocalStorageProvider.delete_all_files()
 
 
+class AzureStorageProvider(StorageProvider):
+    def __init__(self):
+        self.endpoint = AZURE_STORAGE_ENDPOINT
+        self.container_name = AZURE_STORAGE_CONTAINER_NAME
+        storage_key = AZURE_STORAGE_KEY
+
+        if storage_key:
+            # Configure using the Azure Storage Account Endpoint and Key
+            self.blob_service_client = BlobServiceClient(
+                account_url=self.endpoint, credential=storage_key
+            )
+        else:
+            # Configure using the Azure Storage Account Endpoint and DefaultAzureCredential
+            # If the key is not configured, then the DefaultAzureCredential will be used to support Managed Identity authentication
+            self.blob_service_client = BlobServiceClient(
+                account_url=self.endpoint, credential=DefaultAzureCredential()
+            )
+        self.container_client = self.blob_service_client.get_container_client(
+            self.container_name
+        )
+
+    def upload_file(self, file: BinaryIO, filename: str) -> Tuple[bytes, str]:
+        """Handles uploading of the file to Azure Blob Storage."""
+        contents, file_path = LocalStorageProvider.upload_file(file, filename)
+        try:
+            blob_client = self.container_client.get_blob_client(filename)
+            blob_client.upload_blob(contents, overwrite=True)
+            return contents, f"{self.endpoint}/{self.container_name}/{filename}"
+        except Exception as e:
+            raise RuntimeError(f"Error uploading file to Azure Blob Storage: {e}")
+
+    def get_file(self, file_path: str) -> str:
+        """Handles downloading of the file from Azure Blob Storage."""
+        try:
+            filename = file_path.split("/")[-1]
+            local_file_path = f"{UPLOAD_DIR}/{filename}"
+            blob_client = self.container_client.get_blob_client(filename)
+            with open(local_file_path, "wb") as download_file:
+                download_file.write(blob_client.download_blob().readall())
+            return local_file_path
+        except ResourceNotFoundError as e:
+            raise RuntimeError(f"Error downloading file from Azure Blob Storage: {e}")
+
+    def delete_file(self, file_path: str) -> None:
+        """Handles deletion of the file from Azure Blob Storage."""
+        try:
+            filename = file_path.split("/")[-1]
+            blob_client = self.container_client.get_blob_client(filename)
+            blob_client.delete_blob()
+        except ResourceNotFoundError as e:
+            raise RuntimeError(f"Error deleting file from Azure Blob Storage: {e}")
+
+        # Always delete from local storage
+        LocalStorageProvider.delete_file(file_path)
+
+    def delete_all_files(self) -> None:
+        """Handles deletion of all files from Azure Blob Storage."""
+        try:
+            blobs = self.container_client.list_blobs()
+            for blob in blobs:
+                self.container_client.delete_blob(blob.name)
+        except Exception as e:
+            raise RuntimeError(f"Error deleting all files from Azure Blob Storage: {e}")
+
+        # Always delete from local storage
+        LocalStorageProvider.delete_all_files()
+
+
 def get_storage_provider(storage_provider: str):
     if storage_provider == "local":
         Storage = LocalStorageProvider()
@@ -214,6 +331,8 @@ def get_storage_provider(storage_provider: str):
         Storage = S3StorageProvider()
     elif storage_provider == "gcs":
         Storage = GCSStorageProvider()
+    elif storage_provider == "azure":
+        Storage = AzureStorageProvider()
     else:
         raise RuntimeError(f"Unsupported storage provider: {storage_provider}")
     return Storage

backend/open_webui/tasks.py

@@ -5,16 +5,23 @@ from uuid import uuid4
 
 # A dictionary to keep track of active tasks
 tasks: Dict[str, asyncio.Task] = {}
+chat_tasks = {}
 
 
-def cleanup_task(task_id: str):
+def cleanup_task(task_id: str, id=None):
     """
     Remove a completed or canceled task from the global `tasks` dictionary.
     """
     tasks.pop(task_id, None)  # Remove the task if it exists
 
+    # If an ID is provided, remove the task from the chat_tasks dictionary
+    if id and task_id in chat_tasks.get(id, []):
+        chat_tasks[id].remove(task_id)
+        if not chat_tasks[id]:  # If no tasks left for this ID, remove the entry
+            chat_tasks.pop(id, None)
 
-def create_task(coroutine):
+
+def create_task(coroutine, id=None):
     """
     Create a new asyncio task and add it to the global task dictionary.
     """
@@ -22,9 +29,15 @@ def create_task(coroutine):
     task = asyncio.create_task(coroutine)  # Create the task
 
     # Add a done callback for cleanup
-    task.add_done_callback(lambda t: cleanup_task(task_id))
-
+    task.add_done_callback(lambda t: cleanup_task(task_id, id))
     tasks[task_id] = task
+
+    # If an ID is provided, associate the task with that ID
+    if chat_tasks.get(id):
+        chat_tasks[id].append(task_id)
+    else:
+        chat_tasks[id] = [task_id]
+
     return task_id, task
 
 
@@ -42,6 +55,13 @@ def list_tasks():
     return list(tasks.keys())
 
 
+def list_task_ids_by_chat_id(id):
+    """
+    List all tasks associated with a specific ID.
+    """
+    return chat_tasks.get(id, [])
+
+
 async def stop_task(task_id: str):
     """
     Cancel a running task and remove it from the global task list.

backend/open_webui/test/apps/webui/storage/test_provider.py

@@ -7,6 +7,8 @@ from moto import mock_aws
 from open_webui.storage import provider
 from gcp_storage_emulator.server import create_server
 from google.cloud import storage
+from azure.storage.blob import BlobServiceClient, ContainerClient, BlobClient
+from unittest.mock import MagicMock
 
 
 def mock_upload_dir(monkeypatch, tmp_path):
@@ -22,6 +24,7 @@ def test_imports():
     provider.LocalStorageProvider
     provider.S3StorageProvider
     provider.GCSStorageProvider
+    provider.AzureStorageProvider
     provider.Storage
 
 
@@ -32,6 +35,8 @@ def test_get_storage_provider():
     assert isinstance(Storage, provider.S3StorageProvider)
     Storage = provider.get_storage_provider("gcs")
     assert isinstance(Storage, provider.GCSStorageProvider)
+    Storage = provider.get_storage_provider("azure")
+    assert isinstance(Storage, provider.AzureStorageProvider)
     with pytest.raises(RuntimeError):
         provider.get_storage_provider("invalid")
 
@@ -48,6 +53,7 @@ def test_class_instantiation():
     provider.LocalStorageProvider()
     provider.S3StorageProvider()
     provider.GCSStorageProvider()
+    provider.AzureStorageProvider()
 
 
 class TestLocalStorageProvider:
@@ -181,6 +187,17 @@ class TestS3StorageProvider:
         assert not (upload_dir / self.filename).exists()
         assert not (upload_dir / self.filename_extra).exists()
 
+    def test_init_without_credentials(self, monkeypatch):
+        """Test that S3StorageProvider can initialize without explicit credentials."""
+        # Temporarily unset the environment variables
+        monkeypatch.setattr(provider, "S3_ACCESS_KEY_ID", None)
+        monkeypatch.setattr(provider, "S3_SECRET_ACCESS_KEY", None)
+
+        # Should not raise an exception
+        storage = provider.S3StorageProvider()
+        assert storage.s3_client is not None
+        assert storage.bucket_name == provider.S3_BUCKET_NAME
+
 
 class TestGCSStorageProvider:
     Storage = provider.GCSStorageProvider()
@@ -272,3 +289,147 @@ class TestGCSStorageProvider:
         assert not (upload_dir / self.filename_extra).exists()
         assert self.Storage.bucket.get_blob(self.filename) == None
         assert self.Storage.bucket.get_blob(self.filename_extra) == None
+
+
+class TestAzureStorageProvider:
+    def __init__(self):
+        super().__init__()
+
+    @pytest.fixture(scope="class")
+    def setup_storage(self, monkeypatch):
+        # Create mock Blob Service Client and related clients
+        mock_blob_service_client = MagicMock()
+        mock_container_client = MagicMock()
+        mock_blob_client = MagicMock()
+
+        # Set up return values for the mock
+        mock_blob_service_client.get_container_client.return_value = (
+            mock_container_client
+        )
+        mock_container_client.get_blob_client.return_value = mock_blob_client
+
+        # Monkeypatch the Azure classes to return our mocks
+        monkeypatch.setattr(
+            azure.storage.blob,
+            "BlobServiceClient",
+            lambda *args, **kwargs: mock_blob_service_client,
+        )
+        monkeypatch.setattr(
+            azure.storage.blob,
+            "ContainerClient",
+            lambda *args, **kwargs: mock_container_client,
+        )
+        monkeypatch.setattr(
+            azure.storage.blob, "BlobClient", lambda *args, **kwargs: mock_blob_client
+        )
+
+        self.Storage = provider.AzureStorageProvider()
+        self.Storage.endpoint = "https://myaccount.blob.core.windows.net"
+        self.Storage.container_name = "my-container"
+        self.file_content = b"test content"
+        self.filename = "test.txt"
+        self.filename_extra = "test_extra.txt"
+        self.file_bytesio_empty = io.BytesIO()
+
+        # Apply mocks to the Storage instance
+        self.Storage.blob_service_client = mock_blob_service_client
+        self.Storage.container_client = mock_container_client
+
+    def test_upload_file(self, monkeypatch, tmp_path):
+        upload_dir = mock_upload_dir(monkeypatch, tmp_path)
+
+        # Simulate an error when container does not exist
+        self.Storage.container_client.get_blob_client.side_effect = Exception(
+            "Container does not exist"
+        )
+        with pytest.raises(Exception):
+            self.Storage.upload_file(io.BytesIO(self.file_content), self.filename)
+
+        # Reset side effect and create container
+        self.Storage.container_client.get_blob_client.side_effect = None
+        self.Storage.create_container()
+        contents, azure_file_path = self.Storage.upload_file(
+            io.BytesIO(self.file_content), self.filename
+        )
+
+        # Assertions
+        self.Storage.container_client.get_blob_client.assert_called_with(self.filename)
+        self.Storage.container_client.get_blob_client().upload_blob.assert_called_once_with(
+            self.file_content, overwrite=True
+        )
+        assert contents == self.file_content
+        assert (
+            azure_file_path
+            == f"https://myaccount.blob.core.windows.net/{self.Storage.container_name}/{self.filename}"
+        )
+        assert (upload_dir / self.filename).exists()
+        assert (upload_dir / self.filename).read_bytes() == self.file_content
+
+        with pytest.raises(ValueError):
+            self.Storage.upload_file(self.file_bytesio_empty, self.filename)
+
+    def test_get_file(self, monkeypatch, tmp_path):
+        upload_dir = mock_upload_dir(monkeypatch, tmp_path)
+        self.Storage.create_container()
+
+        # Mock upload behavior
+        self.Storage.upload_file(io.BytesIO(self.file_content), self.filename)
+        # Mock blob download behavior
+        self.Storage.container_client.get_blob_client().download_blob().readall.return_value = (
+            self.file_content
+        )
+
+        file_url = f"https://myaccount.blob.core.windows.net/{self.Storage.container_name}/{self.filename}"
+        file_path = self.Storage.get_file(file_url)
+
+        assert file_path == str(upload_dir / self.filename)
+        assert (upload_dir / self.filename).exists()
+        assert (upload_dir / self.filename).read_bytes() == self.file_content
+
+    def test_delete_file(self, monkeypatch, tmp_path):
+        upload_dir = mock_upload_dir(monkeypatch, tmp_path)
+        self.Storage.create_container()
+
+        # Mock file upload
+        self.Storage.upload_file(io.BytesIO(self.file_content), self.filename)
+        # Mock deletion
+        self.Storage.container_client.get_blob_client().delete_blob.return_value = None
+
+        file_url = f"https://myaccount.blob.core.windows.net/{self.Storage.container_name}/{self.filename}"
+        self.Storage.delete_file(file_url)
+
+        self.Storage.container_client.get_blob_client().delete_blob.assert_called_once()
+        assert not (upload_dir / self.filename).exists()
+
+    def test_delete_all_files(self, monkeypatch, tmp_path):
+        upload_dir = mock_upload_dir(monkeypatch, tmp_path)
+        self.Storage.create_container()
+
+        # Mock file uploads
+        self.Storage.upload_file(io.BytesIO(self.file_content), self.filename)
+        self.Storage.upload_file(io.BytesIO(self.file_content), self.filename_extra)
+
+        # Mock listing and deletion behavior
+        self.Storage.container_client.list_blobs.return_value = [
+            {"name": self.filename},
+            {"name": self.filename_extra},
+        ]
+        self.Storage.container_client.get_blob_client().delete_blob.return_value = None
+
+        self.Storage.delete_all_files()
+
+        self.Storage.container_client.list_blobs.assert_called_once()
+        self.Storage.container_client.get_blob_client().delete_blob.assert_any_call()
+        assert not (upload_dir / self.filename).exists()
+        assert not (upload_dir / self.filename_extra).exists()
+
+    def test_get_file_not_found(self, monkeypatch):
+        self.Storage.create_container()
+
+        file_url = f"https://myaccount.blob.core.windows.net/{self.Storage.container_name}/{self.filename}"
+        # Mock behavior to raise an error for missing blobs
+        self.Storage.container_client.get_blob_client().download_blob.side_effect = (
+            Exception("Blob not found")
+        )
+        with pytest.raises(Exception, match="Blob not found"):
+            self.Storage.get_file(file_url)

backend/open_webui/utils/audit.py

@@ -0,0 +1,249 @@
+from contextlib import asynccontextmanager
+from dataclasses import asdict, dataclass
+from enum import Enum
+import re
+from typing import (
+    TYPE_CHECKING,
+    Any,
+    AsyncGenerator,
+    Dict,
+    MutableMapping,
+    Optional,
+    cast,
+)
+import uuid
+
+from asgiref.typing import (
+    ASGI3Application,
+    ASGIReceiveCallable,
+    ASGIReceiveEvent,
+    ASGISendCallable,
+    ASGISendEvent,
+    Scope as ASGIScope,
+)
+from loguru import logger
+from starlette.requests import Request
+
+from open_webui.env import AUDIT_LOG_LEVEL, MAX_BODY_LOG_SIZE
+from open_webui.utils.auth import get_current_user, get_http_authorization_cred
+from open_webui.models.users import UserModel
+
+
+if TYPE_CHECKING:
+    from loguru import Logger
+
+
+@dataclass(frozen=True)
+class AuditLogEntry:
+    # `Metadata` audit level properties
+    id: str
+    user: dict[str, Any]
+    audit_level: str
+    verb: str
+    request_uri: str
+    user_agent: Optional[str] = None
+    source_ip: Optional[str] = None
+    # `Request` audit level properties
+    request_object: Any = None
+    # `Request Response` level
+    response_object: Any = None
+    response_status_code: Optional[int] = None
+
+
+class AuditLevel(str, Enum):
+    NONE = "NONE"
+    METADATA = "METADATA"
+    REQUEST = "REQUEST"
+    REQUEST_RESPONSE = "REQUEST_RESPONSE"
+
+
+class AuditLogger:
+    """
+    A helper class that encapsulates audit logging functionality. It uses Loguru’s logger with an auditable binding to ensure that audit log entries are filtered correctly.
+
+    Parameters:
+    logger (Logger): An instance of Loguru’s logger.
+    """
+
+    def __init__(self, logger: "Logger"):
+        self.logger = logger.bind(auditable=True)
+
+    def write(
+        self,
+        audit_entry: AuditLogEntry,
+        *,
+        log_level: str = "INFO",
+        extra: Optional[dict] = None,
+    ):
+
+        entry = asdict(audit_entry)
+
+        if extra:
+            entry["extra"] = extra
+
+        self.logger.log(
+            log_level,
+            "",
+            **entry,
+        )
+
+
+class AuditContext:
+    """
+    Captures and aggregates the HTTP request and response bodies during the processing of a request. It ensures that only a configurable maximum amount of data is stored to prevent excessive memory usage.
+
+    Attributes:
+    request_body (bytearray): Accumulated request payload.
+    response_body (bytearray): Accumulated response payload.
+    max_body_size (int): Maximum number of bytes to capture.
+    metadata (Dict[str, Any]): A dictionary to store additional audit metadata (user, http verb, user agent, etc.).
+    """
+
+    def __init__(self, max_body_size: int = MAX_BODY_LOG_SIZE):
+        self.request_body = bytearray()
+        self.response_body = bytearray()
+        self.max_body_size = max_body_size
+        self.metadata: Dict[str, Any] = {}
+
+    def add_request_chunk(self, chunk: bytes):
+        if len(self.request_body) < self.max_body_size:
+            self.request_body.extend(
+                chunk[: self.max_body_size - len(self.request_body)]
+            )
+
+    def add_response_chunk(self, chunk: bytes):
+        if len(self.response_body) < self.max_body_size:
+            self.response_body.extend(
+                chunk[: self.max_body_size - len(self.response_body)]
+            )
+
+
+class AuditLoggingMiddleware:
+    """
+    ASGI middleware that intercepts HTTP requests and responses to perform audit logging. It captures request/response bodies (depending on audit level), headers, HTTP methods, and user information, then logs a structured audit entry at the end of the request cycle.
+    """
+
+    AUDITED_METHODS = {"PUT", "PATCH", "DELETE", "POST"}
+
+    def __init__(
+        self,
+        app: ASGI3Application,
+        *,
+        excluded_paths: Optional[list[str]] = None,
+        max_body_size: int = MAX_BODY_LOG_SIZE,
+        audit_level: AuditLevel = AuditLevel.NONE,
+    ) -> None:
+        self.app = app
+        self.audit_logger = AuditLogger(logger)
+        self.excluded_paths = excluded_paths or []
+        self.max_body_size = max_body_size
+        self.audit_level = audit_level
+
+    async def __call__(
+        self,
+        scope: ASGIScope,
+        receive: ASGIReceiveCallable,
+        send: ASGISendCallable,
+    ) -> None:
+        if scope["type"] != "http":
+            return await self.app(scope, receive, send)
+
+        request = Request(scope=cast(MutableMapping, scope))
+
+        if self._should_skip_auditing(request):
+            return await self.app(scope, receive, send)
+
+        async with self._audit_context(request) as context:
+
+            async def send_wrapper(message: ASGISendEvent) -> None:
+                if self.audit_level == AuditLevel.REQUEST_RESPONSE:
+                    await self._capture_response(message, context)
+
+                await send(message)
+
+            original_receive = receive
+
+            async def receive_wrapper() -> ASGIReceiveEvent:
+                nonlocal original_receive
+                message = await original_receive()
+
+                if self.audit_level in (
+                    AuditLevel.REQUEST,
+                    AuditLevel.REQUEST_RESPONSE,
+                ):
+                    await self._capture_request(message, context)
+
+                return message
+
+            await self.app(scope, receive_wrapper, send_wrapper)
+
+    @asynccontextmanager
+    async def _audit_context(
+        self, request: Request
+    ) -> AsyncGenerator[AuditContext, None]:
+        """
+        async context manager that ensures that an audit log entry is recorded after the request is processed.
+        """
+        context = AuditContext()
+        try:
+            yield context
+        finally:
+            await self._log_audit_entry(request, context)
+
+    async def _get_authenticated_user(self, request: Request) -> UserModel:
+
+        auth_header = request.headers.get("Authorization")
+        assert auth_header
+        user = get_current_user(request, None, get_http_authorization_cred(auth_header))
+
+        return user
+
+    def _should_skip_auditing(self, request: Request) -> bool:
+        if (
+            request.method not in {"POST", "PUT", "PATCH", "DELETE"}
+            or AUDIT_LOG_LEVEL == "NONE"
+            or not request.headers.get("authorization")
+        ):
+            return True
+        # match either /api/<resource>/...(for the endpoint /api/chat case) or /api/v1/<resource>/...
+        pattern = re.compile(
+            r"^/api(?:/v1)?/(" + "|".join(self.excluded_paths) + r")\b"
+        )
+        if pattern.match(request.url.path):
+            return True
+
+        return False
+
+    async def _capture_request(self, message: ASGIReceiveEvent, context: AuditContext):
+        if message["type"] == "http.request":
+            body = message.get("body", b"")
+            context.add_request_chunk(body)
+
+    async def _capture_response(self, message: ASGISendEvent, context: AuditContext):
+        if message["type"] == "http.response.start":
+            context.metadata["response_status_code"] = message["status"]
+
+        elif message["type"] == "http.response.body":
+            body = message.get("body", b"")
+            context.add_response_chunk(body)
+
+    async def _log_audit_entry(self, request: Request, context: AuditContext):
+        try:
+            user = await self._get_authenticated_user(request)
+
+            entry = AuditLogEntry(
+                id=str(uuid.uuid4()),
+                user=user.model_dump(include={"id", "name", "email", "role"}),
+                audit_level=self.audit_level.value,
+                verb=request.method,
+                request_uri=str(request.url),
+                response_status_code=context.metadata.get("response_status_code", None),
+                source_ip=request.client.host if request.client else None,
+                user_agent=request.headers.get("user-agent"),
+                request_object=context.request_body.decode("utf-8", errors="replace"),
+                response_object=context.response_body.decode("utf-8", errors="replace"),
+            )
+
+            self.audit_logger.write(entry)
+        except Exception as e:
+            logger.error(f"Failed to log audit entry: {str(e)}")

backend/open_webui/utils/auth.py

@@ -1,21 +1,37 @@
 import logging
 import uuid
 import jwt
+import base64
+import hmac
+import hashlib
+import requests
+import os
 
-from datetime import UTC, datetime, timedelta
+
+from datetime import datetime, timedelta
+import pytz
+from pytz import UTC
 from typing import Optional, Union, List, Dict
 
 from open_webui.models.users import Users
 
 from open_webui.constants import ERROR_MESSAGES
-from open_webui.env import WEBUI_SECRET_KEY
+from open_webui.env import (
+    WEBUI_SECRET_KEY,
+    TRUSTED_SIGNATURE_KEY,
+    STATIC_DIR,
+    SRC_LOG_LEVELS,
+)
 
-from fastapi import Depends, HTTPException, Request, Response, status
+from fastapi import BackgroundTasks, Depends, HTTPException, Request, Response, status
 from fastapi.security import HTTPAuthorizationCredentials, HTTPBearer
 from passlib.context import CryptContext
 
+
 logging.getLogger("passlib").setLevel(logging.ERROR)
 
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["OAUTH"])
 
 SESSION_SECRET = WEBUI_SECRET_KEY
 ALGORITHM = "HS256"
@@ -24,6 +40,67 @@ ALGORITHM = "HS256"
 # Auth Utils
 ##############
 
+
+def verify_signature(payload: str, signature: str) -> bool:
+    """
+    Verifies the HMAC signature of the received payload.
+    """
+    try:
+        expected_signature = base64.b64encode(
+            hmac.new(TRUSTED_SIGNATURE_KEY, payload.encode(), hashlib.sha256).digest()
+        ).decode()
+
+        # Compare securely to prevent timing attacks
+        return hmac.compare_digest(expected_signature, signature)
+
+    except Exception:
+        return False
+
+
+def override_static(path: str, content: str):
+    # Ensure path is safe
+    if "/" in path or ".." in path:
+        log.error(f"Invalid path: {path}")
+        return
+
+    file_path = os.path.join(STATIC_DIR, path)
+    os.makedirs(os.path.dirname(file_path), exist_ok=True)
+
+    with open(file_path, "wb") as f:
+        f.write(base64.b64decode(content))  # Convert Base64 back to raw binary
+
+
+def get_license_data(app, key):
+    if key:
+        try:
+            res = requests.post(
+                "https://api.openwebui.com/api/v1/license/",
+                json={"key": key, "version": "1"},
+                timeout=5,
+            )
+
+            if getattr(res, "ok", False):
+                payload = getattr(res, "json", lambda: {})()
+                for k, v in payload.items():
+                    if k == "resources":
+                        for p, c in v.items():
+                            globals().get("override_static", lambda a, b: None)(p, c)
+                    elif k == "count":
+                        setattr(app.state, "USER_COUNT", v)
+                    elif k == "name":
+                        setattr(app.state, "WEBUI_NAME", v)
+                    elif k == "metadata":
+                        setattr(app.state, "LICENSE_METADATA", v)
+                return True
+            else:
+                log.error(
+                    f"License: retrieval issue: {getattr(res, 'text', 'unknown error')}"
+                )
+        except Exception as ex:
+            log.exception(f"License: Uncaught Exception: {ex}")
+    return False
+
+
 bearer_security = HTTPBearer(auto_error=False)
 pwd_context = CryptContext(schemes=["bcrypt"], deprecated="auto")
 
@@ -66,16 +143,19 @@ def create_api_key():
     return f"sk-{key}"
 
 
-def get_http_authorization_cred(auth_header: str):
+def get_http_authorization_cred(auth_header: Optional[str]):
+    if not auth_header:
+        return None
     try:
         scheme, credentials = auth_header.split(" ")
         return HTTPAuthorizationCredentials(scheme=scheme, credentials=credentials)
     except Exception:
-        raise ValueError(ERROR_MESSAGES.INVALID_TOKEN)
+        return None
 
 
 def get_current_user(
     request: Request,
+    background_tasks: BackgroundTasks,
     auth_token: HTTPAuthorizationCredentials = Depends(bearer_security),
 ):
     token = None
@@ -104,7 +184,12 @@ def get_current_user(
                 ).split(",")
             ]
 
-            if request.url.path not in allowed_paths:
+            # Check if the request path matches any allowed endpoint.
+            if not any(
+                request.url.path == allowed
+                or request.url.path.startswith(allowed + "/")
+                for allowed in allowed_paths
+            ):
                 raise HTTPException(
                     status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.API_KEY_NOT_ALLOWED
                 )
@@ -128,7 +213,10 @@ def get_current_user(
                 detail=ERROR_MESSAGES.INVALID_TOKEN,
             )
         else:
-            Users.update_user_last_active_by_id(user.id)
+            # Refresh the user's last active timestamp asynchronously
+            # to prevent blocking the request
+            if background_tasks:
+                background_tasks.add_task(Users.update_user_last_active_by_id, user.id)
         return user
     else:
         raise HTTPException(

backend/open_webui/utils/chat.py

@@ -7,14 +7,17 @@ from typing import Any, Optional
 import random
 import json
 import inspect
+import uuid
+import asyncio
 
-from fastapi import Request
-from starlette.responses import Response, StreamingResponse
+from fastapi import Request, status
+from starlette.responses import Response, StreamingResponse, JSONResponse
 
 
 from open_webui.models.users import UserModel
 
 from open_webui.socket.main import (
+    sio,
     get_event_call,
     get_event_emitter,
 )
@@ -44,6 +47,10 @@ from open_webui.utils.response import (
     convert_response_ollama_to_openai,
     convert_streaming_response_ollama_to_openai,
 )
+from open_webui.utils.filter import (
+    get_sorted_filter_ids,
+    process_filter_functions,
+)
 
 from open_webui.env import SRC_LOG_LEVELS, GLOBAL_LOG_LEVEL, BYPASS_MODEL_ACCESS_CONTROL
 
@@ -53,108 +60,224 @@ log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["MAIN"])
 
 
+async def generate_direct_chat_completion(
+    request: Request,
+    form_data: dict,
+    user: Any,
+    models: dict,
+):
+    log.info("generate_direct_chat_completion")
+
+    metadata = form_data.pop("metadata", {})
+
+    user_id = metadata.get("user_id")
+    session_id = metadata.get("session_id")
+    request_id = str(uuid.uuid4())  # Generate a unique request ID
+
+    event_caller = get_event_call(metadata)
+
+    channel = f"{user_id}:{session_id}:{request_id}"
+
+    if form_data.get("stream"):
+        q = asyncio.Queue()
+
+        async def message_listener(sid, data):
+            """
+            Handle received socket messages and push them into the queue.
+            """
+            await q.put(data)
+
+        # Register the listener
+        sio.on(channel, message_listener)
+
+        # Start processing chat completion in background
+        res = await event_caller(
+            {
+                "type": "request:chat:completion",
+                "data": {
+                    "form_data": form_data,
+                    "model": models[form_data["model"]],
+                    "channel": channel,
+                    "session_id": session_id,
+                },
+            }
+        )
+
+        log.info(f"res: {res}")
+
+        if res.get("status", False):
+            # Define a generator to stream responses
+            async def event_generator():
+                nonlocal q
+                try:
+                    while True:
+                        data = await q.get()  # Wait for new messages
+                        if isinstance(data, dict):
+                            if "done" in data and data["done"]:
+                                break  # Stop streaming when 'done' is received
+
+                            yield f"data: {json.dumps(data)}\n\n"
+                        elif isinstance(data, str):
+                            yield data
+                except Exception as e:
+                    log.debug(f"Error in event generator: {e}")
+                    pass
+
+            # Define a background task to run the event generator
+            async def background():
+                try:
+                    del sio.handlers["/"][channel]
+                except Exception as e:
+                    pass
+
+            # Return the streaming response
+            return StreamingResponse(
+                event_generator(), media_type="text/event-stream", background=background
+            )
+        else:
+            raise Exception(str(res))
+    else:
+        res = await event_caller(
+            {
+                "type": "request:chat:completion",
+                "data": {
+                    "form_data": form_data,
+                    "model": models[form_data["model"]],
+                    "channel": channel,
+                    "session_id": session_id,
+                },
+            }
+        )
+
+        if "error" in res and res["error"]:
+            raise Exception(res["error"])
+
+        return res
+
+
 async def generate_chat_completion(
     request: Request,
     form_data: dict,
     user: Any,
     bypass_filter: bool = False,
 ):
+    log.debug(f"generate_chat_completion: {form_data}")
     if BYPASS_MODEL_ACCESS_CONTROL:
         bypass_filter = True
 
-    models = request.app.state.MODELS
+    if hasattr(request.state, "metadata"):
+        if "metadata" not in form_data:
+            form_data["metadata"] = request.state.metadata
+        else:
+            form_data["metadata"] = {
+                **form_data["metadata"],
+                **request.state.metadata,
+            }
+
+    if getattr(request.state, "direct", False) and hasattr(request.state, "model"):
+        models = {
+            request.state.model["id"]: request.state.model,
+        }
+        log.debug(f"direct connection to model: {models}")
+    else:
+        models = request.app.state.MODELS
 
     model_id = form_data["model"]
     if model_id not in models:
         raise Exception("Model not found")
 
-    # Process the form_data through the pipeline
-    try:
-        form_data = process_pipeline_inlet_filter(request, form_data, user, models)
-    except Exception as e:
-        raise e
-
     model = models[model_id]
 
-    # Check if user has access to the model
-    if not bypass_filter and user.role == "user":
-        try:
-            check_model_access(user, model)
-        except Exception as e:
-            raise e
-
-    if model["owned_by"] == "arena":
-        model_ids = model.get("info", {}).get("meta", {}).get("model_ids")
-        filter_mode = model.get("info", {}).get("meta", {}).get("filter_mode")
-        if model_ids and filter_mode == "exclude":
-            model_ids = [
-                model["id"]
-                for model in list(request.app.state.MODELS.values())
-                if model.get("owned_by") != "arena" and model["id"] not in model_ids
-            ]
-
-        selected_model_id = None
-        if isinstance(model_ids, list) and model_ids:
-            selected_model_id = random.choice(model_ids)
-        else:
-            model_ids = [
-                model["id"]
-                for model in list(request.app.state.MODELS.values())
-                if model.get("owned_by") != "arena"
-            ]
-            selected_model_id = random.choice(model_ids)
-
-        form_data["model"] = selected_model_id
-
-        if form_data.get("stream") == True:
-
-            async def stream_wrapper(stream):
-                yield f"data: {json.dumps({'selected_model_id': selected_model_id})}\n\n"
-                async for chunk in stream:
-                    yield chunk
-
-            response = await generate_chat_completion(
-                request, form_data, user, bypass_filter=True
-            )
-            return StreamingResponse(
-                stream_wrapper(response.body_iterator),
-                media_type="text/event-stream",
-                background=response.background,
-            )
-        else:
-            return {
-                **(
-                    await generate_chat_completion(
-                        request, form_data, user, bypass_filter=True
-                    )
-                ),
-                "selected_model_id": selected_model_id,
-            }
-
-    if model.get("pipe"):
-        # Below does not require bypass_filter because this is the only route the uses this function and it is already bypassing the filter
-        return await generate_function_chat_completion(
+    if getattr(request.state, "direct", False):
+        return await generate_direct_chat_completion(
             request, form_data, user=user, models=models
         )
-    if model["owned_by"] == "ollama":
-        # Using /ollama/api/chat endpoint
-        form_data = convert_payload_openai_to_ollama(form_data)
-        response = await generate_ollama_chat_completion(
-            request=request, form_data=form_data, user=user, bypass_filter=bypass_filter
-        )
-        if form_data.get("stream"):
-            response.headers["content-type"] = "text/event-stream"
-            return StreamingResponse(
-                convert_streaming_response_ollama_to_openai(response),
-                headers=dict(response.headers),
-                background=response.background,
-            )
-        else:
-            return convert_response_ollama_to_openai(response)
     else:
-        return await generate_openai_chat_completion(
-            request=request, form_data=form_data, user=user, bypass_filter=bypass_filter
-        )
+        # Check if user has access to the model
+        if not bypass_filter and user.role == "user":
+            try:
+                check_model_access(user, model)
+            except Exception as e:
+                raise e
+
+        if model.get("owned_by") == "arena":
+            model_ids = model.get("info", {}).get("meta", {}).get("model_ids")
+            filter_mode = model.get("info", {}).get("meta", {}).get("filter_mode")
+            if model_ids and filter_mode == "exclude":
+                model_ids = [
+                    model["id"]
+                    for model in list(request.app.state.MODELS.values())
+                    if model.get("owned_by") != "arena" and model["id"] not in model_ids
+                ]
+
+            selected_model_id = None
+            if isinstance(model_ids, list) and model_ids:
+                selected_model_id = random.choice(model_ids)
+            else:
+                model_ids = [
+                    model["id"]
+                    for model in list(request.app.state.MODELS.values())
+                    if model.get("owned_by") != "arena"
+                ]
+                selected_model_id = random.choice(model_ids)
+
+            form_data["model"] = selected_model_id
+
+            if form_data.get("stream") == True:
+
+                async def stream_wrapper(stream):
+                    yield f"data: {json.dumps({'selected_model_id': selected_model_id})}\n\n"
+                    async for chunk in stream:
+                        yield chunk
+
+                response = await generate_chat_completion(
+                    request, form_data, user, bypass_filter=True
+                )
+                return StreamingResponse(
+                    stream_wrapper(response.body_iterator),
+                    media_type="text/event-stream",
+                    background=response.background,
+                )
+            else:
+                return {
+                    **(
+                        await generate_chat_completion(
+                            request, form_data, user, bypass_filter=True
+                        )
+                    ),
+                    "selected_model_id": selected_model_id,
+                }
+
+        if model.get("pipe"):
+            # Below does not require bypass_filter because this is the only route the uses this function and it is already bypassing the filter
+            return await generate_function_chat_completion(
+                request, form_data, user=user, models=models
+            )
+        if model.get("owned_by") == "ollama":
+            # Using /ollama/api/chat endpoint
+            form_data = convert_payload_openai_to_ollama(form_data)
+            response = await generate_ollama_chat_completion(
+                request=request,
+                form_data=form_data,
+                user=user,
+                bypass_filter=bypass_filter,
+            )
+            if form_data.get("stream"):
+                response.headers["content-type"] = "text/event-stream"
+                return StreamingResponse(
+                    convert_streaming_response_ollama_to_openai(response),
+                    headers=dict(response.headers),
+                    background=response.background,
+                )
+            else:
+                return convert_response_ollama_to_openai(response)
+        else:
+            return await generate_openai_chat_completion(
+                request=request,
+                form_data=form_data,
+                user=user,
+                bypass_filter=bypass_filter,
+            )
 
 
 chat_completion = generate_chat_completion
@@ -162,8 +285,14 @@ chat_completion = generate_chat_completion
 
 async def chat_completed(request: Request, form_data: dict, user: Any):
     if not request.app.state.MODELS:
-        await get_all_models(request)
-    models = request.app.state.MODELS
+        await get_all_models(request, user=user)
+
+    if getattr(request.state, "direct", False) and hasattr(request.state, "model"):
+        models = {
+            request.state.model["id"]: request.state.model,
+        }
+    else:
+        models = request.app.state.MODELS
 
     data = form_data
     model_id = data["model"]
@@ -173,120 +302,47 @@ async def chat_completed(request: Request, form_data: dict, user: Any):
     model = models[model_id]
 
     try:
-        data = process_pipeline_outlet_filter(request, data, user, models)
+        data = await process_pipeline_outlet_filter(request, data, user, models)
     except Exception as e:
         return Exception(f"Error: {e}")
 
-    __event_emitter__ = get_event_emitter(
-        {
-            "chat_id": data["chat_id"],
-            "message_id": data["id"],
-            "session_id": data["session_id"],
-            "user_id": user.id,
-        }
-    )
+    metadata = {
+        "chat_id": data["chat_id"],
+        "message_id": data["id"],
+        "session_id": data["session_id"],
+        "user_id": user.id,
+    }
 
-    __event_call__ = get_event_call(
-        {
-            "chat_id": data["chat_id"],
-            "message_id": data["id"],
-            "session_id": data["session_id"],
-            "user_id": user.id,
-        }
-    )
+    extra_params = {
+        "__event_emitter__": get_event_emitter(metadata),
+        "__event_call__": get_event_call(metadata),
+        "__user__": {
+            "id": user.id,
+            "email": user.email,
+            "name": user.name,
+            "role": user.role,
+        },
+        "__metadata__": metadata,
+        "__request__": request,
+        "__model__": model,
+    }
 
-    def get_priority(function_id):
-        function = Functions.get_function_by_id(function_id)
-        if function is not None and hasattr(function, "valves"):
-            # TODO: Fix FunctionModel to include vavles
-            return (function.valves if function.valves else {}).get("priority", 0)
-        return 0
+    try:
+        filter_functions = [
+            Functions.get_function_by_id(filter_id)
+            for filter_id in get_sorted_filter_ids(model)
+        ]
 
-    filter_ids = [function.id for function in Functions.get_global_filter_functions()]
-    if "info" in model and "meta" in model["info"]:
-        filter_ids.extend(model["info"]["meta"].get("filterIds", []))
-        filter_ids = list(set(filter_ids))
-
-    enabled_filter_ids = [
-        function.id
-        for function in Functions.get_functions_by_type("filter", active_only=True)
-    ]
-    filter_ids = [
-        filter_id for filter_id in filter_ids if filter_id in enabled_filter_ids
-    ]
-
-    # Sort filter_ids by priority, using the get_priority function
-    filter_ids.sort(key=get_priority)
-
-    for filter_id in filter_ids:
-        filter = Functions.get_function_by_id(filter_id)
-        if not filter:
-            continue
-
-        if filter_id in request.app.state.FUNCTIONS:
-            function_module = request.app.state.FUNCTIONS[filter_id]
-        else:
-            function_module, _, _ = load_function_module_by_id(filter_id)
-            request.app.state.FUNCTIONS[filter_id] = function_module
-
-        if hasattr(function_module, "valves") and hasattr(function_module, "Valves"):
-            valves = Functions.get_function_valves_by_id(filter_id)
-            function_module.valves = function_module.Valves(
-                **(valves if valves else {})
-            )
-
-        if not hasattr(function_module, "outlet"):
-            continue
-        try:
-            outlet = function_module.outlet
-
-            # Get the signature of the function
-            sig = inspect.signature(outlet)
-            params = {"body": data}
-
-            # Extra parameters to be passed to the function
-            extra_params = {
-                "__model__": model,
-                "__id__": filter_id,
-                "__event_emitter__": __event_emitter__,
-                "__event_call__": __event_call__,
-                "__request__": request,
-            }
-
-            # Add extra params in contained in function signature
-            for key, value in extra_params.items():
-                if key in sig.parameters:
-                    params[key] = value
-
-            if "__user__" in sig.parameters:
-                __user__ = {
-                    "id": user.id,
-                    "email": user.email,
-                    "name": user.name,
-                    "role": user.role,
-                }
-
-                try:
-                    if hasattr(function_module, "UserValves"):
-                        __user__["valves"] = function_module.UserValves(
-                            **Functions.get_user_valves_by_id_and_user_id(
-                                filter_id, user.id
-                            )
-                        )
-                except Exception as e:
-                    print(e)
-
-                params = {**params, "__user__": __user__}
-
-            if inspect.iscoroutinefunction(outlet):
-                data = await outlet(**params)
-            else:
-                data = outlet(**params)
-
-        except Exception as e:
-            return Exception(f"Error: {e}")
-
-    return data
+        result, _ = await process_filter_functions(
+            request=request,
+            filter_functions=filter_functions,
+            filter_type="outlet",
+            form_data=data,
+            extra_params=extra_params,
+        )
+        return result
+    except Exception as e:
+        return Exception(f"Error: {e}")
 
 
 async def chat_action(request: Request, action_id: str, form_data: dict, user: Any):
@@ -300,8 +356,14 @@ async def chat_action(request: Request, action_id: str, form_data: dict, user: A
         raise Exception(f"Action not found: {action_id}")
 
     if not request.app.state.MODELS:
-        await get_all_models(request)
-    models = request.app.state.MODELS
+        await get_all_models(request, user=user)
+
+    if getattr(request.state, "direct", False) and hasattr(request.state, "model"):
+        models = {
+            request.state.model["id"]: request.state.model,
+        }
+    else:
+        models = request.app.state.MODELS
 
     data = form_data
     model_id = data["model"]
@@ -375,7 +437,7 @@ async def chat_action(request: Request, action_id: str, form_data: dict, user: A
                             )
                         )
                 except Exception as e:
-                    print(e)
+                    log.exception(f"Failed to get user values: {e}")
 
                 params = {**params, "__user__": __user__}
 

backend/open_webui/utils/code_interpreter.py

@@ -0,0 +1,210 @@
+import asyncio
+import json
+import logging
+import uuid
+from typing import Optional
+
+import aiohttp
+import websockets
+from pydantic import BaseModel
+
+from open_webui.env import SRC_LOG_LEVELS
+
+logger = logging.getLogger(__name__)
+logger.setLevel(SRC_LOG_LEVELS["MAIN"])
+
+
+class ResultModel(BaseModel):
+    """
+    Execute Code Result Model
+    """
+
+    stdout: Optional[str] = ""
+    stderr: Optional[str] = ""
+    result: Optional[str] = ""
+
+
+class JupyterCodeExecuter:
+    """
+    Execute code in jupyter notebook
+    """
+
+    def __init__(
+        self,
+        base_url: str,
+        code: str,
+        token: str = "",
+        password: str = "",
+        timeout: int = 60,
+    ):
+        """
+        :param base_url: Jupyter server URL (e.g., "http://localhost:8888")
+        :param code: Code to execute
+        :param token: Jupyter authentication token (optional)
+        :param password: Jupyter password (optional)
+        :param timeout: WebSocket timeout in seconds (default: 60s)
+        """
+        self.base_url = base_url.rstrip("/")
+        self.code = code
+        self.token = token
+        self.password = password
+        self.timeout = timeout
+        self.kernel_id = ""
+        self.session = aiohttp.ClientSession(base_url=self.base_url)
+        self.params = {}
+        self.result = ResultModel()
+
+    async def __aenter__(self):
+        return self
+
+    async def __aexit__(self, exc_type, exc_val, exc_tb):
+        if self.kernel_id:
+            try:
+                async with self.session.delete(
+                    f"/api/kernels/{self.kernel_id}", params=self.params
+                ) as response:
+                    response.raise_for_status()
+            except Exception as err:
+                logger.exception("close kernel failed, %s", err)
+        await self.session.close()
+
+    async def run(self) -> ResultModel:
+        try:
+            await self.sign_in()
+            await self.init_kernel()
+            await self.execute_code()
+        except Exception as err:
+            logger.exception("execute code failed, %s", err)
+            self.result.stderr = f"Error: {err}"
+        return self.result
+
+    async def sign_in(self) -> None:
+        # password authentication
+        if self.password and not self.token:
+            async with self.session.get("/login") as response:
+                response.raise_for_status()
+                xsrf_token = response.cookies["_xsrf"].value
+                if not xsrf_token:
+                    raise ValueError("_xsrf token not found")
+                self.session.cookie_jar.update_cookies(response.cookies)
+                self.session.headers.update({"X-XSRFToken": xsrf_token})
+            async with self.session.post(
+                "/login",
+                data={"_xsrf": xsrf_token, "password": self.password},
+                allow_redirects=False,
+            ) as response:
+                response.raise_for_status()
+                self.session.cookie_jar.update_cookies(response.cookies)
+
+        # token authentication
+        if self.token:
+            self.params.update({"token": self.token})
+
+    async def init_kernel(self) -> None:
+        async with self.session.post(
+            url="/api/kernels", params=self.params
+        ) as response:
+            response.raise_for_status()
+            kernel_data = await response.json()
+            self.kernel_id = kernel_data["id"]
+
+    def init_ws(self) -> (str, dict):
+        ws_base = self.base_url.replace("http", "ws")
+        ws_params = "?" + "&".join([f"{key}={val}" for key, val in self.params.items()])
+        websocket_url = f"{ws_base}/api/kernels/{self.kernel_id}/channels{ws_params if len(ws_params) > 1 else ''}"
+        ws_headers = {}
+        if self.password and not self.token:
+            ws_headers = {
+                "Cookie": "; ".join(
+                    [
+                        f"{cookie.key}={cookie.value}"
+                        for cookie in self.session.cookie_jar
+                    ]
+                ),
+                **self.session.headers,
+            }
+        return websocket_url, ws_headers
+
+    async def execute_code(self) -> None:
+        # initialize ws
+        websocket_url, ws_headers = self.init_ws()
+        # execute
+        async with websockets.connect(
+            websocket_url, additional_headers=ws_headers
+        ) as ws:
+            await self.execute_in_jupyter(ws)
+
+    async def execute_in_jupyter(self, ws) -> None:
+        # send message
+        msg_id = uuid.uuid4().hex
+        await ws.send(
+            json.dumps(
+                {
+                    "header": {
+                        "msg_id": msg_id,
+                        "msg_type": "execute_request",
+                        "username": "user",
+                        "session": uuid.uuid4().hex,
+                        "date": "",
+                        "version": "5.3",
+                    },
+                    "parent_header": {},
+                    "metadata": {},
+                    "content": {
+                        "code": self.code,
+                        "silent": False,
+                        "store_history": True,
+                        "user_expressions": {},
+                        "allow_stdin": False,
+                        "stop_on_error": True,
+                    },
+                    "channel": "shell",
+                }
+            )
+        )
+        # parse message
+        stdout, stderr, result = "", "", []
+        while True:
+            try:
+                # wait for message
+                message = await asyncio.wait_for(ws.recv(), self.timeout)
+                message_data = json.loads(message)
+                # msg id not match, skip
+                if message_data.get("parent_header", {}).get("msg_id") != msg_id:
+                    continue
+                # check message type
+                msg_type = message_data.get("msg_type")
+                match msg_type:
+                    case "stream":
+                        if message_data["content"]["name"] == "stdout":
+                            stdout += message_data["content"]["text"]
+                        elif message_data["content"]["name"] == "stderr":
+                            stderr += message_data["content"]["text"]
+                    case "execute_result" | "display_data":
+                        data = message_data["content"]["data"]
+                        if "image/png" in data:
+                            result.append(f"data:image/png;base64,{data['image/png']}")
+                        elif "text/plain" in data:
+                            result.append(data["text/plain"])
+                    case "error":
+                        stderr += "\n".join(message_data["content"]["traceback"])
+                    case "status":
+                        if message_data["content"]["execution_state"] == "idle":
+                            break
+
+            except asyncio.TimeoutError:
+                stderr += "\nExecution timed out."
+                break
+        self.result.stdout = stdout.strip()
+        self.result.stderr = stderr.strip()
+        self.result.result = "\n".join(result).strip() if result else ""
+
+
+async def execute_code_jupyter(
+    base_url: str, code: str, token: str = "", password: str = "", timeout: int = 60
+) -> dict:
+    async with JupyterCodeExecuter(
+        base_url, code, token, password, timeout
+    ) as executor:
+        result = await executor.run()
+        return result.model_dump()

backend/open_webui/utils/filter.py

@@ -0,0 +1,112 @@
+import inspect
+import logging
+
+from open_webui.utils.plugin import load_function_module_by_id
+from open_webui.models.functions import Functions
+from open_webui.env import SRC_LOG_LEVELS
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["MAIN"])
+
+
+def get_sorted_filter_ids(model: dict):
+    def get_priority(function_id):
+        function = Functions.get_function_by_id(function_id)
+        if function is not None:
+            valves = Functions.get_function_valves_by_id(function_id)
+            return valves.get("priority", 0) if valves else 0
+        return 0
+
+    filter_ids = [function.id for function in Functions.get_global_filter_functions()]
+    if "info" in model and "meta" in model["info"]:
+        filter_ids.extend(model["info"]["meta"].get("filterIds", []))
+        filter_ids = list(set(filter_ids))
+
+    enabled_filter_ids = [
+        function.id
+        for function in Functions.get_functions_by_type("filter", active_only=True)
+    ]
+
+    filter_ids = [fid for fid in filter_ids if fid in enabled_filter_ids]
+    filter_ids.sort(key=get_priority)
+    return filter_ids
+
+
+async def process_filter_functions(
+    request, filter_functions, filter_type, form_data, extra_params
+):
+    skip_files = None
+
+    for function in filter_functions:
+        filter = function
+        filter_id = function.id
+        if not filter:
+            continue
+
+        if filter_id in request.app.state.FUNCTIONS:
+            function_module = request.app.state.FUNCTIONS[filter_id]
+        else:
+            function_module, _, _ = load_function_module_by_id(filter_id)
+            request.app.state.FUNCTIONS[filter_id] = function_module
+
+        # Prepare handler function
+        handler = getattr(function_module, filter_type, None)
+        if not handler:
+            continue
+
+        # Check if the function has a file_handler variable
+        if filter_type == "inlet" and hasattr(function_module, "file_handler"):
+            skip_files = function_module.file_handler
+
+        # Apply valves to the function
+        if hasattr(function_module, "valves") and hasattr(function_module, "Valves"):
+            valves = Functions.get_function_valves_by_id(filter_id)
+            function_module.valves = function_module.Valves(
+                **(valves if valves else {})
+            )
+
+        try:
+            # Prepare parameters
+            sig = inspect.signature(handler)
+
+            params = {"body": form_data}
+            if filter_type == "stream":
+                params = {"event": form_data}
+
+            params = params | {
+                k: v
+                for k, v in {
+                    **extra_params,
+                    "__id__": filter_id,
+                }.items()
+                if k in sig.parameters
+            }
+
+            # Handle user parameters
+            if "__user__" in sig.parameters:
+                if hasattr(function_module, "UserValves"):
+                    try:
+                        params["__user__"]["valves"] = function_module.UserValves(
+                            **Functions.get_user_valves_by_id_and_user_id(
+                                filter_id, params["__user__"]["id"]
+                            )
+                        )
+                    except Exception as e:
+                        log.exception(f"Failed to get user values: {e}")
+
+            # Execute handler
+            if inspect.iscoroutinefunction(handler):
+                form_data = await handler(**params)
+            else:
+                form_data = handler(**params)
+
+        except Exception as e:
+            log.debug(f"Error in {filter_type} handler {filter_id}: {e}")
+            raise e
+
+    # Handle file cleanup for inlet
+    if skip_files and "files" in form_data.get("metadata", {}):
+        del form_data["files"]
+        del form_data["metadata"]["files"]
+
+    return form_data, {}

backend/open_webui/utils/images/comfyui.py

@@ -161,7 +161,7 @@ async def comfyui_generate_image(
                 seed = (
                     payload.seed
                     if payload.seed
-                    else random.randint(0, 18446744073709551614)
+                    else random.randint(0, 1125899906842624)
                 )
                 for node_id in node.node_ids:
                     workflow[node_id]["inputs"][node.key] = seed

backend/open_webui/utils/logger.py

@@ -0,0 +1,140 @@
+import json
+import logging
+import sys
+from typing import TYPE_CHECKING
+
+from loguru import logger
+
+from open_webui.env import (
+    AUDIT_LOG_FILE_ROTATION_SIZE,
+    AUDIT_LOG_LEVEL,
+    AUDIT_LOGS_FILE_PATH,
+    GLOBAL_LOG_LEVEL,
+)
+
+
+if TYPE_CHECKING:
+    from loguru import Record
+
+
+def stdout_format(record: "Record") -> str:
+    """
+    Generates a formatted string for log records that are output to the console. This format includes a timestamp, log level, source location (module, function, and line), the log message, and any extra data (serialized as JSON).
+
+    Parameters:
+    record (Record): A Loguru record that contains logging details including time, level, name, function, line, message, and any extra context.
+    Returns:
+    str: A formatted log string intended for stdout.
+    """
+    record["extra"]["extra_json"] = json.dumps(record["extra"])
+    return (
+        "<green>{time:YYYY-MM-DD HH:mm:ss.SSS}</green> | "
+        "<level>{level: <8}</level> | "
+        "<cyan>{name}</cyan>:<cyan>{function}</cyan>:<cyan>{line}</cyan> - "
+        "<level>{message}</level> - {extra[extra_json]}"
+        "\n{exception}"
+    )
+
+
+class InterceptHandler(logging.Handler):
+    """
+    Intercepts log records from Python's standard logging module
+    and redirects them to Loguru's logger.
+    """
+
+    def emit(self, record):
+        """
+        Called by the standard logging module for each log event.
+        It transforms the standard `LogRecord` into a format compatible with Loguru
+        and passes it to Loguru's logger.
+        """
+        try:
+            level = logger.level(record.levelname).name
+        except ValueError:
+            level = record.levelno
+
+        frame, depth = sys._getframe(6), 6
+        while frame and frame.f_code.co_filename == logging.__file__:
+            frame = frame.f_back
+            depth += 1
+
+        logger.opt(depth=depth, exception=record.exc_info).log(
+            level, record.getMessage()
+        )
+
+
+def file_format(record: "Record"):
+    """
+    Formats audit log records into a structured JSON string for file output.
+
+    Parameters:
+    record (Record): A Loguru record containing extra audit data.
+    Returns:
+    str: A JSON-formatted string representing the audit data.
+    """
+
+    audit_data = {
+        "id": record["extra"].get("id", ""),
+        "timestamp": int(record["time"].timestamp()),
+        "user": record["extra"].get("user", dict()),
+        "audit_level": record["extra"].get("audit_level", ""),
+        "verb": record["extra"].get("verb", ""),
+        "request_uri": record["extra"].get("request_uri", ""),
+        "response_status_code": record["extra"].get("response_status_code", 0),
+        "source_ip": record["extra"].get("source_ip", ""),
+        "user_agent": record["extra"].get("user_agent", ""),
+        "request_object": record["extra"].get("request_object", b""),
+        "response_object": record["extra"].get("response_object", b""),
+        "extra": record["extra"].get("extra", {}),
+    }
+
+    record["extra"]["file_extra"] = json.dumps(audit_data, default=str)
+    return "{extra[file_extra]}\n"
+
+
+def start_logger():
+    """
+    Initializes and configures Loguru's logger with distinct handlers:
+
+    A console (stdout) handler for general log messages (excluding those marked as auditable).
+    An optional file handler for audit logs if audit logging is enabled.
+    Additionally, this function reconfigures Python’s standard logging to route through Loguru and adjusts logging levels for Uvicorn.
+
+    Parameters:
+    enable_audit_logging (bool): Determines whether audit-specific log entries should be recorded to file.
+    """
+    logger.remove()
+
+    logger.add(
+        sys.stdout,
+        level=GLOBAL_LOG_LEVEL,
+        format=stdout_format,
+        filter=lambda record: "auditable" not in record["extra"],
+    )
+
+    if AUDIT_LOG_LEVEL != "NONE":
+        try:
+            logger.add(
+                AUDIT_LOGS_FILE_PATH,
+                level="INFO",
+                rotation=AUDIT_LOG_FILE_ROTATION_SIZE,
+                compression="zip",
+                format=file_format,
+                filter=lambda record: record["extra"].get("auditable") is True,
+            )
+        except Exception as e:
+            logger.error(f"Failed to initialize audit log file handler: {str(e)}")
+
+    logging.basicConfig(
+        handlers=[InterceptHandler()], level=GLOBAL_LOG_LEVEL, force=True
+    )
+    for uvicorn_logger_name in ["uvicorn", "uvicorn.error"]:
+        uvicorn_logger = logging.getLogger(uvicorn_logger_name)
+        uvicorn_logger.setLevel(GLOBAL_LOG_LEVEL)
+        uvicorn_logger.handlers = []
+    for uvicorn_logger_name in ["uvicorn.access"]:
+        uvicorn_logger = logging.getLogger(uvicorn_logger_name)
+        uvicorn_logger.setLevel(GLOBAL_LOG_LEVEL)
+        uvicorn_logger.handlers = [InterceptHandler()]
+
+    logger.info(f"GLOBAL_LOG_LEVEL: {GLOBAL_LOG_LEVEL}")