Merge pull request #13530 from open-webui/dev

0.6.7

.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
 
@@ -70,3 +70,7 @@
 ### Screenshots or Videos
 
 - [Attach any relevant screenshots or videos demonstrating the changes]
+
+### Contributor License Agreement
+
+By submitting this pull request, I confirm that I have read and fully agree to the [Contributor License Agreement (CLA)](/CONTRIBUTOR_LICENSE_AGREEMENT), and I am providing my contributions under its terms.

.github/workflows/codespell.disabled

@@ -0,0 +1,25 @@
+# Codespell configuration is within pyproject.toml
+---
+name: Codespell
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+permissions:
+  contents: read
+
+jobs:
+  codespell:
+    name: Check for spelling errors
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Annotate locations with typos
+        uses: codespell-project/codespell-problem-matcher@v1
+      - name: Codespell
+        uses: codespell-project/actions-codespell@v2

.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/integration-test.disabled

@@ -52,6 +52,8 @@ jobs:
 
       - name: Cypress run
         uses: cypress-io/github-action@v6
+        env:
+          LIBGL_ALWAYS_SOFTWARE: 1
         with:
           browser: chrome
           wait-on: 'http://localhost:3000'

.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 +5,572 @@ 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).
 
+## [0.6.7] - 2025-05-07
+
+### Added
+
+- 🌐 **Custom Azure TTS API URL Support Added**: You can now define a custom Azure Text-to-Speech endpoint—enabling flexibility for enterprise deployments and regional compliance.
+- ⚙️ **TOOL_SERVER_CONNECTIONS Environment Variable Suppor**: Easily configure and deploy tool servers via environment variables, streamlining setup and enabling faster enterprise provisioning.
+- 👥 **Enhanced OAuth Group Handling as String or List**: OAuth group data can now be passed as either a list or a comma-separated string, improving compatibility with varied identity provider formats and reducing onboarding friction.
+
+### Fixed
+
+- 🧠 **Embedding with Ollama Proxy Endpoints Restored**: Fixed an issue where missing API config broke embedding for proxied Ollama models—ensuring consistent performance and compatibility.
+- 🔐 **OIDC OAuth Login Issue Resolved**: Users can once again sign in seamlessly using OpenID Connect-based OAuth, eliminating login interruptions and improving reliability.
+- 📝 **Notes Feature Access Fixed for Non-Admins**: Fixed an issue preventing non-admin users from accessing the Notes feature, restoring full cross-role collaboration capabilities.
+- 🖼️ **Tika Loader Image Extraction Problem Resolved**: Ensured TikaLoader now processes 'extract_images' parameter correctly, restoring complete file extraction functionality in document workflows.
+- 🎨 **Automatic1111 Image Model Setting Applied Properly**: Fixed an issue where switching to a specific image model via the UI wasn’t reflected in generation, re-enabling full visual creativity control.
+- 🏷️ **Multiple XML Tags in Messages Now Parsed Correctly**: Fixed parsing issues when messages included multiple XML-style tags, ensuring clean and unbroken rendering of rich content in chats.
+- 🖌️ **OpenAI Image Generation Issues Resolved**: Resolved broken image output when using OpenAI’s image generation, ensuring fully functional visual creation workflows.
+- 🔎 **Tool Server Settings UI Privacy Restored**: Prevented restricted users from accessing tool server settings via search—restoring tight permissions control and safeguarding sensitive configurations.
+- 🎧 **WebM Audio Transcription Now Supported**: Fixed an issue where WebM files failed during audio transcription—these formats are now fully supported, ensuring smoother voice note workflows and broader file compatibility.
+
+## [0.6.6] - 2025-05-05
+
+### Added
+
+- 📝 **AI-Enhanced Notes (With Audio Transcription)**: Effortlessly create notes, attach meeting or voice audio, and let the AI instantly enhance, summarize, or refine your notes using audio transcriptions—making your documentation smarter, cleaner, and more insightful with minimal effort.
+- 🔊 **Meeting Audio Recording & Import**: Seamlessly record audio from your meetings or capture screen audio and attach it to your notes—making it easier to revisit, annotate, and extract insights from important discussions.
+- 📁 **Import Markdown Notes Effortlessly**: Bring your existing knowledge library into Open WebUI by importing your Markdown notes, so you can leverage all advanced note management and AI features right away.
+- 👥 **Notes Permissions by User Group**: Fine-tune access and editing rights for notes based on user roles or groups, so you can delegate writing or restrict sensitive information as needed.
+- ☁️ **OneDrive & SharePoint Integration**: Keep your content in sync by connecting notes and files directly with OneDrive or SharePoint—unlocking fast enterprise import/export and seamless collaboration with your existing workflows.
+- 🗂️ **Paginated User List in Admin Panel**: Effortlessly manage and search through large teams via the new paginated user list—saving time and streamlining user administration in big organizations.
+- 🕹️ **Granular Chat Share & Export Permissions**: Enjoy enhanced control over who can share or export chats, enabling tighter governance and privacy in team and enterprise settings.
+- 🛑 **User Role Change Confirmation Dialog**: Reduce accidental privilege changes with a required confirmation step before updating user roles—improving security and preventing costly mistakes in team management.
+- 🚨 **Audit Log for Failed Login Attempts**: Quickly detect unauthorized access attempts or troubleshoot user login problems with detailed logs of failed authentication right in the audit trail.
+- 💡 **Dedicated 'Generate Title' Button for Chats**: Swiftly organize every conversation—tap the new button to let AI create relevant, clear titles for all your chats, saving time and reducing clutter.
+- 💬 **Notification Sound Always-On Option**: Take control of your notifications by setting sound alerts to always play—helping you stay on top of important updates in busy environments.
+- 🆔 **S3 File Tagging Support**: Uploaded files to S3 now include tags for better organization, searching, and integration with your file management policies.
+- 🛡️ **OAuth Blocked Groups Support**: Gain more control over group-based access by explicitly blocking specified OAuth groups—ideal for complex identity or security requirements.
+- 🚀 **Optimized Faster Web Search & Multi-Threaded Queries**: Enjoy dramatically faster web search and RAG (retrieval augmented generation) with revamped multi-threaded search—get richer, more accurate results in less time.
+- 🔍 **All-Knowledge Parallel Search**: Searches across your entire knowledge base now happen in parallel even in non-hybrid mode, speeding up responses and improving knowledge accuracy for every question.
+- 🌐 **New Firecrawl & Yacy Web Search Integrations**: Expand your world of information with two new advanced search engines—Firecrawl for deeper web insight and Yacy for decentralized, privacy-friendly search capabilities.
+- 🧠 **Configurable Docling OCR Engine & Language**: Use environment variables to fine-tune Docling OCR engine and supported languages for smarter, more tailored document extraction and RAG workflows.
+- 🗝️ **Enhanced Sentence Transformers Configuration**: Added new environment variables for easier set up and advanced customization of Sentence Transformers—ensuring best fit for your embedding needs.
+- 🌲 **Pinecone Vector Database Integration**: Index, search, and manage knowledge at enterprise scale with full native support for Pinecone as your vector database—effortlessly handle even the biggest document sets.
+- 🔄 **Automatic Requirements Installation for Tools & Functions**: Never worry about lost dependencies on restart—external function and tool requirements are now auto-installed at boot, ensuring tools always “just work.”
+- 🔒 **Automatic Sign-Out on Token Expiry**: Security is smarter—users are now automatically logged out if their authentication token expires, protecting sensitive content and ensuring compliance without disruption.
+- 🎬 **Automatic YouTube Embed Detection**: Paste YouTube links and see instant in-chat video embeds—no more manual embedding, making knowledge sharing and media consumption even easier for every team.
+- 🔄 **Expanded Language & Locale Support**: Translations for Danish, French, Russian, Traditional Chinese, Simplified Chinese, Thai, Catalan, German, and Korean have been upgraded, offering smoother, more natural user experiences across the platform.
+
+### Fixed
+
+- 🔒 **Tighter HTML Token Security**: HTML rendering is now restricted to admin-uploaded tokens only, reducing any risk of XSS and keeping your data safe.
+- 🔐 **Refined HTML Security and Token Handling**: Further hardened how HTML tokens and content are handled, guaranteeing even stronger resistance to security vulnerabilities and attacks.
+- 🔏 **Correct Model Usage with Ollama Proxy Prefixes**: Enhanced model reference handling so proxied models in Ollama always download and run correctly—even when using custom prefixes.
+- 📥 **Video File Upload Handling**: Prevented video files from being misclassified as text, fixing bugs with uploads and ensuring media files work as expected.
+- 🔄 **No More Dependent WebSocket Sequential Delays**: Streamlined WebSocket operation to prevent delays and maintain snappy real-time collaboration, especially in multi-user environments.
+- 🛠️ **More Robust Action Module Execution**: Multiple actions in a module now trigger as designed, increasing automation and scripting flexibility.
+- 📧 **Notification Webhooks**: Ensured that notification webhooks are always sent for user events, even when the user isn’t currently active.
+- 🗂️ **Smarter Knowledge Base Reindexing**: Knowledge reindexing continues even when corrupt or missing collections are encountered, keeping your search features running reliably.
+- 🏷️ **User Import with Profile Images**: When importing users, their profile images now come along—making onboarding and collaboration visually clearer from day one.
+- 💬 **OpenAI o-Series Universal Support**: All OpenAI o-series models are now seamlessly recognized and supported, unlocking more advanced capabilities and model choices for every workflow.
+
+### Changed
+
+- 📜 **Custom License Update & Contributor Agreement**: Open WebUI now operates under a custom license with Contributor License Agreement required by default—see https://docs.openwebui.com/license/ for details, ensuring sustainable open innovation for the community.
+- 🔨 **CUDA Docker Images Updated to 12.8**: Upgraded CUDA image support for faster, more compatible model inference and futureproof GPU performance in your AI infrastructure.
+- 🧱 **General Backend Refactoring for Reliability**: Continuous stability improvements streamline backend logic, reduce errors, and lay a stronger foundation for the next wave of feature releases—all under the hood for a more dependable WebUI.
+
+## [0.6.5] - 2025-04-14
+
+### Added
+
+- 🛂 **Granular Voice Feature Permissions Per User Group**: Admins can now separately manage access to Speech-to-Text (record voice), Text-to-Speech (read aloud), and Tool Calls for each user group—giving teams tighter control over voice features and enhanced governance across roles.
+- 🗣️ **Toggle Voice Activity Detection (VAD) for Whisper STT**: New environment variable lets you enable/disable VAD filtering with built-in Whisper speech-to-text, giving you flexibility to optimize for different audio quality and response accuracy levels.
+- 📋 **Copy Formatted Response Mode**: You can now enable “Copy Formatted” in Settings > Interface to copy AI responses exactly as styled (with rich formatting, links, and structure preserved), making it faster and cleaner to paste into documents, emails, or reports.
+- ⚙️ **Backend Stability and Performance Enhancements**: General backend refactoring improves system resilience, consistency, and overall reliability—offering smoother performance across workflows whether chatting, generating media, or using external tools.
+- 🌎 **Translation Refinements Across Multiple Languages**: Updated translations deliver smoother language localization, clearer labels, and improved international usability throughout the UI—ensuring a better experience for non-English speakers.
+
+### Fixed
+
+- 🛠️ **LDAP Login Reliability Restored**: Resolved a critical issue where some LDAP setups failed due to attribute parsing—ensuring consistent, secure, and seamless user authentication across enterprise deployments.
+- 🖼️ **Image Generation in Temporary Chats Now Works Properly**: Fixed a bug where image outputs weren’t generated during temporary chats—visual content can now be used reliably in all chat modes without interruptions.
+
+## [0.6.4] - 2025-04-12
+
+### Fixed
+
+- 🛠️ **RAG_TEMPLATE Display Issue Resolved**: Fixed a formatting problem where the custom RAG_TEMPLATE wasn't correctly rendered in the interface—ensuring that custom retrieval prompts now appear exactly as intended for more reliable prompt engineering.
+
+## [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
+
+- **🌍 Enhanced Internationalization (i18n)**: Refined and expanded translations for greater global accessibility and a smoother experience for international users.
+
+### Fixed
+
+- **🔗 Connection Model ID Resolution**: Resolved an issue preventing model IDs from registering in connections.
+- **💡 Prefix ID for Ollama Connections**: Fixed a bug where prefix IDs in Ollama connections were non-functional.
+- **🔧 Ollama Model Enable/Disable Functionality**: Addressed the issue of enable/disable toggles not working for Ollama base models.
+- **🔒 RBAC Permissions for Tools and Models**: Corrected incorrect Role-Based Access Control (RBAC) permissions for tools and models, ensuring that users now only access features according to their assigned privileges, enhancing security and role clarity.
+
+## [0.5.6] - 2025-01-22
+
+### Added
+
+- **🧠 Effortful Reasoning Control for OpenAI Models**: Introduced the reasoning_effort parameter in chat controls for supported OpenAI models, enabling users to fine-tune how much cognitive effort a model dedicates to its responses, offering greater customization for complex queries and reasoning tasks.
+
+### Fixed
+
+- **🔄 Chat Controls Loading UI Bug**: Resolved an issue where collapsible chat controls appeared as "loading," ensuring a smoother and more intuitive user experience for managing chat settings.
+
+### Changed
+
+- **🔧 Updated Ollama Model Creation**: Revamped the Ollama model creation method to align with their new JSON payload format, ensuring seamless compatibility and more efficient model setup workflows.
+
+## [0.5.5] - 2025-01-22
+
+### Added
+
+- **🤔 Native 'Think' Tag Support**: Introduced the new 'think' tag support that visually displays how long the model is thinking, omitting the reasoning content itself until the next turn. Ideal for creating a more streamlined and focused interaction experience.
+- **🖼️ Toggle Image Generation On/Off**: In the chat input menu, you can now easily toggle image generation before initiating chats, providing greater control and flexibility to suit your needs.
+- **🔒 Chat Controls Permissions**: Admins can now disable chat controls access for users, offering tighter management and customization over user interactions.
+- **🔍 Web Search & Image Generation Permissions**: Easily disable web search and image generation for specific users, improving workflow governance and security for certain environments.
+- **🗂️ S3 and GCS Storage Provider Support**: Scaled deployments now benefit from expanded storage options with Amazon S3 and Google Cloud Storage seamlessly integrated as providers.
+- **🎨 Enhanced Model Management**: Reintroduced the ability to download and delete models directly in the admin models settings page to minimize user confusion and aid efficient model management.
+- **🔗 Improved Connection Handling**: Enhanced backend to smoothly handle multiple identical base URLs, allowing more flexible multi-instance configurations with fewer hiccups.
+- **✨ General UI/UX Refinements**: Numerous tweaks across the WebUI make navigation and usability even more user-friendly and intuitive.
+- **🌍 Translation Enhancements**: Various translation updates ensure smoother and more polished interactions for international users.
+
+### Fixed
+
+- **⚡ MPS Functionality for Mac Users**: Fixed MPS support, ensuring smooth performance and compatibility for Mac users leveraging MPS.
+- **📡 Ollama Connection Management**: Resolved the issue where deleting all Ollama connections prevented adding new ones.
+
+### Changed
+
+- **⚙️ General Stability Refac**: Backend refactoring delivers a more stable, robust platform.
+- **🖥️ Desktop App Preparations**: Ongoing work to support the upcoming Open WebUI desktop app. Follow our progress and updates here: https://github.com/open-webui/desktop
+
+## [0.5.4] - 2025-01-05
+
+### Added
+
+- **🔄 Clone Shared Chats**: Effortlessly clone shared chats to save time and streamline collaboration, perfect for reusing insightful discussions or custom setups.
+- **📣 Native Notifications for Channel Messages**: Stay informed with integrated desktop notifications for channel messages, ensuring you never miss important updates while multitasking.
+- **🔥 Torch MPS Support**: MPS support for Mac users when Open WebUI is installed directly, offering better performance and compatibility for AI workloads.
+- **🌍 Enhanced Translations**: Small improvements to various translations, ensuring a smoother global user experience.
+
+### Fixed
+
+- **🖼️ Image-Only Messages in Channels**: You can now send images without accompanying text or content in channels.
+- **❌ Proper Exception Handling**: Enhanced error feedback by ensuring exceptions are raised clearly, reducing confusion and promoting smoother debugging.
+- **🔍 RAG Query Generation Restored**: Fixed query generation issues for Retrieval-Augmented Generation, improving retrieval accuracy and ensuring seamless functionality.
+- **📩 MOA Response Functionality Fixed**: Addressed an error with the MOA response generation feature.
+- **💬 Channel Thread Loading with 50+ Messages**: Resolved an issue where channel threads stalled when exceeding 50 messages, ensuring smooth navigation in active discussions.
+- **🔑 API Endpoint Restrictions Resolution**: Fixed a critical bug where the 'API_KEY_ALLOWED_ENDPOINTS' setting was not functioning as intended, ensuring API access is limited to specified endpoints for enhanced security.
+- **🛠️ Action Functions Restored**: Corrected an issue preventing action functions from working, restoring their utility for customized automations and workflows.
+- **📂 Temporary Chat JSON Export Fix**: Resolved a bug blocking temporary chats from being exported in JSON format, ensuring seamless data portability.
+
+### Changed
+
+- **🎛️ Sidebar UI Tweaks**: Chat folders, including pinned folders, now display below the Chats section for better organization; the "New Folder" button has been relocated to the Chats section for a more intuitive workflow.
+- **🏗️ Real-Time Save Disabled by Default**: The 'ENABLE_REALTIME_CHAT_SAVE' setting is now off by default, boosting response speed for users who prioritize performance in high-paced workflows or less critical scenarios.
+- **🎤 Audio Input Echo Cancellation**: Audio input now features echo cancellation enabled by default, reducing audio feedback for improved clarity during conversations or voice-based interactions.
+- **🔧 General Reliability Improvements**: Numerous under-the-hood enhancements have been made to improve platform stability, boost overall performance, and ensure a more seamless, dependable experience across workflows.
+
+## [0.5.3] - 2024-12-31
+
+### Added
+
+- **💬 Channel Reactions with Built-In Emoji Picker**: Easily express yourself in channel threads and messages with reactions, featuring an intuitive built-in emoji picker for seamless selection.
+- **🧵 Threads for Channels**: Organize discussions within channels by creating threads, improving clarity and fostering focused conversations.
+- **🔄 Reset Button for SVG Pan/Zoom**: Added a handy reset button to SVG Pan/Zoom, allowing users to quickly return diagrams or visuals to their default state without hassle.
+- **⚡ Realtime Chat Save Environment Variable**: Introduced the ENABLE_REALTIME_CHAT_SAVE environment variable. Choose between faster responses by disabling realtime chat saving or ensuring chunk-by-chunk data persistency for critical operations.
+- **🌍 Translation Enhancements**: Updated and refined translations across multiple languages, providing a smoother experience for international users.
+- **📚 Improved Documentation**: Expanded documentation on functions, including clearer guidance on function plugins and detailed instructions for migrating to v0.5. This ensures users can adapt and harness new updates more effectively. (https://docs.openwebui.com/features/plugin/)
+
+### Fixed
+
+- **🛠️ Ollama Parameters Respected**: Resolved an issue where input parameters for Ollama were being ignored, ensuring precise and consistent model behavior.
+- **🔧 Function Plugin Outlet Hook Reliability**: Fixed a bug causing issues with 'event_emitter' and outlet hooks in filter function plugins, guaranteeing smoother operation within custom extensions.
+- **🖋️ Weird Custom Status Descriptions**: Adjusted the formatting and functionality for custom user statuses, ensuring they display correctly and intuitively.
+- **🔗 Restored API Functionality**: Fixed a critical issue where APIs were not operational for certain configurations, ensuring uninterrupted access.
+- **⏳ Custom Pipe Function Completion**: Resolved an issue where chats using specific custom pipe function plugins weren’t finishing properly, restoring consistent chat workflows.
+- **✅ General Stability Enhancements**: Implemented various under-the-hood improvements to boost overall reliability, ensuring smoother and more consistent performance across the WebUI.
+
+## [0.5.2] - 2024-12-26
+
+### Added
+
+- **🖊️ Typing Indicators in Channels**: Know exactly who’s typing in real-time within your channels, enhancing collaboration and keeping everyone engaged.
+- **👤 User Status Indicators**: Quickly view a user’s status by clicking their profile image in channels for better coordination and availability insights.
+- **🔒 Configurable API Key Authentication Restrictions**: Flexibly configure endpoint restrictions for API key authentication, now off by default for a smoother setup in trusted environments.
+
+### Fixed
+
+- **🔧 Playground Functionality Restored**: Resolved a critical issue where the playground wasn’t working, ensuring seamless experimentation and troubleshooting workflows.
+- **📊 Corrected Ollama Usage Statistics**: Fixed a calculation error in Ollama’s usage statistics, providing more accurate tracking and insights for better resource management.
+- **🔗 Pipelines Outlet Hook Registration**: Addressed an issue where outlet hooks for pipelines weren’t registered, restoring functionality and consistency in pipeline workflows.
+- **🎨 Image Generation Error**: Resolved a persistent issue causing errors with 'get_automatic1111_api_auth()' to ensure smooth image generation workflows.
+- **🎙️ Text-to-Speech Error**: Fixed the missing argument in Eleven Labs’ 'get_available_voices()', restoring full text-to-speech capabilities for uninterrupted voice interactions.
+- **🖋️ Title Generation Issue**: Fixed a bug where title generation was not working in certain cases, ensuring consistent and reliable chat organization.
+
+## [0.5.1] - 2024-12-25
+
+### Added
+
+- **🔕 Notification Sound Toggle**: Added a new setting under Settings > Interface to disable notification sounds, giving you greater control over your workspace environment and focus.
+
+### Fixed
+
+- **🔄 Non-Streaming Response Visibility**: Resolved an issue where non-streaming responses were not displayed, ensuring all responses are now reliably shown in your conversations.
+- **🖋️ Title Generation with OpenAI APIs**: Fixed a bug preventing title generation when using OpenAI APIs, restoring the ability to automatically generate chat titles for smoother organization.
+- **👥 Admin Panel User List**: Addressed the issue where only 50 users were visible in the admin panel. You can now manage and view all users without restrictions.
+- **🖼️ Image Generation Error**: Fixed the issue causing 'get_automatic1111_api_auth()' errors in image generation, ensuring seamless creative workflows.
+- **⚙️ Pipeline Settings Loading Issue**: Resolved a problem where pipeline settings were stuck at the loading screen, restoring full configurability in the admin panel.
+
+## [0.5.0] - 2024-12-25
+
+### Added
+
+- **💬 True Asynchronous Chat Support**: Create chats, navigate away, and return anytime with responses ready. Ideal for reasoning models and multi-agent workflows, enhancing multitasking like never before.
+- **🔔 Chat Completion Notifications**: Never miss a completed response. Receive instant in-UI notifications when a chat finishes in a non-active tab, keeping you updated while you work elsewhere.
+- **🌐 Notification Webhook Integration**: Get alerts via webhooks even when your tab is closed! Configure your webhook URL in Settings > Account and receive timely updates for long-running chats or external integration needs.
+- **📚 Channels (Beta)**: Explore Discord/Slack-style chat rooms designed for real-time collaboration between users and AIs. Build bots for channels and unlock asynchronous communication for proactive multi-agent workflows. Opt-in via Admin Settings > General. A Comprehensive Bot SDK tutorial (https://github.com/open-webui/bot) is incoming, so stay tuned!
+- **🖼️ Client-Side Image Compression**: Now compress images before upload (Settings > Interface), saving bandwidth and improving performance seamlessly.
+- **🛠️ OAuth Management for User Groups**: Enable group-level management via OAuth integration for enhanced control and scalability in collaborative environments.
+- **✅ Structured Output for Ollama**: Pass structured data output directly to Ollama, unlocking new possibilities for streamlined automation and precise data handling.
+- **📜 Offline Swagger Documentation**: Developer-friendly Swagger API docs are now available offline, ensuring full accessibility wherever you are.
+- **📸 Quick Screen Capture Button**: Effortlessly capture your screen with a single click from the message input menu.
+- **🌍 i18n Updates**: Improved and refined translations across several languages, including Ukrainian, German, Brazilian Portuguese, Catalan, and more, ensuring a seamless global user experience.
+
+### Fixed
+
+- **📋 Table Export to CSV**: Resolved issues with CSV export where headers were missing or errors occurred due to values with commas, ensuring smooth and reliable data handling.
+- **🔓 BYPASS_MODEL_ACCESS_CONTROL**: Fixed an issue where users could see models but couldn’t use them with 'BYPASS_MODEL_ACCESS_CONTROL=True', restoring proper functionality for environments leveraging this setting.
+
+### Changed
+
+- **💡 API Key Authentication Restriction**: Narrowed API key auth permissions to '/api/models' and '/api/chat/completions' for enhanced security and better API governance.
+- **⚙️ Backend Overhaul for Performance**: Major backend restructuring; a heads-up that some "Functions" using internal variables may face compatibility issues. Moving forward, websocket support is mandatory to ensure Open WebUI operates seamlessly.
+
+### Removed
+
+- **⚠️ Legacy Functionality Clean-Up**: Deprecated outdated backend systems that were non-essential or overlapped with newer implementations, allowing for a leaner, more efficient platform.
+
 ## [0.4.8] - 2024-12-07
 
 ### Added

CODE_OF_CONDUCT.md

@@ -2,13 +2,13 @@
 
 ## Our Pledge
 
-As members, contributors, and leaders of this community, we pledge to make participation in our open-source project a harassment-free experience for everyone, regardless of age, body size, visible or invisible disability, ethnicity, sex characteristics, gender identity and expression, level of experience, education, socio-economic status, nationality, personal appearance, race, religion, or sexual identity and orientation.
+As members, contributors, and leaders of this community, we pledge to make participation in our project a harassment-free experience for everyone, regardless of age, body size, visible or invisible disability, ethnicity, sex characteristics, gender identity and expression, level of experience, education, socioeconomic status, nationality, personal appearance, race, religion, or sexual identity and orientation.
 
 We are committed to creating and maintaining an open, respectful, and professional environment where positive contributions and meaningful discussions can flourish. By participating in this project, you agree to uphold these values and align your behavior to the standards outlined in this Code of Conduct.
 
 ## Why These Standards Are Important
 
-Open-source projects rely on a community of volunteers dedicating their time, expertise, and effort toward a shared goal. These projects are inherently collaborative but also fragile, as the success of the project depends on the goodwill, energy, and productivity of those involved.
+Projects rely on a community of volunteers dedicating their time, expertise, and effort toward a shared goal. These projects are inherently collaborative but also fragile, as the success of the project depends on the goodwill, energy, and productivity of those involved.
 
 Maintaining a positive and respectful environment is essential to safeguarding the integrity of this project and protecting contributors' efforts. Behavior that disrupts this atmosphere—whether through hostility, entitlement, or unprofessional conduct—can severely harm the morale and productivity of the community. **Strict enforcement of these standards ensures a safe and supportive space for meaningful collaboration.**
 
@@ -79,7 +79,7 @@ This approach ensures that disruptive behaviors are addressed swiftly and decisi
 
 ## Why Zero Tolerance Is Necessary
 
-Open-source projects thrive on collaboration, goodwill, and mutual respect. Toxic behaviors—such as entitlement, hostility, or persistent negativity—threaten not just individual contributors but the health of the project as a whole. Allowing such behaviors to persist robs contributors of their time, energy, and enthusiasm for the work they do.
+Projects thrive on collaboration, goodwill, and mutual respect. Toxic behaviors—such as entitlement, hostility, or persistent negativity—threaten not just individual contributors but the health of the project as a whole. Allowing such behaviors to persist robs contributors of their time, energy, and enthusiasm for the work they do.
 
 By enforcing a zero-tolerance policy, we ensure that the community remains a safe, welcoming space for all participants. These measures are not about harshness—they are about protecting contributors and fostering a productive environment where innovation can thrive.
 

CONTRIBUTOR_LICENSE_AGREEMENT

@@ -0,0 +1,7 @@
+# Open WebUI Contributor License Agreement
+
+By submitting my contributions to Open WebUI, I grant Open WebUI full freedom to use my work in any way they choose, under any terms they like, both now and in the future. This approach helps ensure the project remains unified, flexible, and easy to maintain, while empowering Open WebUI to respond quickly to the needs of its users and the wider community.
+
+Taking part in this process means my work can be seamlessly integrated and combined with others, ensuring longevity and adaptability for everyone who benefits from the Open WebUI project. This collaborative approach strengthens the project’s future and helps guarantee that improvements can always be shared and distributed in the most effective way possible.
+
+**_To the fullest extent permitted by law, my contributions are provided on an “as is” basis, with no warranties or guarantees of any kind, and I disclaim any liability for any issues or damages arising from their use or incorporation into the project, regardless of the type of legal claim._**

Dockerfile

@@ -4,7 +4,7 @@
 ARG USE_CUDA=false
 ARG USE_OLLAMA=false
 # Tested with cu117 for CUDA 11 and cu121 for CUDA 12 (default)
-ARG USE_CUDA_VER=cu121
+ARG USE_CUDA_VER=cu128
 # any sentence transformer model; models to use can be found at https://huggingface.co/models?library=sentence-transformers
 # Leaderboard: https://huggingface.co/spaces/mteb/leaderboard 
 # for better performance and multilangauge support use "intfloat/multilingual-e5-large" (~2.5GB) or "intfloat/multilingual-e5-base" (~1.5GB)
@@ -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 && \

LICENSE

@@ -1,21 +1,33 @@
-MIT License
+Copyright (c) 2023-2025 Timothy Jaeryang Baek (Open WebUI)
+All rights reserved.
 
-Copyright (c) 2023 Timothy Jaeryang Baek
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
 
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
+1. Redistributions of source code must retain the above copyright notice, this
+   list of conditions and the following disclaimer.
 
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
+2. Redistributions in binary form must reproduce the above copyright notice,
+   this list of conditions and the following disclaimer in the documentation
+   and/or other materials provided with the distribution.
 
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
+3. Neither the name of the copyright holder nor the names of its
+   contributors may be used to endorse or promote products derived from
+   this software without specific prior written permission.
+
+4. Notwithstanding any other provision of this License, and as a material condition of the rights granted herein, licensees are strictly prohibited from altering, removing, obscuring, or replacing any "Open WebUI" branding, including but not limited to the name, logo, or any visual, textual, or symbolic identifiers that distinguish the software and its interfaces, in any deployment or distribution, regardless of the number of users, except as explicitly set forth in Clauses 5 and 6 below.
+
+5. The branding restriction enumerated in Clause 4 shall not apply in the following limited circumstances: (i) deployments or distributions where the total number of end users (defined as individual natural persons with direct access to the application) does not exceed fifty (50) within any rolling thirty (30) day period; (ii) cases in which the licensee is an official contributor to the codebase—with a substantive code change successfully merged into the main branch of the official codebase maintained by the copyright holder—who has obtained specific prior written permission for branding adjustment from the copyright holder; or (iii) where the licensee has obtained a duly executed enterprise license expressly permitting such modification. For all other cases, any removal or alteration of the "Open WebUI" branding shall constitute a material breach of license.
+
+6. All code, modifications, or derivative works incorporated into this project prior to the incorporation of this branding clause remain licensed under the BSD 3-Clause License, and prior contributors retain all BSD-3 rights therein; if any such contributor requests the removal of their BSD-3-licensed code, the copyright holder will do so, and any replacement code will be licensed under the project's primary license then in effect. By contributing after this clause's adoption, you agree to the project's Contributor License Agreement (CLA) and to these updated terms for all new contributions.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

README.md

@@ -7,14 +7,20 @@
 ![GitHub language count](https://img.shields.io/github/languages/count/open-webui/open-webui)
 ![GitHub top language](https://img.shields.io/github/languages/top/open-webui/open-webui)
 ![GitHub last commit](https://img.shields.io/github/last-commit/open-webui/open-webui?color=red)
-![Hits](https://hits.seeyoufarm.com/api/count/incr/badge.svg?url=https%3A%2F%2Fgithub.com%2Follama-webui%2Follama-wbui&count_bg=%2379C83D&title_bg=%23555555&icon=&icon_color=%23E7E7E7&title=hits&edge_flat=false)
 [![Discord](https://img.shields.io/badge/Discord-Open_WebUI-blue?logo=discord&logoColor=white)](https://discord.gg/5rJgQTnV4s)
 [![](https://img.shields.io/static/v1?label=Sponsor&message=%E2%9D%A4&logo=GitHub&color=%23fe8e86)](https://github.com/sponsors/tjbck)
 
-Open WebUI is an [extensible](https://github.com/open-webui/pipelines), feature-rich, and user-friendly self-hosted WebUI designed to operate entirely offline. It supports various LLM runners, including Ollama and OpenAI-compatible APIs. For more information, be sure to check out our [Open WebUI Documentation](https://docs.openwebui.com/).
+**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**.
 
 ![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.
@@ -55,9 +61,26 @@ Open WebUI is an [extensible](https://github.com/open-webui/pipelines), feature-
 
 Want to learn more about Open WebUI's features? Check out our [Open WebUI documentation](https://docs.openwebui.com/features) for a comprehensive overview!
 
-## 🔗 Also Check Out Open WebUI Community!
+## Sponsors 🙌
 
-Don't forget to explore our sibling project, [Open WebUI Community](https://openwebui.com/), where you can discover, download, and explore customized Modelfiles. Open WebUI Community offers a wide range of exciting possibilities for enhancing your chat interactions with Open WebUI! 🚀
+#### Emerald
+
+<table>
+  <tr>
+    <td>
+      <a href="https://n8n.io/" target="_blank">
+        <img src="https://docs.openwebui.com/sponsors/logos/n8n.png" alt="n8n" style="width: 8rem; height: 8rem; border-radius: .75rem;" />
+      </a>
+    </td>
+    <td>
+      Does your interface have a backend yet?<br>Try <a href="https://n8n.io/">n8n</a>
+    </td>
+  </tr>
+</table>
+
+---
+
+We are incredibly grateful for the generous support of our sponsors. Their contributions help us to maintain and improve our project, ensuring we can continue to deliver quality work to our community. Thank you!
 
 ## How to Install 🚀
 
@@ -172,7 +195,7 @@ 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 🌙
 
@@ -185,13 +208,21 @@ If you want to try out the latest bleeding-edge features and are okay with occas
 docker run -d -p 3000:8080 -v open-webui:/app/backend/data --name open-webui --add-host=host.docker.internal:host-gateway --restart always ghcr.io/open-webui/open-webui:dev
 ```
 
+### Offline Mode
+
+If you are running Open WebUI in an offline environment, you can set the `HF_HUB_OFFLINE` environment variable to `1` to prevent attempts to download models from the internet.
+
+```bash
+export HF_HUB_OFFLINE=1
+```
+
 ## What's Next? 🌟
 
 Discover upcoming features on our roadmap in the [Open WebUI Documentation](https://docs.openwebui.com/roadmap/).
 
 ## License 📜
 
-This project is licensed under the [MIT License](LICENSE) - see the [LICENSE](LICENSE) file for details. 📄
+This project is licensed under the [Open WebUI License](LICENSE), a revised BSD-3-Clause license. You receive all the same rights as the classic BSD-3 license: you can use, modify, and distribute the software, including in proprietary and commercial products, with minimal restrictions. The only additional requirement is to preserve the "Open WebUI" branding, as detailed in the LICENSE file. For full terms, see the [LICENSE](LICENSE) document. 📄
 
 ## Support 💬
 

backend/open_webui/__init__.py

@@ -5,12 +5,31 @@ from pathlib import Path
 
 import typer
 import uvicorn
+from typing import Optional
+from typing_extensions import Annotated
 
 app = typer.Typer()
 
 KEY_FILE = Path.cwd() / ".webui_secret_key"
 
 
+def version_callback(value: bool):
+    if value:
+        from open_webui.env import VERSION
+
+        typer.echo(f"Open WebUI version: {VERSION}")
+        raise typer.Exit()
+
+
+@app.command()
+def main(
+    version: Annotated[
+        Optional[bool], typer.Option("--version", callback=version_callback)
+    ] = None,
+):
+    pass
+
+
 @app.command()
 def serve(
     host: str = "0.0.0.0",
@@ -54,8 +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/apps/audio/main.py

@@ -1,703 +0,0 @@
-import hashlib
-import json
-import logging
-import os
-import uuid
-from functools import lru_cache
-from pathlib import Path
-from pydub import AudioSegment
-from pydub.silence import split_on_silence
-
-import aiohttp
-import aiofiles
-import requests
-from open_webui.config import (
-    AUDIO_STT_ENGINE,
-    AUDIO_STT_MODEL,
-    AUDIO_STT_OPENAI_API_BASE_URL,
-    AUDIO_STT_OPENAI_API_KEY,
-    AUDIO_TTS_API_KEY,
-    AUDIO_TTS_ENGINE,
-    AUDIO_TTS_MODEL,
-    AUDIO_TTS_OPENAI_API_BASE_URL,
-    AUDIO_TTS_OPENAI_API_KEY,
-    AUDIO_TTS_SPLIT_ON,
-    AUDIO_TTS_VOICE,
-    AUDIO_TTS_AZURE_SPEECH_REGION,
-    AUDIO_TTS_AZURE_SPEECH_OUTPUT_FORMAT,
-    CACHE_DIR,
-    CORS_ALLOW_ORIGIN,
-    WHISPER_MODEL,
-    WHISPER_MODEL_AUTO_UPDATE,
-    WHISPER_MODEL_DIR,
-    AppConfig,
-)
-
-from open_webui.constants import ERROR_MESSAGES
-from open_webui.env import (
-    ENV,
-    SRC_LOG_LEVELS,
-    DEVICE_TYPE,
-    ENABLE_FORWARD_USER_INFO_HEADERS,
-)
-
-from fastapi import Depends, FastAPI, File, HTTPException, Request, UploadFile, status
-from fastapi.middleware.cors import CORSMiddleware
-from fastapi.responses import FileResponse
-from pydantic import BaseModel
-from open_webui.utils.utils import get_admin_user, get_verified_user
-
-# Constants
-MAX_FILE_SIZE_MB = 25
-MAX_FILE_SIZE = MAX_FILE_SIZE_MB * 1024 * 1024  # Convert MB to bytes
-
-
-log = logging.getLogger(__name__)
-log.setLevel(SRC_LOG_LEVELS["AUDIO"])
-
-app = FastAPI(
-    docs_url="/docs" if ENV == "dev" else None,
-    openapi_url="/openapi.json" if ENV == "dev" else None,
-    redoc_url=None,
-)
-
-app.add_middleware(
-    CORSMiddleware,
-    allow_origins=CORS_ALLOW_ORIGIN,
-    allow_credentials=True,
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
-
-app.state.config = AppConfig()
-
-app.state.config.STT_OPENAI_API_BASE_URL = AUDIO_STT_OPENAI_API_BASE_URL
-app.state.config.STT_OPENAI_API_KEY = AUDIO_STT_OPENAI_API_KEY
-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.faster_whisper_model = None
-
-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
-app.state.config.TTS_ENGINE = AUDIO_TTS_ENGINE
-app.state.config.TTS_MODEL = AUDIO_TTS_MODEL
-app.state.config.TTS_VOICE = AUDIO_TTS_VOICE
-app.state.config.TTS_API_KEY = AUDIO_TTS_API_KEY
-app.state.config.TTS_SPLIT_ON = AUDIO_TTS_SPLIT_ON
-
-
-app.state.speech_synthesiser = None
-app.state.speech_speaker_embeddings_dataset = None
-
-app.state.config.TTS_AZURE_SPEECH_REGION = AUDIO_TTS_AZURE_SPEECH_REGION
-app.state.config.TTS_AZURE_SPEECH_OUTPUT_FORMAT = AUDIO_TTS_AZURE_SPEECH_OUTPUT_FORMAT
-
-# setting device type for whisper model
-whisper_device_type = DEVICE_TYPE if DEVICE_TYPE and DEVICE_TYPE == "cuda" else "cpu"
-log.info(f"whisper_device_type: {whisper_device_type}")
-
-SPEECH_CACHE_DIR = Path(CACHE_DIR).joinpath("./audio/speech/")
-SPEECH_CACHE_DIR.mkdir(parents=True, exist_ok=True)
-
-
-def set_faster_whisper_model(model: str, auto_update: bool = False):
-    if model and app.state.config.STT_ENGINE == "":
-        from faster_whisper import WhisperModel
-
-        faster_whisper_kwargs = {
-            "model_size_or_path": model,
-            "device": whisper_device_type,
-            "compute_type": "int8",
-            "download_root": WHISPER_MODEL_DIR,
-            "local_files_only": not auto_update,
-        }
-
-        try:
-            app.state.faster_whisper_model = WhisperModel(**faster_whisper_kwargs)
-        except Exception:
-            log.warning(
-                "WhisperModel initialization failed, attempting download with local_files_only=False"
-            )
-            faster_whisper_kwargs["local_files_only"] = False
-            app.state.faster_whisper_model = WhisperModel(**faster_whisper_kwargs)
-
-    else:
-        app.state.faster_whisper_model = None
-
-
-class TTSConfigForm(BaseModel):
-    OPENAI_API_BASE_URL: str
-    OPENAI_API_KEY: str
-    API_KEY: str
-    ENGINE: str
-    MODEL: str
-    VOICE: str
-    SPLIT_ON: str
-    AZURE_SPEECH_REGION: str
-    AZURE_SPEECH_OUTPUT_FORMAT: str
-
-
-class STTConfigForm(BaseModel):
-    OPENAI_API_BASE_URL: str
-    OPENAI_API_KEY: str
-    ENGINE: str
-    MODEL: str
-    WHISPER_MODEL: str
-
-
-class AudioConfigUpdateForm(BaseModel):
-    tts: TTSConfigForm
-    stt: STTConfigForm
-
-
-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."""
-    if not os.path.isfile(file_path):
-        print(f"File not found: {file_path}")
-        return False
-
-    info = mediainfo(file_path)
-    if (
-        info.get("codec_name") == "aac"
-        and info.get("codec_type") == "audio"
-        and info.get("codec_tag_string") == "mp4a"
-    ):
-        return True
-    return False
-
-
-def convert_mp4_to_wav(file_path, output_path):
-    """Convert MP4 audio file to WAV format."""
-    audio = AudioSegment.from_file(file_path, format="mp4")
-    audio.export(output_path, format="wav")
-    print(f"Converted {file_path} to {output_path}")
-
-
-@app.get("/config")
-async def get_audio_config(user=Depends(get_admin_user)):
-    return {
-        "tts": {
-            "OPENAI_API_BASE_URL": app.state.config.TTS_OPENAI_API_BASE_URL,
-            "OPENAI_API_KEY": app.state.config.TTS_OPENAI_API_KEY,
-            "API_KEY": app.state.config.TTS_API_KEY,
-            "ENGINE": app.state.config.TTS_ENGINE,
-            "MODEL": app.state.config.TTS_MODEL,
-            "VOICE": app.state.config.TTS_VOICE,
-            "SPLIT_ON": app.state.config.TTS_SPLIT_ON,
-            "AZURE_SPEECH_REGION": app.state.config.TTS_AZURE_SPEECH_REGION,
-            "AZURE_SPEECH_OUTPUT_FORMAT": app.state.config.TTS_AZURE_SPEECH_OUTPUT_FORMAT,
-        },
-        "stt": {
-            "OPENAI_API_BASE_URL": app.state.config.STT_OPENAI_API_BASE_URL,
-            "OPENAI_API_KEY": app.state.config.STT_OPENAI_API_KEY,
-            "ENGINE": app.state.config.STT_ENGINE,
-            "MODEL": app.state.config.STT_MODEL,
-            "WHISPER_MODEL": app.state.config.WHISPER_MODEL,
-        },
-    }
-
-
-@app.post("/config/update")
-async def update_audio_config(
-    form_data: AudioConfigUpdateForm, user=Depends(get_admin_user)
-):
-    app.state.config.TTS_OPENAI_API_BASE_URL = form_data.tts.OPENAI_API_BASE_URL
-    app.state.config.TTS_OPENAI_API_KEY = form_data.tts.OPENAI_API_KEY
-    app.state.config.TTS_API_KEY = form_data.tts.API_KEY
-    app.state.config.TTS_ENGINE = form_data.tts.ENGINE
-    app.state.config.TTS_MODEL = form_data.tts.MODEL
-    app.state.config.TTS_VOICE = form_data.tts.VOICE
-    app.state.config.TTS_SPLIT_ON = form_data.tts.SPLIT_ON
-    app.state.config.TTS_AZURE_SPEECH_REGION = form_data.tts.AZURE_SPEECH_REGION
-    app.state.config.TTS_AZURE_SPEECH_OUTPUT_FORMAT = (
-        form_data.tts.AZURE_SPEECH_OUTPUT_FORMAT
-    )
-
-    app.state.config.STT_OPENAI_API_BASE_URL = form_data.stt.OPENAI_API_BASE_URL
-    app.state.config.STT_OPENAI_API_KEY = form_data.stt.OPENAI_API_KEY
-    app.state.config.STT_ENGINE = form_data.stt.ENGINE
-    app.state.config.STT_MODEL = form_data.stt.MODEL
-    app.state.config.WHISPER_MODEL = form_data.stt.WHISPER_MODEL
-    set_faster_whisper_model(form_data.stt.WHISPER_MODEL, WHISPER_MODEL_AUTO_UPDATE)
-
-    return {
-        "tts": {
-            "OPENAI_API_BASE_URL": app.state.config.TTS_OPENAI_API_BASE_URL,
-            "OPENAI_API_KEY": app.state.config.TTS_OPENAI_API_KEY,
-            "API_KEY": app.state.config.TTS_API_KEY,
-            "ENGINE": app.state.config.TTS_ENGINE,
-            "MODEL": app.state.config.TTS_MODEL,
-            "VOICE": app.state.config.TTS_VOICE,
-            "SPLIT_ON": app.state.config.TTS_SPLIT_ON,
-            "AZURE_SPEECH_REGION": app.state.config.TTS_AZURE_SPEECH_REGION,
-            "AZURE_SPEECH_OUTPUT_FORMAT": app.state.config.TTS_AZURE_SPEECH_OUTPUT_FORMAT,
-        },
-        "stt": {
-            "OPENAI_API_BASE_URL": app.state.config.STT_OPENAI_API_BASE_URL,
-            "OPENAI_API_KEY": app.state.config.STT_OPENAI_API_KEY,
-            "ENGINE": app.state.config.STT_ENGINE,
-            "MODEL": app.state.config.STT_MODEL,
-            "WHISPER_MODEL": app.state.config.WHISPER_MODEL,
-        },
-    }
-
-
-def load_speech_pipeline():
-    from transformers import pipeline
-    from datasets import load_dataset
-
-    if app.state.speech_synthesiser is None:
-        app.state.speech_synthesiser = pipeline(
-            "text-to-speech", "microsoft/speecht5_tts"
-        )
-
-    if app.state.speech_speaker_embeddings_dataset is None:
-        app.state.speech_speaker_embeddings_dataset = load_dataset(
-            "Matthijs/cmu-arctic-xvectors", split="validation"
-        )
-
-
-@app.post("/speech")
-async def speech(request: Request, user=Depends(get_verified_user)):
-    body = await request.body()
-    name = hashlib.sha256(body).hexdigest()
-
-    file_path = SPEECH_CACHE_DIR.joinpath(f"{name}.mp3")
-    file_body_path = SPEECH_CACHE_DIR.joinpath(f"{name}.json")
-
-    # Check if the file already exists in the cache
-    if file_path.is_file():
-        return FileResponse(file_path)
-
-    if app.state.config.TTS_ENGINE == "openai":
-        headers = {}
-        headers["Authorization"] = f"Bearer {app.state.config.TTS_OPENAI_API_KEY}"
-        headers["Content-Type"] = "application/json"
-
-        if ENABLE_FORWARD_USER_INFO_HEADERS:
-            headers["X-OpenWebUI-User-Name"] = user.name
-            headers["X-OpenWebUI-User-Id"] = user.id
-            headers["X-OpenWebUI-User-Email"] = user.email
-            headers["X-OpenWebUI-User-Role"] = user.role
-
-        try:
-            body = body.decode("utf-8")
-            body = json.loads(body)
-            body["model"] = app.state.config.TTS_MODEL
-            body = json.dumps(body).encode("utf-8")
-        except Exception:
-            pass
-
-        try:
-            async with aiohttp.ClientSession() as session:
-                async with session.post(
-                    url=f"{app.state.config.TTS_OPENAI_API_BASE_URL}/audio/speech",
-                    data=body,
-                    headers=headers,
-                ) as r:
-                    r.raise_for_status()
-                    async with aiofiles.open(file_path, "wb") as f:
-                        await f.write(await r.read())
-
-                    async with aiofiles.open(file_body_path, "w") as f:
-                        await f.write(json.dumps(json.loads(body.decode("utf-8"))))
-
-            return FileResponse(file_path)
-
-        except Exception as e:
-            log.exception(e)
-            error_detail = "Open WebUI: Server Connection Error"
-            try:
-                if r.status != 200:
-                    res = await r.json()
-                    if "error" in res:
-                        error_detail = f"External: {res['error']['message']}"
-            except Exception:
-                error_detail = f"External: {e}"
-
-            raise HTTPException(
-                status_code=getattr(r, "status", 500),
-                detail=error_detail,
-            )
-
-    elif app.state.config.TTS_ENGINE == "elevenlabs":
-        try:
-            payload = json.loads(body.decode("utf-8"))
-        except Exception as e:
-            log.exception(e)
-            raise HTTPException(status_code=400, detail="Invalid JSON payload")
-
-        voice_id = payload.get("voice", "")
-        if voice_id not in get_available_voices():
-            raise HTTPException(
-                status_code=400,
-                detail="Invalid voice id",
-            )
-
-        url = f"https://api.elevenlabs.io/v1/text-to-speech/{voice_id}"
-        headers = {
-            "Accept": "audio/mpeg",
-            "Content-Type": "application/json",
-            "xi-api-key": app.state.config.TTS_API_KEY,
-        }
-        data = {
-            "text": payload["input"],
-            "model_id": app.state.config.TTS_MODEL,
-            "voice_settings": {"stability": 0.5, "similarity_boost": 0.5},
-        }
-
-        try:
-            async with aiohttp.ClientSession() as session:
-                async with session.post(url, json=data, headers=headers) as r:
-                    r.raise_for_status()
-                    async with aiofiles.open(file_path, "wb") as f:
-                        await f.write(await r.read())
-
-                    async with aiofiles.open(file_body_path, "w") as f:
-                        await f.write(json.dumps(json.loads(body.decode("utf-8"))))
-
-            return FileResponse(file_path)
-
-        except Exception as e:
-            log.exception(e)
-            error_detail = "Open WebUI: Server Connection Error"
-            try:
-                if r.status != 200:
-                    res = await r.json()
-                    if "error" in res:
-                        error_detail = f"External: {res['error']['message']}"
-            except Exception:
-                error_detail = f"External: {e}"
-
-            raise HTTPException(
-                status_code=getattr(r, "status", 500),
-                detail=error_detail,
-            )
-
-    elif app.state.config.TTS_ENGINE == "azure":
-        try:
-            payload = json.loads(body.decode("utf-8"))
-        except Exception as e:
-            log.exception(e)
-            raise HTTPException(status_code=400, detail="Invalid JSON payload")
-
-        region = app.state.config.TTS_AZURE_SPEECH_REGION
-        language = app.state.config.TTS_VOICE
-        locale = "-".join(app.state.config.TTS_VOICE.split("-")[:1])
-        output_format = app.state.config.TTS_AZURE_SPEECH_OUTPUT_FORMAT
-        url = f"https://{region}.tts.speech.microsoft.com/cognitiveservices/v1"
-
-        headers = {
-            "Ocp-Apim-Subscription-Key": app.state.config.TTS_API_KEY,
-            "Content-Type": "application/ssml+xml",
-            "X-Microsoft-OutputFormat": output_format,
-        }
-
-        data = f"""<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xml:lang="{locale}">
-                <voice name="{language}">{payload["input"]}</voice>
-            </speak>"""
-
-        try:
-            async with aiohttp.ClientSession() as session:
-                async with session.post(url, headers=headers, data=data) as response:
-                    if response.status == 200:
-                        async with aiofiles.open(file_path, "wb") as f:
-                            await f.write(await response.read())
-                        return FileResponse(file_path)
-                    else:
-                        error_msg = f"Error synthesizing speech - {response.reason}"
-                        log.error(error_msg)
-                        raise HTTPException(status_code=500, detail=error_msg)
-        except Exception as e:
-            log.exception(e)
-            raise HTTPException(status_code=500, detail=str(e))
-    elif app.state.config.TTS_ENGINE == "transformers":
-        payload = None
-        try:
-            payload = json.loads(body.decode("utf-8"))
-        except Exception as e:
-            log.exception(e)
-            raise HTTPException(status_code=400, detail="Invalid JSON payload")
-
-        import torch
-        import soundfile as sf
-
-        load_speech_pipeline()
-
-        embeddings_dataset = app.state.speech_speaker_embeddings_dataset
-
-        speaker_index = 6799
-        try:
-            speaker_index = embeddings_dataset["filename"].index(
-                app.state.config.TTS_MODEL
-            )
-        except Exception:
-            pass
-
-        speaker_embedding = torch.tensor(
-            embeddings_dataset[speaker_index]["xvector"]
-        ).unsqueeze(0)
-
-        speech = app.state.speech_synthesiser(
-            payload["input"],
-            forward_params={"speaker_embeddings": speaker_embedding},
-        )
-
-        sf.write(file_path, speech["audio"], samplerate=speech["sampling_rate"])
-        with open(file_body_path, "w") as f:
-            json.dump(json.loads(body.decode("utf-8")), f)
-
-        return FileResponse(file_path)
-
-
-def transcribe(file_path):
-    print("transcribe", file_path)
-    filename = os.path.basename(file_path)
-    file_dir = os.path.dirname(file_path)
-    id = filename.split(".")[0]
-
-    if app.state.config.STT_ENGINE == "":
-        if app.state.faster_whisper_model is None:
-            set_faster_whisper_model(app.state.config.WHISPER_MODEL)
-
-        model = app.state.faster_whisper_model
-        segments, info = model.transcribe(file_path, beam_size=5)
-        log.info(
-            "Detected language '%s' with probability %f"
-            % (info.language, info.language_probability)
-        )
-
-        transcript = "".join([segment.text for segment in list(segments)])
-        data = {"text": transcript.strip()}
-
-        # save the transcript to a json file
-        transcript_file = f"{file_dir}/{id}.json"
-        with open(transcript_file, "w") as f:
-            json.dump(data, f)
-
-        log.debug(data)
-        return data
-    elif app.state.config.STT_ENGINE == "openai":
-        if is_mp4_audio(file_path):
-            print("is_mp4_audio")
-            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)
-
-        headers = {"Authorization": f"Bearer {app.state.config.STT_OPENAI_API_KEY}"}
-
-        files = {"file": (filename, open(file_path, "rb"))}
-        data = {"model": app.state.config.STT_MODEL}
-
-        log.debug(files, data)
-
-        r = None
-        try:
-            r = requests.post(
-                url=f"{app.state.config.STT_OPENAI_API_BASE_URL}/audio/transcriptions",
-                headers=headers,
-                files=files,
-                data=data,
-            )
-
-            r.raise_for_status()
-
-            data = r.json()
-
-            # save the transcript to a json file
-            transcript_file = f"{file_dir}/{id}.json"
-            with open(transcript_file, "w") as f:
-                json.dump(data, f)
-
-            print(data)
-            return data
-        except Exception as e:
-            log.exception(e)
-            error_detail = "Open WebUI: Server Connection Error"
-            if r is not None:
-                try:
-                    res = r.json()
-                    if "error" in res:
-                        error_detail = f"External: {res['error']['message']}"
-                except Exception:
-                    error_detail = f"External: {e}"
-
-            raise Exception(error_detail)
-
-
-@app.post("/transcriptions")
-def transcription(
-    file: UploadFile = File(...),
-    user=Depends(get_verified_user),
-):
-    log.info(f"file.content_type: {file.content_type}")
-
-    if file.content_type not in ["audio/mpeg", "audio/wav", "audio/ogg", "audio/x-m4a"]:
-        raise HTTPException(
-            status_code=status.HTTP_400_BAD_REQUEST,
-            detail=ERROR_MESSAGES.FILE_NOT_SUPPORTED,
-        )
-
-    try:
-        ext = file.filename.split(".")[-1]
-        id = uuid.uuid4()
-
-        filename = f"{id}.{ext}"
-        contents = file.file.read()
-
-        file_dir = f"{CACHE_DIR}/audio/transcriptions"
-        os.makedirs(file_dir, exist_ok=True)
-        file_path = f"{file_dir}/{filename}"
-
-        with open(file_path, "wb") as f:
-            f.write(contents)
-
-        try:
-            if os.path.getsize(file_path) > MAX_FILE_SIZE:  # file is bigger than 25MB
-                log.debug(f"File size is larger than {MAX_FILE_SIZE_MB}MB")
-                audio = AudioSegment.from_file(file_path)
-                audio = audio.set_frame_rate(16000).set_channels(1)  # Compress audio
-                compressed_path = f"{file_dir}/{id}_compressed.opus"
-                audio.export(compressed_path, format="opus", bitrate="32k")
-                log.debug(f"Compressed audio to {compressed_path}")
-                file_path = compressed_path
-
-                if (
-                    os.path.getsize(file_path) > MAX_FILE_SIZE
-                ):  # Still larger than 25MB after compression
-                    log.debug(
-                        f"Compressed file size is still larger than {MAX_FILE_SIZE_MB}MB: {os.path.getsize(file_path)}"
-                    )
-                    raise HTTPException(
-                        status_code=status.HTTP_400_BAD_REQUEST,
-                        detail=ERROR_MESSAGES.FILE_TOO_LARGE(
-                            size=f"{MAX_FILE_SIZE_MB}MB"
-                        ),
-                    )
-
-                data = transcribe(file_path)
-            else:
-                data = transcribe(file_path)
-
-            file_path = file_path.split("/")[-1]
-            return {**data, "filename": file_path}
-        except Exception as e:
-            log.exception(e)
-            raise HTTPException(
-                status_code=status.HTTP_400_BAD_REQUEST,
-                detail=ERROR_MESSAGES.DEFAULT(e),
-            )
-
-    except Exception as e:
-        log.exception(e)
-
-        raise HTTPException(
-            status_code=status.HTTP_400_BAD_REQUEST,
-            detail=ERROR_MESSAGES.DEFAULT(e),
-        )
-
-
-def get_available_models() -> list[dict]:
-    if app.state.config.TTS_ENGINE == "openai":
-        return [{"id": "tts-1"}, {"id": "tts-1-hd"}]
-    elif app.state.config.TTS_ENGINE == "elevenlabs":
-        headers = {
-            "xi-api-key": app.state.config.TTS_API_KEY,
-            "Content-Type": "application/json",
-        }
-
-        try:
-            response = requests.get(
-                "https://api.elevenlabs.io/v1/models", headers=headers, timeout=5
-            )
-            response.raise_for_status()
-            models = response.json()
-            return [
-                {"name": model["name"], "id": model["model_id"]} for model in models
-            ]
-        except requests.RequestException as e:
-            log.error(f"Error fetching voices: {str(e)}")
-    return []
-
-
-@app.get("/models")
-async def get_models(user=Depends(get_verified_user)):
-    return {"models": get_available_models()}
-
-
-def get_available_voices() -> dict:
-    """Returns {voice_id: voice_name} dict"""
-    ret = {}
-    if app.state.config.TTS_ENGINE == "openai":
-        ret = {
-            "alloy": "alloy",
-            "echo": "echo",
-            "fable": "fable",
-            "onyx": "onyx",
-            "nova": "nova",
-            "shimmer": "shimmer",
-        }
-    elif app.state.config.TTS_ENGINE == "elevenlabs":
-        try:
-            ret = get_elevenlabs_voices()
-        except Exception:
-            # Avoided @lru_cache with exception
-            pass
-    elif app.state.config.TTS_ENGINE == "azure":
-        try:
-            region = app.state.config.TTS_AZURE_SPEECH_REGION
-            url = f"https://{region}.tts.speech.microsoft.com/cognitiveservices/voices/list"
-            headers = {"Ocp-Apim-Subscription-Key": app.state.config.TTS_API_KEY}
-
-            response = requests.get(url, headers=headers)
-            response.raise_for_status()
-            voices = response.json()
-            for voice in voices:
-                ret[voice["ShortName"]] = (
-                    f"{voice['DisplayName']} ({voice['ShortName']})"
-                )
-        except requests.RequestException as e:
-            log.error(f"Error fetching voices: {str(e)}")
-
-    return ret
-
-
-@lru_cache
-def get_elevenlabs_voices() -> dict:
-    """
-    Note, set the following in your .env file to use Elevenlabs:
-    AUDIO_TTS_ENGINE=elevenlabs
-    AUDIO_TTS_API_KEY=sk_...  # Your Elevenlabs API key
-    AUDIO_TTS_VOICE=EXAVITQu4vr4xnSDxMaL  # From https://api.elevenlabs.io/v1/voices
-    AUDIO_TTS_MODEL=eleven_multilingual_v2
-    """
-    headers = {
-        "xi-api-key": app.state.config.TTS_API_KEY,
-        "Content-Type": "application/json",
-    }
-    try:
-        # TODO: Add retries
-        response = requests.get("https://api.elevenlabs.io/v1/voices", headers=headers)
-        response.raise_for_status()
-        voices_data = response.json()
-
-        voices = {}
-        for voice in voices_data.get("voices", []):
-            voices[voice["voice_id"]] = voice["name"]
-    except requests.RequestException as e:
-        # Avoid @lru_cache with exception
-        log.error(f"Error fetching voices: {str(e)}")
-        raise RuntimeError(f"Error fetching voices: {str(e)}")
-
-    return voices
-
-
-@app.get("/voices")
-async def get_voices(user=Depends(get_verified_user)):
-    return {"voices": [{"id": k, "name": v} for k, v in get_available_voices().items()]}

backend/open_webui/apps/images/main.py

@@ -1,609 +0,0 @@
-import asyncio
-import base64
-import json
-import logging
-import mimetypes
-import re
-import uuid
-from pathlib import Path
-from typing import Optional
-
-import requests
-from open_webui.apps.images.utils.comfyui import (
-    ComfyUIGenerateImageForm,
-    ComfyUIWorkflow,
-    comfyui_generate_image,
-)
-from open_webui.config import (
-    AUTOMATIC1111_API_AUTH,
-    AUTOMATIC1111_BASE_URL,
-    AUTOMATIC1111_CFG_SCALE,
-    AUTOMATIC1111_SAMPLER,
-    AUTOMATIC1111_SCHEDULER,
-    CACHE_DIR,
-    COMFYUI_BASE_URL,
-    COMFYUI_WORKFLOW,
-    COMFYUI_WORKFLOW_NODES,
-    CORS_ALLOW_ORIGIN,
-    ENABLE_IMAGE_GENERATION,
-    IMAGE_GENERATION_ENGINE,
-    IMAGE_GENERATION_MODEL,
-    IMAGE_SIZE,
-    IMAGE_STEPS,
-    IMAGES_OPENAI_API_BASE_URL,
-    IMAGES_OPENAI_API_KEY,
-    AppConfig,
-)
-from open_webui.constants import ERROR_MESSAGES
-from open_webui.env import ENV, SRC_LOG_LEVELS, ENABLE_FORWARD_USER_INFO_HEADERS
-
-from fastapi import Depends, FastAPI, HTTPException, Request
-from fastapi.middleware.cors import CORSMiddleware
-from pydantic import BaseModel
-from open_webui.utils.utils import get_admin_user, get_verified_user
-
-log = logging.getLogger(__name__)
-log.setLevel(SRC_LOG_LEVELS["IMAGES"])
-
-IMAGE_CACHE_DIR = Path(CACHE_DIR).joinpath("./image/generations/")
-IMAGE_CACHE_DIR.mkdir(parents=True, exist_ok=True)
-
-app = FastAPI(
-    docs_url="/docs" if ENV == "dev" else None,
-    openapi_url="/openapi.json" if ENV == "dev" else None,
-    redoc_url=None,
-)
-
-app.add_middleware(
-    CORSMiddleware,
-    allow_origins=CORS_ALLOW_ORIGIN,
-    allow_credentials=True,
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
-
-app.state.config = AppConfig()
-
-app.state.config.ENGINE = IMAGE_GENERATION_ENGINE
-app.state.config.ENABLED = ENABLE_IMAGE_GENERATION
-
-app.state.config.OPENAI_API_BASE_URL = IMAGES_OPENAI_API_BASE_URL
-app.state.config.OPENAI_API_KEY = IMAGES_OPENAI_API_KEY
-
-app.state.config.MODEL = IMAGE_GENERATION_MODEL
-
-app.state.config.AUTOMATIC1111_BASE_URL = AUTOMATIC1111_BASE_URL
-app.state.config.AUTOMATIC1111_API_AUTH = AUTOMATIC1111_API_AUTH
-app.state.config.AUTOMATIC1111_CFG_SCALE = AUTOMATIC1111_CFG_SCALE
-app.state.config.AUTOMATIC1111_SAMPLER = AUTOMATIC1111_SAMPLER
-app.state.config.AUTOMATIC1111_SCHEDULER = AUTOMATIC1111_SCHEDULER
-app.state.config.COMFYUI_BASE_URL = COMFYUI_BASE_URL
-app.state.config.COMFYUI_WORKFLOW = COMFYUI_WORKFLOW
-app.state.config.COMFYUI_WORKFLOW_NODES = COMFYUI_WORKFLOW_NODES
-
-app.state.config.IMAGE_SIZE = IMAGE_SIZE
-app.state.config.IMAGE_STEPS = IMAGE_STEPS
-
-
-@app.get("/config")
-async def get_config(request: Request, user=Depends(get_admin_user)):
-    return {
-        "enabled": app.state.config.ENABLED,
-        "engine": app.state.config.ENGINE,
-        "openai": {
-            "OPENAI_API_BASE_URL": app.state.config.OPENAI_API_BASE_URL,
-            "OPENAI_API_KEY": app.state.config.OPENAI_API_KEY,
-        },
-        "automatic1111": {
-            "AUTOMATIC1111_BASE_URL": app.state.config.AUTOMATIC1111_BASE_URL,
-            "AUTOMATIC1111_API_AUTH": app.state.config.AUTOMATIC1111_API_AUTH,
-            "AUTOMATIC1111_CFG_SCALE": app.state.config.AUTOMATIC1111_CFG_SCALE,
-            "AUTOMATIC1111_SAMPLER": app.state.config.AUTOMATIC1111_SAMPLER,
-            "AUTOMATIC1111_SCHEDULER": app.state.config.AUTOMATIC1111_SCHEDULER,
-        },
-        "comfyui": {
-            "COMFYUI_BASE_URL": app.state.config.COMFYUI_BASE_URL,
-            "COMFYUI_WORKFLOW": app.state.config.COMFYUI_WORKFLOW,
-            "COMFYUI_WORKFLOW_NODES": app.state.config.COMFYUI_WORKFLOW_NODES,
-        },
-    }
-
-
-class OpenAIConfigForm(BaseModel):
-    OPENAI_API_BASE_URL: str
-    OPENAI_API_KEY: str
-
-
-class Automatic1111ConfigForm(BaseModel):
-    AUTOMATIC1111_BASE_URL: str
-    AUTOMATIC1111_API_AUTH: str
-    AUTOMATIC1111_CFG_SCALE: Optional[str]
-    AUTOMATIC1111_SAMPLER: Optional[str]
-    AUTOMATIC1111_SCHEDULER: Optional[str]
-
-
-class ComfyUIConfigForm(BaseModel):
-    COMFYUI_BASE_URL: str
-    COMFYUI_WORKFLOW: str
-    COMFYUI_WORKFLOW_NODES: list[dict]
-
-
-class ConfigForm(BaseModel):
-    enabled: bool
-    engine: str
-    openai: OpenAIConfigForm
-    automatic1111: Automatic1111ConfigForm
-    comfyui: ComfyUIConfigForm
-
-
-@app.post("/config/update")
-async def update_config(form_data: ConfigForm, user=Depends(get_admin_user)):
-    app.state.config.ENGINE = form_data.engine
-    app.state.config.ENABLED = form_data.enabled
-
-    app.state.config.OPENAI_API_BASE_URL = form_data.openai.OPENAI_API_BASE_URL
-    app.state.config.OPENAI_API_KEY = form_data.openai.OPENAI_API_KEY
-
-    app.state.config.AUTOMATIC1111_BASE_URL = (
-        form_data.automatic1111.AUTOMATIC1111_BASE_URL
-    )
-    app.state.config.AUTOMATIC1111_API_AUTH = (
-        form_data.automatic1111.AUTOMATIC1111_API_AUTH
-    )
-
-    app.state.config.AUTOMATIC1111_CFG_SCALE = (
-        float(form_data.automatic1111.AUTOMATIC1111_CFG_SCALE)
-        if form_data.automatic1111.AUTOMATIC1111_CFG_SCALE
-        else None
-    )
-    app.state.config.AUTOMATIC1111_SAMPLER = (
-        form_data.automatic1111.AUTOMATIC1111_SAMPLER
-        if form_data.automatic1111.AUTOMATIC1111_SAMPLER
-        else None
-    )
-    app.state.config.AUTOMATIC1111_SCHEDULER = (
-        form_data.automatic1111.AUTOMATIC1111_SCHEDULER
-        if form_data.automatic1111.AUTOMATIC1111_SCHEDULER
-        else None
-    )
-
-    app.state.config.COMFYUI_BASE_URL = form_data.comfyui.COMFYUI_BASE_URL.strip("/")
-    app.state.config.COMFYUI_WORKFLOW = form_data.comfyui.COMFYUI_WORKFLOW
-    app.state.config.COMFYUI_WORKFLOW_NODES = form_data.comfyui.COMFYUI_WORKFLOW_NODES
-
-    return {
-        "enabled": app.state.config.ENABLED,
-        "engine": app.state.config.ENGINE,
-        "openai": {
-            "OPENAI_API_BASE_URL": app.state.config.OPENAI_API_BASE_URL,
-            "OPENAI_API_KEY": app.state.config.OPENAI_API_KEY,
-        },
-        "automatic1111": {
-            "AUTOMATIC1111_BASE_URL": app.state.config.AUTOMATIC1111_BASE_URL,
-            "AUTOMATIC1111_API_AUTH": app.state.config.AUTOMATIC1111_API_AUTH,
-            "AUTOMATIC1111_CFG_SCALE": app.state.config.AUTOMATIC1111_CFG_SCALE,
-            "AUTOMATIC1111_SAMPLER": app.state.config.AUTOMATIC1111_SAMPLER,
-            "AUTOMATIC1111_SCHEDULER": app.state.config.AUTOMATIC1111_SCHEDULER,
-        },
-        "comfyui": {
-            "COMFYUI_BASE_URL": app.state.config.COMFYUI_BASE_URL,
-            "COMFYUI_WORKFLOW": app.state.config.COMFYUI_WORKFLOW,
-            "COMFYUI_WORKFLOW_NODES": app.state.config.COMFYUI_WORKFLOW_NODES,
-        },
-    }
-
-
-def get_automatic1111_api_auth():
-    if app.state.config.AUTOMATIC1111_API_AUTH is None:
-        return ""
-    else:
-        auth1111_byte_string = app.state.config.AUTOMATIC1111_API_AUTH.encode("utf-8")
-        auth1111_base64_encoded_bytes = base64.b64encode(auth1111_byte_string)
-        auth1111_base64_encoded_string = auth1111_base64_encoded_bytes.decode("utf-8")
-        return f"Basic {auth1111_base64_encoded_string}"
-
-
-@app.get("/config/url/verify")
-async def verify_url(user=Depends(get_admin_user)):
-    if app.state.config.ENGINE == "automatic1111":
-        try:
-            r = requests.get(
-                url=f"{app.state.config.AUTOMATIC1111_BASE_URL}/sdapi/v1/options",
-                headers={"authorization": get_automatic1111_api_auth()},
-            )
-            r.raise_for_status()
-            return True
-        except Exception:
-            app.state.config.ENABLED = False
-            raise HTTPException(status_code=400, detail=ERROR_MESSAGES.INVALID_URL)
-    elif app.state.config.ENGINE == "comfyui":
-        try:
-            r = requests.get(url=f"{app.state.config.COMFYUI_BASE_URL}/object_info")
-            r.raise_for_status()
-            return True
-        except Exception:
-            app.state.config.ENABLED = False
-            raise HTTPException(status_code=400, detail=ERROR_MESSAGES.INVALID_URL)
-    else:
-        return True
-
-
-def set_image_model(model: str):
-    log.info(f"Setting image model to {model}")
-    app.state.config.MODEL = model
-    if app.state.config.ENGINE in ["", "automatic1111"]:
-        api_auth = get_automatic1111_api_auth()
-        r = requests.get(
-            url=f"{app.state.config.AUTOMATIC1111_BASE_URL}/sdapi/v1/options",
-            headers={"authorization": api_auth},
-        )
-        options = r.json()
-        if model != options["sd_model_checkpoint"]:
-            options["sd_model_checkpoint"] = model
-            r = requests.post(
-                url=f"{app.state.config.AUTOMATIC1111_BASE_URL}/sdapi/v1/options",
-                json=options,
-                headers={"authorization": api_auth},
-            )
-    return app.state.config.MODEL
-
-
-def get_image_model():
-    if app.state.config.ENGINE == "openai":
-        return app.state.config.MODEL if app.state.config.MODEL else "dall-e-2"
-    elif app.state.config.ENGINE == "comfyui":
-        return app.state.config.MODEL if app.state.config.MODEL else ""
-    elif app.state.config.ENGINE == "automatic1111" or app.state.config.ENGINE == "":
-        try:
-            r = requests.get(
-                url=f"{app.state.config.AUTOMATIC1111_BASE_URL}/sdapi/v1/options",
-                headers={"authorization": get_automatic1111_api_auth()},
-            )
-            options = r.json()
-            return options["sd_model_checkpoint"]
-        except Exception as e:
-            app.state.config.ENABLED = False
-            raise HTTPException(status_code=400, detail=ERROR_MESSAGES.DEFAULT(e))
-
-
-class ImageConfigForm(BaseModel):
-    MODEL: str
-    IMAGE_SIZE: str
-    IMAGE_STEPS: int
-
-
-@app.get("/image/config")
-async def get_image_config(user=Depends(get_admin_user)):
-    return {
-        "MODEL": app.state.config.MODEL,
-        "IMAGE_SIZE": app.state.config.IMAGE_SIZE,
-        "IMAGE_STEPS": app.state.config.IMAGE_STEPS,
-    }
-
-
-@app.post("/image/config/update")
-async def update_image_config(form_data: ImageConfigForm, user=Depends(get_admin_user)):
-
-    set_image_model(form_data.MODEL)
-
-    pattern = r"^\d+x\d+$"
-    if re.match(pattern, form_data.IMAGE_SIZE):
-        app.state.config.IMAGE_SIZE = form_data.IMAGE_SIZE
-    else:
-        raise HTTPException(
-            status_code=400,
-            detail=ERROR_MESSAGES.INCORRECT_FORMAT("  (e.g., 512x512)."),
-        )
-
-    if form_data.IMAGE_STEPS >= 0:
-        app.state.config.IMAGE_STEPS = form_data.IMAGE_STEPS
-    else:
-        raise HTTPException(
-            status_code=400,
-            detail=ERROR_MESSAGES.INCORRECT_FORMAT("  (e.g., 50)."),
-        )
-
-    return {
-        "MODEL": app.state.config.MODEL,
-        "IMAGE_SIZE": app.state.config.IMAGE_SIZE,
-        "IMAGE_STEPS": app.state.config.IMAGE_STEPS,
-    }
-
-
-@app.get("/models")
-def get_models(user=Depends(get_verified_user)):
-    try:
-        if app.state.config.ENGINE == "openai":
-            return [
-                {"id": "dall-e-2", "name": "DALL·E 2"},
-                {"id": "dall-e-3", "name": "DALL·E 3"},
-            ]
-        elif app.state.config.ENGINE == "comfyui":
-            # TODO - get models from comfyui
-            r = requests.get(url=f"{app.state.config.COMFYUI_BASE_URL}/object_info")
-            info = r.json()
-
-            workflow = json.loads(app.state.config.COMFYUI_WORKFLOW)
-            model_node_id = None
-
-            for node in app.state.config.COMFYUI_WORKFLOW_NODES:
-                if node["type"] == "model":
-                    if node["node_ids"]:
-                        model_node_id = node["node_ids"][0]
-                    break
-
-            if model_node_id:
-                model_list_key = None
-
-                print(workflow[model_node_id]["class_type"])
-                for key in info[workflow[model_node_id]["class_type"]]["input"][
-                    "required"
-                ]:
-                    if "_name" in key:
-                        model_list_key = key
-                        break
-
-                if model_list_key:
-                    return list(
-                        map(
-                            lambda model: {"id": model, "name": model},
-                            info[workflow[model_node_id]["class_type"]]["input"][
-                                "required"
-                            ][model_list_key][0],
-                        )
-                    )
-            else:
-                return list(
-                    map(
-                        lambda model: {"id": model, "name": model},
-                        info["CheckpointLoaderSimple"]["input"]["required"][
-                            "ckpt_name"
-                        ][0],
-                    )
-                )
-        elif (
-            app.state.config.ENGINE == "automatic1111" or app.state.config.ENGINE == ""
-        ):
-            r = requests.get(
-                url=f"{app.state.config.AUTOMATIC1111_BASE_URL}/sdapi/v1/sd-models",
-                headers={"authorization": get_automatic1111_api_auth()},
-            )
-            models = r.json()
-            return list(
-                map(
-                    lambda model: {"id": model["title"], "name": model["model_name"]},
-                    models,
-                )
-            )
-    except Exception as e:
-        app.state.config.ENABLED = False
-        raise HTTPException(status_code=400, detail=ERROR_MESSAGES.DEFAULT(e))
-
-
-class GenerateImageForm(BaseModel):
-    model: Optional[str] = None
-    prompt: str
-    size: Optional[str] = None
-    n: int = 1
-    negative_prompt: Optional[str] = None
-
-
-def save_b64_image(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)
-
-            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
-
-    except Exception as e:
-        log.exception(f"Error saving image: {e}")
-        return None
-
-
-def save_url_image(url):
-    image_id = str(uuid.uuid4())
-    try:
-        r = requests.get(url)
-        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
-        else:
-            log.error("Url does not point to an image.")
-            return None
-
-    except Exception as e:
-        log.exception(f"Error saving image: {e}")
-        return None
-
-
-@app.post("/generations")
-async def image_generations(
-    form_data: GenerateImageForm,
-    user=Depends(get_verified_user),
-):
-    width, height = tuple(map(int, app.state.config.IMAGE_SIZE.split("x")))
-
-    r = None
-    try:
-        if app.state.config.ENGINE == "openai":
-            headers = {}
-            headers["Authorization"] = f"Bearer {app.state.config.OPENAI_API_KEY}"
-            headers["Content-Type"] = "application/json"
-
-            if ENABLE_FORWARD_USER_INFO_HEADERS:
-                headers["X-OpenWebUI-User-Name"] = user.name
-                headers["X-OpenWebUI-User-Id"] = user.id
-                headers["X-OpenWebUI-User-Email"] = user.email
-                headers["X-OpenWebUI-User-Role"] = user.role
-
-            data = {
-                "model": (
-                    app.state.config.MODEL
-                    if app.state.config.MODEL != ""
-                    else "dall-e-2"
-                ),
-                "prompt": form_data.prompt,
-                "n": form_data.n,
-                "size": (
-                    form_data.size if form_data.size else app.state.config.IMAGE_SIZE
-                ),
-                "response_format": "b64_json",
-            }
-
-            # Use asyncio.to_thread for the requests.post call
-            r = await asyncio.to_thread(
-                requests.post,
-                url=f"{app.state.config.OPENAI_API_BASE_URL}/images/generations",
-                json=data,
-                headers=headers,
-            )
-
-            r.raise_for_status()
-            res = r.json()
-
-            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")
-
-                with open(file_body_path, "w") as f:
-                    json.dump(data, f)
-
-            return images
-
-        elif app.state.config.ENGINE == "comfyui":
-            data = {
-                "prompt": form_data.prompt,
-                "width": width,
-                "height": height,
-                "n": form_data.n,
-            }
-
-            if app.state.config.IMAGE_STEPS is not None:
-                data["steps"] = app.state.config.IMAGE_STEPS
-
-            if form_data.negative_prompt is not None:
-                data["negative_prompt"] = form_data.negative_prompt
-
-            form_data = ComfyUIGenerateImageForm(
-                **{
-                    "workflow": ComfyUIWorkflow(
-                        **{
-                            "workflow": app.state.config.COMFYUI_WORKFLOW,
-                            "nodes": app.state.config.COMFYUI_WORKFLOW_NODES,
-                        }
-                    ),
-                    **data,
-                }
-            )
-            res = await comfyui_generate_image(
-                app.state.config.MODEL,
-                form_data,
-                user.id,
-                app.state.config.COMFYUI_BASE_URL,
-            )
-            log.debug(f"res: {res}")
-
-            images = []
-
-            for image in res["data"]:
-                image_filename = save_url_image(image["url"])
-                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}")
-            return images
-        elif (
-            app.state.config.ENGINE == "automatic1111" or app.state.config.ENGINE == ""
-        ):
-            if form_data.model:
-                set_image_model(form_data.model)
-
-            data = {
-                "prompt": form_data.prompt,
-                "batch_size": form_data.n,
-                "width": width,
-                "height": height,
-            }
-
-            if app.state.config.IMAGE_STEPS is not None:
-                data["steps"] = app.state.config.IMAGE_STEPS
-
-            if form_data.negative_prompt is not None:
-                data["negative_prompt"] = form_data.negative_prompt
-
-            if app.state.config.AUTOMATIC1111_CFG_SCALE:
-                data["cfg_scale"] = app.state.config.AUTOMATIC1111_CFG_SCALE
-
-            if app.state.config.AUTOMATIC1111_SAMPLER:
-                data["sampler_name"] = app.state.config.AUTOMATIC1111_SAMPLER
-
-            if app.state.config.AUTOMATIC1111_SCHEDULER:
-                data["scheduler"] = app.state.config.AUTOMATIC1111_SCHEDULER
-
-            # Use asyncio.to_thread for the requests.post call
-            r = await asyncio.to_thread(
-                requests.post,
-                url=f"{app.state.config.AUTOMATIC1111_BASE_URL}/sdapi/v1/txt2img",
-                json=data,
-                headers={"authorization": get_automatic1111_api_auth()},
-            )
-
-            res = r.json()
-            log.debug(f"res: {res}")
-
-            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)
-
-            return images
-    except Exception as e:
-        error = e
-        if r != None:
-            data = r.json()
-            if "error" in data:
-                error = data["error"]["message"]
-        raise HTTPException(status_code=400, detail=ERROR_MESSAGES.DEFAULT(error))

backend/open_webui/apps/openai/main.py

@@ -1,719 +0,0 @@
-import asyncio
-import hashlib
-import json
-import logging
-from pathlib import Path
-from typing import Literal, Optional, overload
-
-import aiohttp
-from aiocache import cached
-import requests
-
-
-from open_webui.apps.webui.models.models import Models
-from open_webui.config import (
-    CACHE_DIR,
-    CORS_ALLOW_ORIGIN,
-    ENABLE_OPENAI_API,
-    OPENAI_API_BASE_URLS,
-    OPENAI_API_KEYS,
-    OPENAI_API_CONFIGS,
-    AppConfig,
-)
-from open_webui.env import (
-    AIOHTTP_CLIENT_TIMEOUT,
-    AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST,
-    ENABLE_FORWARD_USER_INFO_HEADERS,
-    BYPASS_MODEL_ACCESS_CONTROL,
-)
-
-from open_webui.constants import ERROR_MESSAGES
-from open_webui.env import ENV, SRC_LOG_LEVELS
-from fastapi import Depends, FastAPI, HTTPException, Request
-from fastapi.middleware.cors import CORSMiddleware
-from fastapi.responses import FileResponse, StreamingResponse
-from pydantic import BaseModel
-from starlette.background import BackgroundTask
-
-from open_webui.utils.payload import (
-    apply_model_params_to_body_openai,
-    apply_model_system_prompt_to_body,
-)
-
-from open_webui.utils.utils import get_admin_user, get_verified_user
-from open_webui.utils.access_control import has_access
-
-
-log = logging.getLogger(__name__)
-log.setLevel(SRC_LOG_LEVELS["OPENAI"])
-
-
-app = FastAPI(
-    docs_url="/docs" if ENV == "dev" else None,
-    openapi_url="/openapi.json" if ENV == "dev" else None,
-    redoc_url=None,
-)
-
-
-app.add_middleware(
-    CORSMiddleware,
-    allow_origins=CORS_ALLOW_ORIGIN,
-    allow_credentials=True,
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
-
-app.state.config = AppConfig()
-
-app.state.config.ENABLE_OPENAI_API = ENABLE_OPENAI_API
-app.state.config.OPENAI_API_BASE_URLS = OPENAI_API_BASE_URLS
-app.state.config.OPENAI_API_KEYS = OPENAI_API_KEYS
-app.state.config.OPENAI_API_CONFIGS = OPENAI_API_CONFIGS
-
-
-@app.get("/config")
-async def get_config(user=Depends(get_admin_user)):
-    return {
-        "ENABLE_OPENAI_API": app.state.config.ENABLE_OPENAI_API,
-        "OPENAI_API_BASE_URLS": app.state.config.OPENAI_API_BASE_URLS,
-        "OPENAI_API_KEYS": app.state.config.OPENAI_API_KEYS,
-        "OPENAI_API_CONFIGS": app.state.config.OPENAI_API_CONFIGS,
-    }
-
-
-class OpenAIConfigForm(BaseModel):
-    ENABLE_OPENAI_API: Optional[bool] = None
-    OPENAI_API_BASE_URLS: list[str]
-    OPENAI_API_KEYS: list[str]
-    OPENAI_API_CONFIGS: dict
-
-
-@app.post("/config/update")
-async def update_config(form_data: OpenAIConfigForm, user=Depends(get_admin_user)):
-    app.state.config.ENABLE_OPENAI_API = form_data.ENABLE_OPENAI_API
-
-    app.state.config.OPENAI_API_BASE_URLS = form_data.OPENAI_API_BASE_URLS
-    app.state.config.OPENAI_API_KEYS = form_data.OPENAI_API_KEYS
-
-    # Check if API KEYS length is same than API URLS length
-    if len(app.state.config.OPENAI_API_KEYS) != len(
-        app.state.config.OPENAI_API_BASE_URLS
-    ):
-        if len(app.state.config.OPENAI_API_KEYS) > len(
-            app.state.config.OPENAI_API_BASE_URLS
-        ):
-            app.state.config.OPENAI_API_KEYS = app.state.config.OPENAI_API_KEYS[
-                : len(app.state.config.OPENAI_API_BASE_URLS)
-            ]
-        else:
-            app.state.config.OPENAI_API_KEYS += [""] * (
-                len(app.state.config.OPENAI_API_BASE_URLS)
-                - len(app.state.config.OPENAI_API_KEYS)
-            )
-
-    app.state.config.OPENAI_API_CONFIGS = form_data.OPENAI_API_CONFIGS
-
-    # Remove any extra configs
-    config_urls = app.state.config.OPENAI_API_CONFIGS.keys()
-    for idx, url in enumerate(app.state.config.OPENAI_API_BASE_URLS):
-        if url not in config_urls:
-            app.state.config.OPENAI_API_CONFIGS.pop(url, None)
-
-    return {
-        "ENABLE_OPENAI_API": app.state.config.ENABLE_OPENAI_API,
-        "OPENAI_API_BASE_URLS": app.state.config.OPENAI_API_BASE_URLS,
-        "OPENAI_API_KEYS": app.state.config.OPENAI_API_KEYS,
-        "OPENAI_API_CONFIGS": app.state.config.OPENAI_API_CONFIGS,
-    }
-
-
-@app.post("/audio/speech")
-async def speech(request: Request, user=Depends(get_verified_user)):
-    idx = None
-    try:
-        idx = app.state.config.OPENAI_API_BASE_URLS.index("https://api.openai.com/v1")
-        body = await request.body()
-        name = hashlib.sha256(body).hexdigest()
-
-        SPEECH_CACHE_DIR = Path(CACHE_DIR).joinpath("./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")
-
-        # Check if the file already exists in the cache
-        if file_path.is_file():
-            return FileResponse(file_path)
-
-        headers = {}
-        headers["Authorization"] = f"Bearer {app.state.config.OPENAI_API_KEYS[idx]}"
-        headers["Content-Type"] = "application/json"
-        if "openrouter.ai" in app.state.config.OPENAI_API_BASE_URLS[idx]:
-            headers["HTTP-Referer"] = "https://openwebui.com/"
-            headers["X-Title"] = "Open WebUI"
-        if ENABLE_FORWARD_USER_INFO_HEADERS:
-            headers["X-OpenWebUI-User-Name"] = user.name
-            headers["X-OpenWebUI-User-Id"] = user.id
-            headers["X-OpenWebUI-User-Email"] = user.email
-            headers["X-OpenWebUI-User-Role"] = user.role
-        r = None
-        try:
-            r = requests.post(
-                url=f"{app.state.config.OPENAI_API_BASE_URLS[idx]}/audio/speech",
-                data=body,
-                headers=headers,
-                stream=True,
-            )
-
-            r.raise_for_status()
-
-            # Save the streaming content to a file
-            with open(file_path, "wb") as f:
-                for chunk in r.iter_content(chunk_size=8192):
-                    f.write(chunk)
-
-            with open(file_body_path, "w") as f:
-                json.dump(json.loads(body.decode("utf-8")), f)
-
-            # Return the saved file
-            return FileResponse(file_path)
-
-        except Exception as e:
-            log.exception(e)
-            error_detail = "Open WebUI: Server Connection Error"
-            if r is not None:
-                try:
-                    res = r.json()
-                    if "error" in res:
-                        error_detail = f"External: {res['error']}"
-                except Exception:
-                    error_detail = f"External: {e}"
-
-            raise HTTPException(
-                status_code=r.status_code if r else 500, detail=error_detail
-            )
-
-    except ValueError:
-        raise HTTPException(status_code=401, detail=ERROR_MESSAGES.OPENAI_NOT_FOUND)
-
-
-async def aiohttp_get(url, key=None):
-    timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST)
-    try:
-        headers = {"Authorization": f"Bearer {key}"} if key else {}
-        async with aiohttp.ClientSession(timeout=timeout, trust_env=True) as session:
-            async with session.get(url, headers=headers) as response:
-                return await response.json()
-    except Exception as e:
-        # Handle connection error here
-        log.error(f"Connection error: {e}")
-        return None
-
-
-async def cleanup_response(
-    response: Optional[aiohttp.ClientResponse],
-    session: Optional[aiohttp.ClientSession],
-):
-    if response:
-        response.close()
-    if session:
-        await session.close()
-
-
-def merge_models_lists(model_lists):
-    log.debug(f"merge_models_lists {model_lists}")
-    merged_list = []
-
-    for idx, models in enumerate(model_lists):
-        if models is not None and "error" not in models:
-            merged_list.extend(
-                [
-                    {
-                        **model,
-                        "name": model.get("name", model["id"]),
-                        "owned_by": "openai",
-                        "openai": model,
-                        "urlIdx": idx,
-                    }
-                    for model in models
-                    if "api.openai.com"
-                    not in 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",
-                        ]
-                    )
-                ]
-            )
-
-    return merged_list
-
-
-async def get_all_models_responses() -> list:
-    if not app.state.config.ENABLE_OPENAI_API:
-        return []
-
-    # Check if API KEYS length is same than API URLS length
-    num_urls = len(app.state.config.OPENAI_API_BASE_URLS)
-    num_keys = len(app.state.config.OPENAI_API_KEYS)
-
-    if num_keys != num_urls:
-        # if there are more keys than urls, remove the extra keys
-        if num_keys > num_urls:
-            new_keys = app.state.config.OPENAI_API_KEYS[:num_urls]
-            app.state.config.OPENAI_API_KEYS = new_keys
-        # if there are more urls than keys, add empty keys
-        else:
-            app.state.config.OPENAI_API_KEYS += [""] * (num_urls - num_keys)
-
-    tasks = []
-    for idx, url in enumerate(app.state.config.OPENAI_API_BASE_URLS):
-        if url not in app.state.config.OPENAI_API_CONFIGS:
-            tasks.append(
-                aiohttp_get(f"{url}/models", app.state.config.OPENAI_API_KEYS[idx])
-            )
-        else:
-            api_config = app.state.config.OPENAI_API_CONFIGS.get(url, {})
-
-            enable = api_config.get("enable", True)
-            model_ids = api_config.get("model_ids", [])
-
-            if enable:
-                if len(model_ids) == 0:
-                    tasks.append(
-                        aiohttp_get(
-                            f"{url}/models", app.state.config.OPENAI_API_KEYS[idx]
-                        )
-                    )
-                else:
-                    model_list = {
-                        "object": "list",
-                        "data": [
-                            {
-                                "id": model_id,
-                                "name": model_id,
-                                "owned_by": "openai",
-                                "openai": {"id": model_id},
-                                "urlIdx": idx,
-                            }
-                            for model_id in model_ids
-                        ],
-                    }
-
-                    tasks.append(asyncio.ensure_future(asyncio.sleep(0, model_list)))
-            else:
-                tasks.append(asyncio.ensure_future(asyncio.sleep(0, None)))
-
-    responses = await asyncio.gather(*tasks)
-
-    for idx, response in enumerate(responses):
-        if response:
-            url = app.state.config.OPENAI_API_BASE_URLS[idx]
-            api_config = app.state.config.OPENAI_API_CONFIGS.get(url, {})
-
-            prefix_id = api_config.get("prefix_id", None)
-
-            if prefix_id:
-                for model in (
-                    response if isinstance(response, list) else response.get("data", [])
-                ):
-                    model["id"] = f"{prefix_id}.{model['id']}"
-
-    log.debug(f"get_all_models:responses() {responses}")
-
-    return responses
-
-
-@cached(ttl=3)
-async def get_all_models() -> dict[str, list]:
-    log.info("get_all_models()")
-
-    if not app.state.config.ENABLE_OPENAI_API:
-        return {"data": []}
-
-    responses = await get_all_models_responses()
-
-    def extract_data(response):
-        if response and "data" in response:
-            return response["data"]
-        if isinstance(response, list):
-            return response
-        return None
-
-    models = {"data": merge_models_lists(map(extract_data, responses))}
-    log.debug(f"models: {models}")
-
-    return models
-
-
-@app.get("/models")
-@app.get("/models/{url_idx}")
-async def get_models(url_idx: Optional[int] = None, user=Depends(get_verified_user)):
-    models = {
-        "data": [],
-    }
-
-    if url_idx is None:
-        models = await get_all_models()
-    else:
-        url = app.state.config.OPENAI_API_BASE_URLS[url_idx]
-        key = app.state.config.OPENAI_API_KEYS[url_idx]
-
-        headers = {}
-        headers["Authorization"] = f"Bearer {key}"
-        headers["Content-Type"] = "application/json"
-
-        if ENABLE_FORWARD_USER_INFO_HEADERS:
-            headers["X-OpenWebUI-User-Name"] = user.name
-            headers["X-OpenWebUI-User-Id"] = user.id
-            headers["X-OpenWebUI-User-Email"] = user.email
-            headers["X-OpenWebUI-User-Role"] = user.role
-
-        r = None
-
-        timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST)
-        async with aiohttp.ClientSession(timeout=timeout) as session:
-            try:
-                async with session.get(f"{url}/models", headers=headers) as r:
-                    if r.status != 200:
-                        # Extract response error details if available
-                        error_detail = f"HTTP Error: {r.status}"
-                        res = await r.json()
-                        if "error" in res:
-                            error_detail = f"External Error: {res['error']}"
-                        raise Exception(error_detail)
-
-                    response_data = await r.json()
-
-                    # Check if we're calling OpenAI API based on the URL
-                    if "api.openai.com" in url:
-                        # Filter models according to the specified conditions
-                        response_data["data"] = [
-                            model
-                            for model in response_data.get("data", [])
-                            if not any(
-                                name in model["id"]
-                                for name in [
-                                    "babbage",
-                                    "dall-e",
-                                    "davinci",
-                                    "embedding",
-                                    "tts",
-                                    "whisper",
-                                ]
-                            )
-                        ]
-
-                    models = response_data
-            except aiohttp.ClientError as e:
-                # ClientError covers all aiohttp requests issues
-                log.exception(f"Client error: {str(e)}")
-                # Handle aiohttp-specific connection issues, timeout etc.
-                raise HTTPException(
-                    status_code=500, detail="Open WebUI: Server Connection Error"
-                )
-            except Exception as e:
-                log.exception(f"Unexpected error: {e}")
-                # Generic error handler in case parsing JSON or other steps fail
-                error_detail = f"Unexpected error: {str(e)}"
-                raise HTTPException(status_code=500, detail=error_detail)
-
-    if user.role == "user" and not BYPASS_MODEL_ACCESS_CONTROL:
-        # Filter models based on user access control
-        filtered_models = []
-        for model in models.get("data", []):
-            model_info = Models.get_model_by_id(model["id"])
-            if model_info:
-                if user.id == model_info.user_id or has_access(
-                    user.id, type="read", access_control=model_info.access_control
-                ):
-                    filtered_models.append(model)
-        models["data"] = filtered_models
-
-    return models
-
-
-class ConnectionVerificationForm(BaseModel):
-    url: str
-    key: str
-
-
-@app.post("/verify")
-async def verify_connection(
-    form_data: ConnectionVerificationForm, user=Depends(get_admin_user)
-):
-    url = form_data.url
-    key = form_data.key
-
-    headers = {}
-    headers["Authorization"] = f"Bearer {key}"
-    headers["Content-Type"] = "application/json"
-
-    timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST)
-    async with aiohttp.ClientSession(timeout=timeout) as session:
-        try:
-            async with session.get(f"{url}/models", headers=headers) as r:
-                if r.status != 200:
-                    # Extract response error details if available
-                    error_detail = f"HTTP Error: {r.status}"
-                    res = await r.json()
-                    if "error" in res:
-                        error_detail = f"External Error: {res['error']}"
-                    raise Exception(error_detail)
-
-                response_data = await r.json()
-                return response_data
-
-        except aiohttp.ClientError as e:
-            # ClientError covers all aiohttp requests issues
-            log.exception(f"Client error: {str(e)}")
-            # Handle aiohttp-specific connection issues, timeout etc.
-            raise HTTPException(
-                status_code=500, detail="Open WebUI: Server Connection Error"
-            )
-        except Exception as e:
-            log.exception(f"Unexpected error: {e}")
-            # Generic error handler in case parsing JSON or other steps fail
-            error_detail = f"Unexpected error: {str(e)}"
-            raise HTTPException(status_code=500, detail=error_detail)
-
-
-@app.post("/chat/completions")
-async def generate_chat_completion(
-    form_data: dict,
-    user=Depends(get_verified_user),
-    bypass_filter: Optional[bool] = False,
-):
-    idx = 0
-    payload = {**form_data}
-
-    if "metadata" in payload:
-        del payload["metadata"]
-
-    model_id = form_data.get("model")
-    model_info = Models.get_model_by_id(model_id)
-
-    # Check model info and override the payload
-    if model_info:
-        if model_info.base_model_id:
-            payload["model"] = model_info.base_model_id
-
-        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)
-
-        # Check if user has access to the model
-        if not bypass_filter and user.role == "user":
-            if not (
-                user.id == model_info.user_id
-                or has_access(
-                    user.id, type="read", access_control=model_info.access_control
-                )
-            ):
-                raise HTTPException(
-                    status_code=403,
-                    detail="Model not found",
-                )
-    elif not bypass_filter:
-        if user.role != "admin":
-            raise HTTPException(
-                status_code=403,
-                detail="Model not found",
-            )
-
-    # Attemp to get urlIdx from the model
-    models = await get_all_models()
-
-    # Find the model from the list
-    model = next(
-        (model for model in models["data"] if model["id"] == payload.get("model")),
-        None,
-    )
-
-    if model:
-        idx = model["urlIdx"]
-    else:
-        raise HTTPException(
-            status_code=404,
-            detail="Model not found",
-        )
-
-    # Get the API config for the model
-    api_config = app.state.config.OPENAI_API_CONFIGS.get(
-        app.state.config.OPENAI_API_BASE_URLS[idx], {}
-    )
-    prefix_id = api_config.get("prefix_id", None)
-
-    if prefix_id:
-        payload["model"] = payload["model"].replace(f"{prefix_id}.", "")
-
-    # Add user info to the payload if the model is a pipeline
-    if "pipeline" in model and model.get("pipeline"):
-        payload["user"] = {
-            "name": user.name,
-            "id": user.id,
-            "email": user.email,
-            "role": user.role,
-        }
-
-    url = app.state.config.OPENAI_API_BASE_URLS[idx]
-    key = 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-")
-    # Change max_completion_tokens to max_tokens (Backward compatible)
-    if "api.openai.com" not in url and not is_o1:
-        if "max_completion_tokens" in payload:
-            # Remove "max_completion_tokens" from the payload
-            payload["max_tokens"] = payload["max_completion_tokens"]
-            del payload["max_completion_tokens"]
-    else:
-        if is_o1 and "max_tokens" in payload:
-            payload["max_completion_tokens"] = payload["max_tokens"]
-            del payload["max_tokens"]
-        if "max_tokens" in payload and "max_completion_tokens" in payload:
-            del payload["max_tokens"]
-
-    # Fix: O1 does not support the "system" parameter, Modify "system" to "user"
-    if is_o1 and payload["messages"][0]["role"] == "system":
-        payload["messages"][0]["role"] = "user"
-
-    # Convert the modified body back to JSON
-    payload = json.dumps(payload)
-
-    headers = {}
-    headers["Authorization"] = f"Bearer {key}"
-    headers["Content-Type"] = "application/json"
-    if "openrouter.ai" in app.state.config.OPENAI_API_BASE_URLS[idx]:
-        headers["HTTP-Referer"] = "https://openwebui.com/"
-        headers["X-Title"] = "Open WebUI"
-    if ENABLE_FORWARD_USER_INFO_HEADERS:
-        headers["X-OpenWebUI-User-Name"] = user.name
-        headers["X-OpenWebUI-User-Id"] = user.id
-        headers["X-OpenWebUI-User-Email"] = user.email
-        headers["X-OpenWebUI-User-Role"] = user.role
-
-    r = None
-    session = None
-    streaming = False
-    response = None
-
-    try:
-        session = aiohttp.ClientSession(
-            trust_env=True, timeout=aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT)
-        )
-        r = await session.request(
-            method="POST",
-            url=f"{url}/chat/completions",
-            data=payload,
-            headers=headers,
-        )
-
-        # Check if response is SSE
-        if "text/event-stream" in r.headers.get("Content-Type", ""):
-            streaming = True
-            return StreamingResponse(
-                r.content,
-                status_code=r.status,
-                headers=dict(r.headers),
-                background=BackgroundTask(
-                    cleanup_response, response=r, session=session
-                ),
-            )
-        else:
-            try:
-                response = await r.json()
-            except Exception as e:
-                log.error(e)
-                response = await r.text()
-
-            r.raise_for_status()
-            return response
-    except Exception as e:
-        log.exception(e)
-        error_detail = "Open WebUI: Server Connection Error"
-        if isinstance(response, dict):
-            if "error" in response:
-                error_detail = f"{response['error']['message'] if 'message' in response['error'] else response['error']}"
-        elif isinstance(response, str):
-            error_detail = response
-
-        raise HTTPException(status_code=r.status if r else 500, detail=error_detail)
-    finally:
-        if not streaming and session:
-            if r:
-                r.close()
-            await session.close()
-
-
-@app.api_route("/{path:path}", methods=["GET", "POST", "PUT", "DELETE"])
-async def proxy(path: str, request: Request, user=Depends(get_verified_user)):
-    idx = 0
-
-    body = await request.body()
-
-    url = app.state.config.OPENAI_API_BASE_URLS[idx]
-    key = app.state.config.OPENAI_API_KEYS[idx]
-
-    target_url = f"{url}/{path}"
-
-    headers = {}
-    headers["Authorization"] = f"Bearer {key}"
-    headers["Content-Type"] = "application/json"
-    if ENABLE_FORWARD_USER_INFO_HEADERS:
-        headers["X-OpenWebUI-User-Name"] = user.name
-        headers["X-OpenWebUI-User-Id"] = user.id
-        headers["X-OpenWebUI-User-Email"] = user.email
-        headers["X-OpenWebUI-User-Role"] = user.role
-
-    r = None
-    session = None
-    streaming = False
-
-    try:
-        session = aiohttp.ClientSession(trust_env=True)
-        r = await session.request(
-            method=request.method,
-            url=target_url,
-            data=body,
-            headers=headers,
-        )
-
-        r.raise_for_status()
-
-        # Check if response is SSE
-        if "text/event-stream" in r.headers.get("Content-Type", ""):
-            streaming = True
-            return StreamingResponse(
-                r.content,
-                status_code=r.status,
-                headers=dict(r.headers),
-                background=BackgroundTask(
-                    cleanup_response, response=r, session=session
-                ),
-            )
-        else:
-            response_data = await r.json()
-            return response_data
-    except Exception as e:
-        log.exception(e)
-        error_detail = "Open WebUI: Server Connection Error"
-        if r is not None:
-            try:
-                res = await r.json()
-                print(res)
-                if "error" in res:
-                    error_detail = f"External: {res['error']['message'] if 'message' in res['error'] else res['error']}"
-            except Exception:
-                error_detail = f"External: {e}"
-        raise HTTPException(status_code=r.status if r else 500, detail=error_detail)
-    finally:
-        if not streaming and session:
-            if r:
-                r.close()
-            await session.close()

backend/open_webui/apps/retrieval/loaders/main.py

@@ -1,190 +0,0 @@
-import requests
-import logging
-import ftfy
-
-from langchain_community.document_loaders import (
-    BSHTMLLoader,
-    CSVLoader,
-    Docx2txtLoader,
-    OutlookMessageLoader,
-    PyPDFLoader,
-    TextLoader,
-    UnstructuredEPubLoader,
-    UnstructuredExcelLoader,
-    UnstructuredMarkdownLoader,
-    UnstructuredPowerPointLoader,
-    UnstructuredRSTLoader,
-    UnstructuredXMLLoader,
-    YoutubeLoader,
-)
-from langchain_core.documents import Document
-from open_webui.env import SRC_LOG_LEVELS
-
-log = logging.getLogger(__name__)
-log.setLevel(SRC_LOG_LEVELS["RAG"])
-
-known_source_ext = [
-    "go",
-    "py",
-    "java",
-    "sh",
-    "bat",
-    "ps1",
-    "cmd",
-    "js",
-    "ts",
-    "css",
-    "cpp",
-    "hpp",
-    "h",
-    "c",
-    "cs",
-    "sql",
-    "log",
-    "ini",
-    "pl",
-    "pm",
-    "r",
-    "dart",
-    "dockerfile",
-    "env",
-    "php",
-    "hs",
-    "hsc",
-    "lua",
-    "nginxconf",
-    "conf",
-    "m",
-    "mm",
-    "plsql",
-    "perl",
-    "rb",
-    "rs",
-    "db2",
-    "scala",
-    "bash",
-    "swift",
-    "vue",
-    "svelte",
-    "msg",
-    "ex",
-    "exs",
-    "erl",
-    "tsx",
-    "jsx",
-    "hs",
-    "lhs",
-]
-
-
-class TikaLoader:
-    def __init__(self, url, file_path, mime_type=None):
-        self.url = url
-        self.file_path = file_path
-        self.mime_type = mime_type
-
-    def load(self) -> list[Document]:
-        with open(self.file_path, "rb") as f:
-            data = f.read()
-
-        if self.mime_type is not None:
-            headers = {"Content-Type": self.mime_type}
-        else:
-            headers = {}
-
-        endpoint = self.url
-        if not endpoint.endswith("/"):
-            endpoint += "/"
-        endpoint += "tika/text"
-
-        r = requests.put(endpoint, data=data, headers=headers)
-
-        if r.ok:
-            raw_metadata = r.json()
-            text = raw_metadata.get("X-TIKA:content", "<No text content found>")
-
-            if "Content-Type" in raw_metadata:
-                headers["Content-Type"] = raw_metadata["Content-Type"]
-
-            log.info("Tika extracted text: %s", text)
-
-            return [Document(page_content=text, metadata=headers)]
-        else:
-            raise Exception(f"Error calling Tika: {r.reason}")
-
-
-class Loader:
-    def __init__(self, engine: str = "", **kwargs):
-        self.engine = engine
-        self.kwargs = kwargs
-
-    def load(
-        self, filename: str, file_content_type: str, file_path: str
-    ) -> list[Document]:
-        loader = self._get_loader(filename, file_content_type, file_path)
-        docs = loader.load()
-
-        return [
-            Document(
-                page_content=ftfy.fix_text(doc.page_content), metadata=doc.metadata
-            )
-            for doc in docs
-        ]
-
-    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
-            ):
-                loader = TextLoader(file_path, autodetect_encoding=True)
-            else:
-                loader = TikaLoader(
-                    url=self.kwargs.get("TIKA_SERVER_URL"),
-                    file_path=file_path,
-                    mime_type=file_content_type,
-                )
-        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)
-            elif file_ext == "rst":
-                loader = UnstructuredRSTLoader(file_path, mode="elements")
-            elif file_ext == "xml":
-                loader = UnstructuredXMLLoader(file_path)
-            elif file_ext in ["htm", "html"]:
-                loader = BSHTMLLoader(file_path, open_encoding="unicode_escape")
-            elif file_ext == "md":
-                loader = TextLoader(file_path, autodetect_encoding=True)
-            elif file_content_type == "application/epub+zip":
-                loader = UnstructuredEPubLoader(file_path)
-            elif (
-                file_content_type
-                == "application/vnd.openxmlformats-officedocument.wordprocessingml.document"
-                or file_ext == "docx"
-            ):
-                loader = Docx2txtLoader(file_path)
-            elif file_content_type in [
-                "application/vnd.ms-excel",
-                "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
-            ] or file_ext in ["xls", "xlsx"]:
-                loader = UnstructuredExcelLoader(file_path)
-            elif file_content_type in [
-                "application/vnd.ms-powerpoint",
-                "application/vnd.openxmlformats-officedocument.presentationml.presentation",
-            ] or file_ext in ["ppt", "pptx"]:
-                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
-            ):
-                loader = TextLoader(file_path, autodetect_encoding=True)
-            else:
-                loader = TextLoader(file_path, autodetect_encoding=True)
-
-        return loader

backend/open_webui/apps/retrieval/utils.py

@@ -1,532 +0,0 @@
-import logging
-import os
-import uuid
-from typing import Optional, Union
-
-import asyncio
-import requests
-
-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.apps.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
-
-log = logging.getLogger(__name__)
-log.setLevel(SRC_LOG_LEVELS["RAG"])
-
-
-from typing import Any
-
-from langchain_core.callbacks import CallbackManagerForRetrieverRun
-from langchain_core.retrievers import BaseRetriever
-
-
-class VectorSearchRetriever(BaseRetriever):
-    collection_name: Any
-    embedding_function: Any
-    top_k: int
-
-    def _get_relevant_documents(
-        self,
-        query: str,
-        *,
-        run_manager: CallbackManagerForRetrieverRun,
-    ) -> list[Document]:
-        result = VECTOR_DB_CLIENT.search(
-            collection_name=self.collection_name,
-            vectors=[self.embedding_function(query)],
-            limit=self.top_k,
-        )
-
-        ids = result.ids[0]
-        metadatas = result.metadatas[0]
-        documents = result.documents[0]
-
-        results = []
-        for idx in range(len(ids)):
-            results.append(
-                Document(
-                    metadata=metadatas[idx],
-                    page_content=documents[idx],
-                )
-            )
-        return results
-
-
-def query_doc(
-    collection_name: str,
-    query_embedding: list[float],
-    k: int,
-):
-    try:
-        result = VECTOR_DB_CLIENT.search(
-            collection_name=collection_name,
-            vectors=[query_embedding],
-            limit=k,
-        )
-
-        log.info(f"query_doc:result {result.ids} {result.metadatas}")
-        return result
-    except Exception as e:
-        print(e)
-        raise e
-
-
-def query_doc_with_hybrid_search(
-    collection_name: str,
-    query: str,
-    embedding_function,
-    k: int,
-    reranking_function,
-    r: float,
-) -> dict:
-    try:
-        result = VECTOR_DB_CLIENT.get(collection_name=collection_name)
-
-        bm25_retriever = BM25Retriever.from_texts(
-            texts=result.documents[0],
-            metadatas=result.metadatas[0],
-        )
-        bm25_retriever.k = k
-
-        vector_search_retriever = VectorSearchRetriever(
-            collection_name=collection_name,
-            embedding_function=embedding_function,
-            top_k=k,
-        )
-
-        ensemble_retriever = EnsembleRetriever(
-            retrievers=[bm25_retriever, vector_search_retriever], weights=[0.5, 0.5]
-        )
-        compressor = RerankCompressor(
-            embedding_function=embedding_function,
-            top_n=k,
-            reranking_function=reranking_function,
-            r_score=r,
-        )
-
-        compression_retriever = ContextualCompressionRetriever(
-            base_compressor=compressor, base_retriever=ensemble_retriever
-        )
-
-        result = compression_retriever.invoke(query)
-        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]],
-        }
-
-        log.info(
-            "query_doc_with_hybrid_search:result "
-            + f'{result["metadatas"]} {result["distances"]}'
-        )
-        return result
-    except Exception as e:
-        raise e
-
-
-def merge_and_sort_query_results(
-    query_results: list[dict], k: int, reverse: bool = False
-) -> list[dict]:
-    # Initialize lists to store combined data
-    combined_distances = []
-    combined_documents = []
-    combined_metadatas = []
-
-    for data in query_results:
-        combined_distances.extend(data["distances"][0])
-        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]
-
-    # Create the output dictionary
-    result = {
-        "distances": [sorted_distances],
-        "documents": [sorted_documents],
-        "metadatas": [sorted_metadatas],
-    }
-
-    return result
-
-
-def query_collection(
-    collection_names: list[str],
-    queries: list[str],
-    embedding_function,
-    k: int,
-) -> dict:
-    results = []
-    for query in queries:
-        query_embedding = embedding_function(query)
-        for collection_name in collection_names:
-            if collection_name:
-                try:
-                    result = query_doc(
-                        collection_name=collection_name,
-                        k=k,
-                        query_embedding=query_embedding,
-                    )
-                    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_and_sort_query_results(results, k=k)
-
-
-def query_collection_with_hybrid_search(
-    collection_names: list[str],
-    queries: list[str],
-    embedding_function,
-    k: int,
-    reranking_function,
-    r: float,
-) -> dict:
-    results = []
-    error = False
-    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}"
-            )
-            error = True
-
-    if error:
-        raise Exception(
-            "Hybrid search failed for all collections. Using Non hybrid search as fallback."
-        )
-
-    return merge_and_sort_query_results(results, k=k, reverse=True)
-
-
-def get_embedding_function(
-    embedding_engine,
-    embedding_model,
-    embedding_function,
-    url,
-    key,
-    embedding_batch_size,
-):
-    if embedding_engine == "":
-        return lambda query: embedding_function.encode(query).tolist()
-    elif embedding_engine in ["ollama", "openai"]:
-        func = lambda query: generate_embeddings(
-            engine=embedding_engine,
-            model=embedding_model,
-            text=query,
-            url=url,
-            key=key,
-        )
-
-        def generate_multiple(query, 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]))
-                return embeddings
-            else:
-                return func(query)
-
-        return lambda query: generate_multiple(query, func)
-
-
-def get_sources_from_files(
-    files,
-    queries,
-    embedding_function,
-    k,
-    reranking_function,
-    r,
-    hybrid_search,
-):
-    log.debug(f"files: {files} {queries} {embedding_function} {reranking_function}")
-
-    extracted_collections = []
-    relevant_contexts = []
-
-    for file in files:
-        if file.get("context") == "full":
-            context = {
-                "documents": [[file.get("file").get("data", {}).get("content")]],
-                "metadatas": [[{"file_id": file.get("id"), "name": file.get("name")}]],
-            }
-        else:
-            context = None
-
-            collection_names = []
-            if file.get("type") == "collection":
-                if file.get("legacy"):
-                    collection_names = file.get("collection_names", [])
-                else:
-                    collection_names.append(file["id"])
-            elif file.get("collection_name"):
-                collection_names.append(file["collection_name"])
-            elif file.get("id"):
-                if file.get("legacy"):
-                    collection_names.append(f"{file['id']}")
-                else:
-                    collection_names.append(f"file-{file['id']}")
-
-            collection_names = set(collection_names).difference(extracted_collections)
-            if not collection_names:
-                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(
-                                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)
-
-            extracted_collections.extend(collection_names)
-
-        if context:
-            if "data" in file:
-                del file["data"]
-            relevant_contexts.append({**context, "file": file})
-
-    sources = []
-    for context in relevant_contexts:
-        try:
-            if "documents" in context:
-                if "metadatas" in context:
-                    source = {
-                        "source": context["file"],
-                        "document": context["documents"][0],
-                        "metadata": context["metadatas"][0],
-                    }
-                    if "distances" in context and context["distances"]:
-                        source["distances"] = context["distances"][0]
-
-                    sources.append(source)
-        except Exception as e:
-            log.exception(e)
-
-    return sources
-
-
-def get_model_path(model: str, update_model: bool = False):
-    # Construct huggingface_hub kwargs with local_files_only to return the snapshot path
-    cache_dir = os.getenv("SENTENCE_TRANSFORMERS_HOME")
-
-    local_files_only = not update_model
-
-    snapshot_kwargs = {
-        "cache_dir": cache_dir,
-        "local_files_only": local_files_only,
-    }
-
-    log.debug(f"model: {model}")
-    log.debug(f"snapshot_kwargs: {snapshot_kwargs}")
-
-    # Inspiration from upstream sentence_transformers
-    if (
-        os.path.exists(model)
-        or ("\\" in model or model.count("/") > 1)
-        and local_files_only
-    ):
-        # If fully qualified path exists, return input, else set repo_id
-        return model
-    elif "/" not in model:
-        # Set valid repo_id for model short-name
-        model = "sentence-transformers" + "/" + model
-
-    snapshot_kwargs["repo_id"] = model
-
-    # Attempt to query the huggingface_hub library to determine the local path and/or to update
-    try:
-        model_repo_path = snapshot_download(**snapshot_kwargs)
-        log.debug(f"model_repo_path: {model_repo_path}")
-        return model_repo_path
-    except Exception as e:
-        log.exception(f"Cannot determine model snapshot path: {e}")
-        return model
-
-
-def generate_openai_batch_embeddings(
-    model: str, texts: list[str], url: str = "https://api.openai.com/v1", key: str = ""
-) -> Optional[list[list[float]]]:
-    try:
-        r = requests.post(
-            f"{url}/embeddings",
-            headers={
-                "Content-Type": "application/json",
-                "Authorization": f"Bearer {key}",
-            },
-            json={"input": texts, "model": model},
-        )
-        r.raise_for_status()
-        data = r.json()
-        if "data" in data:
-            return [elem["embedding"] for elem in data["data"]]
-        else:
-            raise "Something went wrong :/"
-    except Exception as e:
-        print(e)
-        return None
-
-
-def generate_ollama_batch_embeddings(
-    model: str, texts: list[str], url: str, key: str = ""
-) -> Optional[list[list[float]]]:
-    try:
-        r = requests.post(
-            f"{url}/api/embed",
-            headers={
-                "Content-Type": "application/json",
-                "Authorization": f"Bearer {key}",
-            },
-            json={"input": texts, "model": model},
-        )
-        r.raise_for_status()
-        data = r.json()
-
-        if "embeddings" in data:
-            return data["embeddings"]
-        else:
-            raise "Something went wrong :/"
-    except Exception as e:
-        print(e)
-        return None
-
-
-def generate_embeddings(engine: str, model: str, text: Union[str, list[str]], **kwargs):
-    url = kwargs.get("url", "")
-    key = kwargs.get("key", "")
-
-    if engine == "ollama":
-        if isinstance(text, list):
-            embeddings = generate_ollama_batch_embeddings(
-                **{"model": model, "texts": text, "url": url, "key": key}
-            )
-        else:
-            embeddings = generate_ollama_batch_embeddings(
-                **{"model": model, "texts": [text], "url": url, "key": key}
-            )
-        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)
-        else:
-            embeddings = generate_openai_batch_embeddings(model, [text], url, key)
-
-        return embeddings[0] if isinstance(text, str) else embeddings
-
-
-import operator
-from typing import Optional, Sequence
-
-from langchain_core.callbacks import Callbacks
-from langchain_core.documents import BaseDocumentCompressor, Document
-
-
-class RerankCompressor(BaseDocumentCompressor):
-    embedding_function: Any
-    top_n: int
-    reranking_function: Any
-    r_score: float
-
-    class Config:
-        extra = "forbid"
-        arbitrary_types_allowed = True
-
-    def compress_documents(
-        self,
-        documents: Sequence[Document],
-        query: str,
-        callbacks: Optional[Callbacks] = None,
-    ) -> Sequence[Document]:
-        reranking = self.reranking_function is not None
-
-        if reranking:
-            scores = self.reranking_function.predict(
-                [(query, doc.page_content) for doc in documents]
-            )
-        else:
-            from sentence_transformers import util
-
-            query_embedding = self.embedding_function(query)
-            document_embedding = self.embedding_function(
-                [doc.page_content for doc in documents]
-            )
-            scores = util.cos_sim(query_embedding, document_embedding)[0]
-
-        docs_with_scores = list(zip(documents, scores.tolist()))
-        if self.r_score:
-            docs_with_scores = [
-                (d, s) for d, s in docs_with_scores if s >= self.r_score
-            ]
-
-        result = sorted(docs_with_scores, key=operator.itemgetter(1), reverse=True)
-        final_results = []
-        for doc, doc_score in result[: self.top_n]:
-            metadata = doc.metadata
-            metadata["score"] = doc_score
-            doc = Document(
-                page_content=doc.page_content,
-                metadata=metadata,
-            )
-            final_results.append(doc)
-        return final_results

backend/open_webui/apps/retrieval/vector/connector.py

@@ -1,22 +0,0 @@
-from open_webui.config import VECTOR_DB
-
-if VECTOR_DB == "milvus":
-    from open_webui.apps.retrieval.vector.dbs.milvus import MilvusClient
-
-    VECTOR_DB_CLIENT = MilvusClient()
-elif VECTOR_DB == "qdrant":
-    from open_webui.apps.retrieval.vector.dbs.qdrant import QdrantClient
-
-    VECTOR_DB_CLIENT = QdrantClient()
-elif VECTOR_DB == "opensearch":
-    from open_webui.apps.retrieval.vector.dbs.opensearch import OpenSearchClient
-
-    VECTOR_DB_CLIENT = OpenSearchClient()
-elif VECTOR_DB == "pgvector":
-    from open_webui.apps.retrieval.vector.dbs.pgvector import PgvectorClient
-
-    VECTOR_DB_CLIENT = PgvectorClient()
-else:
-    from open_webui.apps.retrieval.vector.dbs.chroma import ChromaClient
-
-    VECTOR_DB_CLIENT = ChromaClient()

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

@@ -1,178 +0,0 @@
-from opensearchpy import OpenSearch
-from typing import Optional
-
-from open_webui.apps.retrieval.vector.main import VectorItem, SearchResult, GetResult
-from open_webui.config import (
-    OPENSEARCH_URI,
-    OPENSEARCH_SSL,
-    OPENSEARCH_CERT_VERIFY,
-    OPENSEARCH_USERNAME,
-    OPENSEARCH_PASSWORD,
-)
-
-
-class OpenSearchClient:
-    def __init__(self):
-        self.index_prefix = "open_webui"
-        self.client = OpenSearch(
-            hosts=[OPENSEARCH_URI],
-            use_ssl=OPENSEARCH_SSL,
-            verify_certs=OPENSEARCH_CERT_VERIFY,
-            http_auth=(OPENSEARCH_USERNAME, OPENSEARCH_PASSWORD),
-        )
-
-    def _result_to_get_result(self, result) -> GetResult:
-        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)
-
-    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
-        )
-
-    def _create_index(self, index_name: str, dimension: int):
-        body = {
-            "mappings": {
-                "properties": {
-                    "id": {"type": "keyword"},
-                    "vector": {
-                        "type": "dense_vector",
-                        "dims": dimension,  # Adjust based on your vector dimensions
-                        "index": true,
-                        "similarity": "faiss",
-                        "method": {
-                            "name": "hnsw",
-                            "space_type": "ip",  # Use inner product to approximate cosine similarity
-                            "engine": "faiss",
-                            "ef_construction": 128,
-                            "m": 16,
-                        },
-                    },
-                    "text": {"type": "text"},
-                    "metadata": {"type": "object"},
-                }
-            }
-        }
-        self.client.indices.create(index=f"{self.index_prefix}_{index_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:
-        # 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}")
-
-    def delete_colleciton(self, index_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}")
-
-    def search(
-        self, index_name: str, vectors: list[list[float]], limit: int
-    ) -> Optional[SearchResult]:
-        query = {
-            "size": limit,
-            "_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
-        )
-
-        return self._result_to_search_result(result)
-
-    def get_or_create_index(self, index_name: str, dimension: int):
-        if not self.has_index(index_name):
-            self._create_index(index_name, dimension)
-
-    def get(self, index_name: str) -> Optional[GetResult]:
-        query = {"query": {"match_all": {}}, "_source": ["text", "metadata"]}
-
-        result = self.client.search(
-            index=f"{self.index_prefix}_{index_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"]))
-
-        for batch in self._create_batches(items):
-            actions = [
-                {
-                    "index": {
-                        "_id": item["id"],
-                        "_source": {
-                            "vector": item["vector"],
-                            "text": item["text"],
-                            "metadata": item["metadata"],
-                        },
-                    }
-                }
-                for item in batch
-            ]
-            self.client.bulk(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"]))
-
-        for batch in self._create_batches(items):
-            actions = [
-                {
-                    "index": {
-                        "_id": item["id"],
-                        "_source": {
-                            "vector": item["vector"],
-                            "text": item["text"],
-                            "metadata": item["metadata"],
-                        },
-                    }
-                }
-                for item in batch
-            ]
-            self.client.bulk(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 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/apps/retrieval/vector/main.py

@@ -1,19 +0,0 @@
-from pydantic import BaseModel
-from typing import Optional, List, Any
-
-
-class VectorItem(BaseModel):
-    id: str
-    text: str
-    vector: List[float | int]
-    metadata: Any
-
-
-class GetResult(BaseModel):
-    ids: Optional[List[List[str]]]
-    documents: Optional[List[List[str]]]
-    metadatas: Optional[List[List[Any]]]
-
-
-class SearchResult(GetResult):
-    distances: Optional[List[List[float | int]]]

backend/open_webui/apps/retrieval/web/duckduckgo.py

@@ -1,50 +0,0 @@
-import logging
-from typing import Optional
-
-from open_webui.apps.retrieval.web.main import SearchResult, get_filtered_results
-from duckduckgo_search import DDGS
-from open_webui.env import SRC_LOG_LEVELS
-
-log = logging.getLogger(__name__)
-log.setLevel(SRC_LOG_LEVELS["RAG"])
-
-
-def search_duckduckgo(
-    query: str, count: int, filter_list: Optional[list[str]] = None
-) -> list[SearchResult]:
-    """
-    Search using DuckDuckGo's Search API and return the results as a list of SearchResult objects.
-    Args:
-        query (str): The query to search for
-        count (int): The number of results to return
-
-    Returns:
-        list[SearchResult]: A list of search results
-    """
-    # Use the DDGS context manager to create a DDGS object
-    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"),
-            )
-        )
-    if filter_list:
-        results = get_filtered_results(results, filter_list)
-    # Return the list of search results
-    return results

backend/open_webui/apps/retrieval/web/google_pse.py

@@ -1,50 +0,0 @@
-import logging
-from typing import Optional
-
-import requests
-from open_webui.apps.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_google_pse(
-    api_key: str,
-    search_engine_id: str,
-    query: str,
-    count: int,
-    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.
-
-    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
-    """
-    url = "https://www.googleapis.com/customsearch/v1"
-
-    headers = {"Content-Type": "application/json"}
-    params = {
-        "cx": search_engine_id,
-        "q": query,
-        "key": api_key,
-        "num": count,
-    }
-
-    response = requests.request("GET", url, headers=headers, params=params)
-    response.raise_for_status()
-
-    json_response = response.json()
-    results = json_response.get("items", [])
-    if filter_list:
-        results = get_filtered_results(results, filter_list)
-    return [
-        SearchResult(
-            link=result["link"],
-            title=result.get("title"),
-            snippet=result.get("snippet"),
-        )
-        for result in results
-    ]

backend/open_webui/apps/retrieval/web/utils.py

@@ -1,97 +0,0 @@
-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 logging
-
-log = logging.getLogger(__name__)
-log.setLevel(SRC_LOG_LEVELS["RAG"])
-
-
-def validate_url(url: Union[str, Sequence[str]]):
-    if isinstance(url, str):
-        if isinstance(validators.url(url), validators.ValidationError):
-            raise ValueError(ERROR_MESSAGES.INVALID_URL)
-        if not ENABLE_RAG_LOCAL_WEB_FETCH:
-            # Local web fetch is disabled, filter out any URLs that resolve to private IP addresses
-            parsed_url = urllib.parse.urlparse(url)
-            # Get IPv4 and IPv6 addresses
-            ipv4_addresses, ipv6_addresses = resolve_hostname(parsed_url.hostname)
-            # Check if any of the resolved addresses are private
-            # This is technically still vulnerable to DNS rebinding attacks, as we don't control WebBaseLoader
-            for ip in ipv4_addresses:
-                if validators.ipv4(ip, private=True):
-                    raise ValueError(ERROR_MESSAGES.INVALID_URL)
-            for ip in ipv6_addresses:
-                if validators.ipv6(ip, private=True):
-                    raise ValueError(ERROR_MESSAGES.INVALID_URL)
-        return True
-    elif isinstance(url, Sequence):
-        return all(validate_url(u) for u in url)
-    else:
-        return False
-
-
-def resolve_hostname(hostname):
-    # Get address information
-    addr_info = socket.getaddrinfo(hostname, None)
-
-    # Extract IP addresses from address information
-    ipv4_addresses = [info[4][0] for info in addr_info if info[0] == socket.AF_INET]
-    ipv6_addresses = [info[4][0] for info in addr_info if info[0] == socket.AF_INET6]
-
-    return ipv4_addresses, ipv6_addresses
-
-
-class SafeWebBaseLoader(WebBaseLoader):
-    """WebBaseLoader with enhanced error handling for URLs."""
-
-    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:
-            try:
-                soup = self._scrape(path, bs_kwargs=self.bs_kwargs)
-                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.")
-
-                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}")
-
-
-def get_web_loader(
-    url: Union[str, Sequence[str]],
-    verify_ssl: bool = True,
-    requests_per_second: int = 2,
-):
-    # Check if the URL is valid
-    if not validate_url(url):
-        raise ValueError(ERROR_MESSAGES.INVALID_URL)
-    return SafeWebBaseLoader(
-        url,
-        verify_ssl=verify_ssl,
-        requests_per_second=requests_per_second,
-        continue_on_failure=True,
-    )

backend/open_webui/apps/socket/main.py

@@ -1,221 +0,0 @@
-# TODO: move socket to webui app
-
-import asyncio
-import socketio
-import logging
-import sys
-import time
-
-from open_webui.apps.webui.models.users import Users
-from open_webui.env import (
-    ENABLE_WEBSOCKET_SUPPORT,
-    WEBSOCKET_MANAGER,
-    WEBSOCKET_REDIS_URL,
-)
-from open_webui.utils.utils import decode_token
-from open_webui.apps.socket.utils import RedisDict
-
-from open_webui.env import (
-    GLOBAL_LOG_LEVEL,
-    SRC_LOG_LEVELS,
-)
-
-
-logging.basicConfig(stream=sys.stdout, level=GLOBAL_LOG_LEVEL)
-log = logging.getLogger(__name__)
-log.setLevel(SRC_LOG_LEVELS["SOCKET"])
-
-
-if WEBSOCKET_MANAGER == "redis":
-    mgr = socketio.AsyncRedisManager(WEBSOCKET_REDIS_URL)
-    sio = socketio.AsyncServer(
-        cors_allowed_origins=[],
-        async_mode="asgi",
-        transports=(
-            ["polling", "websocket"] if ENABLE_WEBSOCKET_SUPPORT else ["polling"]
-        ),
-        allow_upgrades=ENABLE_WEBSOCKET_SUPPORT,
-        always_connect=True,
-        client_manager=mgr,
-    )
-else:
-    sio = socketio.AsyncServer(
-        cors_allowed_origins=[],
-        async_mode="asgi",
-        transports=(
-            ["polling", "websocket"] if ENABLE_WEBSOCKET_SUPPORT else ["polling"]
-        ),
-        allow_upgrades=ENABLE_WEBSOCKET_SUPPORT,
-        always_connect=True,
-    )
-
-
-# Dictionary to maintain the user pool
-
-if WEBSOCKET_MANAGER == "redis":
-    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)
-else:
-    SESSION_POOL = {}
-    USER_POOL = {}
-    USAGE_POOL = {}
-
-
-# Timeout duration in seconds
-TIMEOUT_DURATION = 3
-
-
-async def periodic_usage_pool_cleanup():
-    while True:
-        now = int(time.time())
-        for model_id, connections in list(USAGE_POOL.items()):
-            # Creating a list of sids to remove if they have timed out
-            expired_sids = [
-                sid
-                for sid, details in connections.items()
-                if now - details["updated_at"] > TIMEOUT_DURATION
-            ]
-
-            for sid in expired_sids:
-                del connections[sid]
-
-            if not connections:
-                log.debug(f"Cleaning up model {model_id} from usage pool")
-                del USAGE_POOL[model_id]
-            else:
-                USAGE_POOL[model_id] = connections
-
-            # Emit updated usage information after cleaning
-            await sio.emit("usage", {"models": get_models_in_use()})
-
-        await asyncio.sleep(TIMEOUT_DURATION)
-
-
-app = socketio.ASGIApp(
-    sio,
-    socketio_path="/ws/socket.io",
-)
-
-
-def get_models_in_use():
-    # List models that are currently in use
-    models_in_use = list(USAGE_POOL.keys())
-    return models_in_use
-
-
-@sio.on("usage")
-async def usage(sid, data):
-    model_id = data["model"]
-    # Record the timestamp for the last update
-    current_time = int(time.time())
-
-    # Store the new usage data and task
-    USAGE_POOL[model_id] = {
-        **(USAGE_POOL[model_id] if model_id in USAGE_POOL else {}),
-        sid: {"updated_at": current_time},
-    }
-
-    # Broadcast the usage data to all clients
-    await sio.emit("usage", {"models": get_models_in_use()})
-
-
-@sio.event
-async def connect(sid, environ, auth):
-    user = None
-    if auth and "token" in auth:
-        data = decode_token(auth["token"])
-
-        if data is not None and "id" in data:
-            user = Users.get_user_by_id(data["id"])
-
-        if user:
-            SESSION_POOL[sid] = user.id
-            if user.id in USER_POOL:
-                USER_POOL[user.id].append(sid)
-            else:
-                USER_POOL[user.id] = [sid]
-
-            # print(f"user {user.name}({user.id}) connected with session ID {sid}")
-            await sio.emit("user-count", {"count": len(USER_POOL.items())})
-            await sio.emit("usage", {"models": get_models_in_use()})
-
-
-@sio.on("user-join")
-async def user_join(sid, data):
-    # print("user-join", sid, data)
-
-    auth = data["auth"] if "auth" in data else None
-    if not auth or "token" not in auth:
-        return
-
-    data = decode_token(auth["token"])
-    if data is None or "id" not in data:
-        return
-
-    user = Users.get_user_by_id(data["id"])
-    if not user:
-        return
-
-    SESSION_POOL[sid] = user.id
-    if user.id in USER_POOL:
-        USER_POOL[user.id].append(sid)
-    else:
-        USER_POOL[user.id] = [sid]
-
-    # print(f"user {user.name}({user.id}) connected with session ID {sid}")
-
-    await sio.emit("user-count", {"count": len(USER_POOL.items())})
-
-
-@sio.on("user-count")
-async def user_count(sid):
-    await sio.emit("user-count", {"count": len(USER_POOL.items())})
-
-
-@sio.event
-async def disconnect(sid):
-    if sid in SESSION_POOL:
-        user_id = SESSION_POOL[sid]
-        del SESSION_POOL[sid]
-
-        USER_POOL[user_id] = [_sid for _sid in USER_POOL[user_id] if _sid != sid]
-
-        if len(USER_POOL[user_id]) == 0:
-            del USER_POOL[user_id]
-
-        await sio.emit("user-count", {"count": len(USER_POOL)})
-    else:
-        pass
-        # print(f"Unknown session ID {sid} disconnected")
-
-
-def get_event_emitter(request_info):
-    async def __event_emitter__(event_data):
-        await sio.emit(
-            "chat-events",
-            {
-                "chat_id": request_info["chat_id"],
-                "message_id": request_info["message_id"],
-                "data": event_data,
-            },
-            to=request_info["session_id"],
-        )
-
-    return __event_emitter__
-
-
-def get_event_call(request_info):
-    async def __event_call__(event_data):
-        response = await sio.call(
-            "chat-events",
-            {
-                "chat_id": request_info["chat_id"],
-                "message_id": request_info["message_id"],
-                "data": event_data,
-            },
-            to=request_info["session_id"],
-        )
-        return response
-
-    return __event_call__

backend/open_webui/apps/webui/main.py

@@ -1,506 +0,0 @@
-import inspect
-import json
-import logging
-import time
-from typing import AsyncGenerator, Generator, Iterator
-
-from open_webui.apps.socket.main import get_event_call, get_event_emitter
-from open_webui.apps.webui.models.functions import Functions
-from open_webui.apps.webui.models.models import Models
-from open_webui.apps.webui.routers import (
-    auths,
-    chats,
-    folders,
-    configs,
-    groups,
-    files,
-    functions,
-    memories,
-    models,
-    knowledge,
-    prompts,
-    evaluations,
-    tools,
-    users,
-    utils,
-)
-from open_webui.apps.webui.utils import load_function_module_by_id
-from open_webui.config import (
-    ADMIN_EMAIL,
-    CORS_ALLOW_ORIGIN,
-    DEFAULT_MODELS,
-    DEFAULT_PROMPT_SUGGESTIONS,
-    DEFAULT_USER_ROLE,
-    MODEL_ORDER_LIST,
-    ENABLE_COMMUNITY_SHARING,
-    ENABLE_LOGIN_FORM,
-    ENABLE_MESSAGE_RATING,
-    ENABLE_SIGNUP,
-    ENABLE_API_KEY,
-    ENABLE_EVALUATION_ARENA_MODELS,
-    EVALUATION_ARENA_MODELS,
-    DEFAULT_ARENA_MODEL,
-    JWT_EXPIRES_IN,
-    ENABLE_OAUTH_ROLE_MANAGEMENT,
-    OAUTH_ROLES_CLAIM,
-    OAUTH_EMAIL_CLAIM,
-    OAUTH_PICTURE_CLAIM,
-    OAUTH_USERNAME_CLAIM,
-    OAUTH_ALLOWED_ROLES,
-    OAUTH_ADMIN_ROLES,
-    SHOW_ADMIN_DETAILS,
-    USER_PERMISSIONS,
-    WEBHOOK_URL,
-    WEBUI_AUTH,
-    WEBUI_BANNERS,
-    ENABLE_LDAP,
-    LDAP_SERVER_LABEL,
-    LDAP_SERVER_HOST,
-    LDAP_SERVER_PORT,
-    LDAP_ATTRIBUTE_FOR_USERNAME,
-    LDAP_SEARCH_FILTERS,
-    LDAP_SEARCH_BASE,
-    LDAP_APP_DN,
-    LDAP_APP_PASSWORD,
-    LDAP_USE_TLS,
-    LDAP_CA_CERT_FILE,
-    LDAP_CIPHERS,
-    AppConfig,
-)
-from open_webui.env import (
-    ENV,
-    SRC_LOG_LEVELS,
-    WEBUI_AUTH_TRUSTED_EMAIL_HEADER,
-    WEBUI_AUTH_TRUSTED_NAME_HEADER,
-)
-from fastapi import FastAPI
-from fastapi.middleware.cors import CORSMiddleware
-from fastapi.responses import StreamingResponse
-from pydantic import BaseModel
-from open_webui.utils.misc import (
-    openai_chat_chunk_message_template,
-    openai_chat_completion_message_template,
-)
-from open_webui.utils.payload import (
-    apply_model_params_to_body_openai,
-    apply_model_system_prompt_to_body,
-)
-
-
-from open_webui.utils.tools import get_tools
-
-app = FastAPI(
-    docs_url="/docs" if ENV == "dev" else None,
-    openapi_url="/openapi.json" if ENV == "dev" else None,
-    redoc_url=None,
-)
-
-log = logging.getLogger(__name__)
-log.setLevel(SRC_LOG_LEVELS["MAIN"])
-
-app.state.config = AppConfig()
-
-app.state.config.ENABLE_SIGNUP = ENABLE_SIGNUP
-app.state.config.ENABLE_LOGIN_FORM = ENABLE_LOGIN_FORM
-app.state.config.ENABLE_API_KEY = ENABLE_API_KEY
-
-app.state.config.JWT_EXPIRES_IN = JWT_EXPIRES_IN
-app.state.AUTH_TRUSTED_EMAIL_HEADER = WEBUI_AUTH_TRUSTED_EMAIL_HEADER
-app.state.AUTH_TRUSTED_NAME_HEADER = WEBUI_AUTH_TRUSTED_NAME_HEADER
-
-
-app.state.config.SHOW_ADMIN_DETAILS = SHOW_ADMIN_DETAILS
-app.state.config.ADMIN_EMAIL = ADMIN_EMAIL
-
-
-app.state.config.DEFAULT_MODELS = DEFAULT_MODELS
-app.state.config.DEFAULT_PROMPT_SUGGESTIONS = DEFAULT_PROMPT_SUGGESTIONS
-app.state.config.DEFAULT_USER_ROLE = DEFAULT_USER_ROLE
-
-
-app.state.config.USER_PERMISSIONS = USER_PERMISSIONS
-app.state.config.WEBHOOK_URL = WEBHOOK_URL
-app.state.config.BANNERS = WEBUI_BANNERS
-app.state.config.MODEL_ORDER_LIST = MODEL_ORDER_LIST
-
-app.state.config.ENABLE_COMMUNITY_SHARING = ENABLE_COMMUNITY_SHARING
-app.state.config.ENABLE_MESSAGE_RATING = ENABLE_MESSAGE_RATING
-
-app.state.config.ENABLE_EVALUATION_ARENA_MODELS = ENABLE_EVALUATION_ARENA_MODELS
-app.state.config.EVALUATION_ARENA_MODELS = EVALUATION_ARENA_MODELS
-
-app.state.config.OAUTH_USERNAME_CLAIM = OAUTH_USERNAME_CLAIM
-app.state.config.OAUTH_PICTURE_CLAIM = OAUTH_PICTURE_CLAIM
-app.state.config.OAUTH_EMAIL_CLAIM = OAUTH_EMAIL_CLAIM
-
-app.state.config.ENABLE_OAUTH_ROLE_MANAGEMENT = ENABLE_OAUTH_ROLE_MANAGEMENT
-app.state.config.OAUTH_ROLES_CLAIM = OAUTH_ROLES_CLAIM
-app.state.config.OAUTH_ALLOWED_ROLES = OAUTH_ALLOWED_ROLES
-app.state.config.OAUTH_ADMIN_ROLES = OAUTH_ADMIN_ROLES
-
-app.state.config.ENABLE_LDAP = ENABLE_LDAP
-app.state.config.LDAP_SERVER_LABEL = LDAP_SERVER_LABEL
-app.state.config.LDAP_SERVER_HOST = LDAP_SERVER_HOST
-app.state.config.LDAP_SERVER_PORT = LDAP_SERVER_PORT
-app.state.config.LDAP_ATTRIBUTE_FOR_USERNAME = LDAP_ATTRIBUTE_FOR_USERNAME
-app.state.config.LDAP_APP_DN = LDAP_APP_DN
-app.state.config.LDAP_APP_PASSWORD = LDAP_APP_PASSWORD
-app.state.config.LDAP_SEARCH_BASE = LDAP_SEARCH_BASE
-app.state.config.LDAP_SEARCH_FILTERS = LDAP_SEARCH_FILTERS
-app.state.config.LDAP_USE_TLS = LDAP_USE_TLS
-app.state.config.LDAP_CA_CERT_FILE = LDAP_CA_CERT_FILE
-app.state.config.LDAP_CIPHERS = LDAP_CIPHERS
-
-app.state.TOOLS = {}
-app.state.FUNCTIONS = {}
-
-app.add_middleware(
-    CORSMiddleware,
-    allow_origins=CORS_ALLOW_ORIGIN,
-    allow_credentials=True,
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
-
-
-app.include_router(configs.router, prefix="/configs", tags=["configs"])
-
-app.include_router(auths.router, prefix="/auths", tags=["auths"])
-app.include_router(users.router, prefix="/users", tags=["users"])
-
-app.include_router(chats.router, prefix="/chats", tags=["chats"])
-
-app.include_router(models.router, prefix="/models", tags=["models"])
-app.include_router(knowledge.router, prefix="/knowledge", tags=["knowledge"])
-app.include_router(prompts.router, prefix="/prompts", tags=["prompts"])
-app.include_router(tools.router, prefix="/tools", tags=["tools"])
-
-app.include_router(memories.router, prefix="/memories", tags=["memories"])
-app.include_router(folders.router, prefix="/folders", tags=["folders"])
-
-app.include_router(groups.router, prefix="/groups", tags=["groups"])
-app.include_router(files.router, prefix="/files", tags=["files"])
-app.include_router(functions.router, prefix="/functions", tags=["functions"])
-app.include_router(evaluations.router, prefix="/evaluations", tags=["evaluations"])
-
-
-app.include_router(utils.router, prefix="/utils", tags=["utils"])
-
-
-@app.get("/")
-async def get_status():
-    return {
-        "status": True,
-        "auth": WEBUI_AUTH,
-        "default_models": app.state.config.DEFAULT_MODELS,
-        "default_prompt_suggestions": app.state.config.DEFAULT_PROMPT_SUGGESTIONS,
-    }
-
-
-async def get_all_models():
-    models = []
-    pipe_models = await get_pipe_models()
-    models = models + pipe_models
-
-    if app.state.config.ENABLE_EVALUATION_ARENA_MODELS:
-        arena_models = []
-        if len(app.state.config.EVALUATION_ARENA_MODELS) > 0:
-            arena_models = [
-                {
-                    "id": model["id"],
-                    "name": model["name"],
-                    "info": {
-                        "meta": model["meta"],
-                    },
-                    "object": "model",
-                    "created": int(time.time()),
-                    "owned_by": "arena",
-                    "arena": True,
-                }
-                for model in app.state.config.EVALUATION_ARENA_MODELS
-            ]
-        else:
-            # Add default arena model
-            arena_models = [
-                {
-                    "id": DEFAULT_ARENA_MODEL["id"],
-                    "name": DEFAULT_ARENA_MODEL["name"],
-                    "info": {
-                        "meta": DEFAULT_ARENA_MODEL["meta"],
-                    },
-                    "object": "model",
-                    "created": int(time.time()),
-                    "owned_by": "arena",
-                    "arena": True,
-                }
-            ]
-        models = models + arena_models
-    return models
-
-
-def get_function_module(pipe_id: str):
-    # Check if function is already loaded
-    if pipe_id not in app.state.FUNCTIONS:
-        function_module, _, _ = load_function_module_by_id(pipe_id)
-        app.state.FUNCTIONS[pipe_id] = function_module
-    else:
-        function_module = app.state.FUNCTIONS[pipe_id]
-
-    if hasattr(function_module, "valves") and hasattr(function_module, "Valves"):
-        valves = Functions.get_function_valves_by_id(pipe_id)
-        function_module.valves = function_module.Valves(**(valves if valves else {}))
-    return function_module
-
-
-async def get_pipe_models():
-    pipes = Functions.get_functions_by_type("pipe", active_only=True)
-    pipe_models = []
-
-    for pipe in pipes:
-        function_module = get_function_module(pipe.id)
-
-        # Check if function is a manifold
-        if hasattr(function_module, "pipes"):
-            sub_pipes = []
-
-            # Check if pipes is a function or a list
-
-            try:
-                if callable(function_module.pipes):
-                    sub_pipes = function_module.pipes()
-                else:
-                    sub_pipes = function_module.pipes
-            except Exception as e:
-                log.exception(e)
-                sub_pipes = []
-
-            log.debug(
-                f"get_pipe_models: function '{pipe.id}' is a manifold of {sub_pipes}"
-            )
-
-            for p in sub_pipes:
-                sub_pipe_id = f'{pipe.id}.{p["id"]}'
-                sub_pipe_name = p["name"]
-
-                if hasattr(function_module, "name"):
-                    sub_pipe_name = f"{function_module.name}{sub_pipe_name}"
-
-                pipe_flag = {"type": pipe.type}
-
-                pipe_models.append(
-                    {
-                        "id": sub_pipe_id,
-                        "name": sub_pipe_name,
-                        "object": "model",
-                        "created": pipe.created_at,
-                        "owned_by": "openai",
-                        "pipe": pipe_flag,
-                    }
-                )
-        else:
-            pipe_flag = {"type": "pipe"}
-
-            log.debug(
-                f"get_pipe_models: function '{pipe.id}' is a single pipe {{ 'id': {pipe.id}, 'name': {pipe.name} }}"
-            )
-
-            pipe_models.append(
-                {
-                    "id": pipe.id,
-                    "name": pipe.name,
-                    "object": "model",
-                    "created": pipe.created_at,
-                    "owned_by": "openai",
-                    "pipe": pipe_flag,
-                }
-            )
-
-    return pipe_models
-
-
-async def execute_pipe(pipe, params):
-    if inspect.iscoroutinefunction(pipe):
-        return await pipe(**params)
-    else:
-        return pipe(**params)
-
-
-async def get_message_content(res: str | Generator | AsyncGenerator) -> str:
-    if isinstance(res, str):
-        return res
-    if isinstance(res, Generator):
-        return "".join(map(str, res))
-    if isinstance(res, AsyncGenerator):
-        return "".join([str(stream) async for stream in res])
-
-
-def process_line(form_data: dict, line):
-    if isinstance(line, BaseModel):
-        line = line.model_dump_json()
-        line = f"data: {line}"
-    if isinstance(line, dict):
-        line = f"data: {json.dumps(line)}"
-
-    try:
-        line = line.decode("utf-8")
-    except Exception:
-        pass
-
-    if line.startswith("data:"):
-        return f"{line}\n\n"
-    else:
-        line = openai_chat_chunk_message_template(form_data["model"], line)
-        return f"data: {json.dumps(line)}\n\n"
-
-
-def get_pipe_id(form_data: dict) -> str:
-    pipe_id = form_data["model"]
-    if "." in pipe_id:
-        pipe_id, _ = pipe_id.split(".", 1)
-
-    return pipe_id
-
-
-def get_function_params(function_module, form_data, user, extra_params=None):
-    if extra_params is None:
-        extra_params = {}
-
-    pipe_id = get_pipe_id(form_data)
-
-    # Get the signature of the function
-    sig = inspect.signature(function_module.pipe)
-    params = {"body": form_data} | {
-        k: v for k, v in extra_params.items() if k in sig.parameters
-    }
-
-    if "__user__" in params and hasattr(function_module, "UserValves"):
-        user_valves = Functions.get_user_valves_by_id_and_user_id(pipe_id, user.id)
-        try:
-            params["__user__"]["valves"] = function_module.UserValves(**user_valves)
-        except Exception as e:
-            log.exception(e)
-            params["__user__"]["valves"] = function_module.UserValves()
-
-    return params
-
-
-async def generate_function_chat_completion(form_data, user, models: dict = {}):
-    model_id = form_data.get("model")
-    model_info = Models.get_model_by_id(model_id)
-
-    metadata = form_data.pop("metadata", {})
-
-    files = metadata.get("files", [])
-    tool_ids = metadata.get("tool_ids", [])
-    # Check if tool_ids is None
-    if tool_ids is None:
-        tool_ids = []
-
-    __event_emitter__ = None
-    __event_call__ = None
-    __task__ = None
-    __task_body__ = None
-
-    if metadata:
-        if all(k in metadata for k in ("session_id", "chat_id", "message_id")):
-            __event_emitter__ = get_event_emitter(metadata)
-            __event_call__ = get_event_call(metadata)
-        __task__ = metadata.get("task", None)
-        __task_body__ = metadata.get("task_body", None)
-
-    extra_params = {
-        "__event_emitter__": __event_emitter__,
-        "__event_call__": __event_call__,
-        "__task__": __task__,
-        "__task_body__": __task_body__,
-        "__files__": files,
-        "__user__": {
-            "id": user.id,
-            "email": user.email,
-            "name": user.name,
-            "role": user.role,
-        },
-        "__metadata__": metadata,
-    }
-    extra_params["__tools__"] = get_tools(
-        app,
-        tool_ids,
-        user,
-        {
-            **extra_params,
-            "__model__": models.get(form_data["model"], None),
-            "__messages__": form_data["messages"],
-            "__files__": files,
-        },
-    )
-
-    if model_info:
-        if model_info.base_model_id:
-            form_data["model"] = model_info.base_model_id
-
-        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)
-
-    pipe_id = get_pipe_id(form_data)
-    function_module = get_function_module(pipe_id)
-
-    pipe = function_module.pipe
-    params = get_function_params(function_module, form_data, user, extra_params)
-
-    if form_data.get("stream", False):
-
-        async def stream_content():
-            try:
-                res = await execute_pipe(pipe, params)
-
-                # Directly return if the response is a StreamingResponse
-                if isinstance(res, StreamingResponse):
-                    async for data in res.body_iterator:
-                        yield data
-                    return
-                if isinstance(res, dict):
-                    yield f"data: {json.dumps(res)}\n\n"
-                    return
-
-            except Exception as e:
-                log.error(f"Error: {e}")
-                yield f"data: {json.dumps({'error': {'detail':str(e)}})}\n\n"
-                return
-
-            if isinstance(res, str):
-                message = openai_chat_chunk_message_template(form_data["model"], res)
-                yield f"data: {json.dumps(message)}\n\n"
-
-            if isinstance(res, Iterator):
-                for line in res:
-                    yield process_line(form_data, line)
-
-            if isinstance(res, AsyncGenerator):
-                async for line in res:
-                    yield process_line(form_data, line)
-
-            if isinstance(res, str) or isinstance(res, Generator):
-                finish_message = openai_chat_chunk_message_template(
-                    form_data["model"], ""
-                )
-                finish_message["choices"][0]["finish_reason"] = "stop"
-                yield f"data: {json.dumps(finish_message)}\n\n"
-                yield "data: [DONE]"
-
-        return StreamingResponse(stream_content(), media_type="text/event-stream")
-    else:
-        try:
-            res = await execute_pipe(pipe, params)
-
-        except Exception as e:
-            log.error(f"Error: {e}")
-            return {"error": {"detail": str(e)}}
-
-        if isinstance(res, StreamingResponse) or isinstance(res, dict):
-            return res
-        if isinstance(res, BaseModel):
-            return res.model_dump()
-
-        message = await get_message_content(res)
-        return openai_chat_completion_message_template(form_data["model"], message)

backend/open_webui/apps/webui/routers/configs.py

@@ -1,112 +0,0 @@
-from fastapi import APIRouter, Depends, Request
-from pydantic import BaseModel
-
-from typing import Optional
-
-from open_webui.utils.utils import get_admin_user, get_verified_user
-from open_webui.config import get_config, save_config
-from open_webui.config import BannerModel
-
-
-router = APIRouter()
-
-
-############################
-# ImportConfig
-############################
-
-
-class ImportConfigForm(BaseModel):
-    config: dict
-
-
-@router.post("/import", response_model=dict)
-async def import_config(form_data: ImportConfigForm, user=Depends(get_admin_user)):
-    save_config(form_data.config)
-    return get_config()
-
-
-############################
-# ExportConfig
-############################
-
-
-@router.get("/export", response_model=dict)
-async def export_config(user=Depends(get_admin_user)):
-    return get_config()
-
-
-############################
-# SetDefaultModels
-############################
-class ModelsConfigForm(BaseModel):
-    DEFAULT_MODELS: Optional[str]
-    MODEL_ORDER_LIST: Optional[list[str]]
-
-
-@router.get("/models", response_model=ModelsConfigForm)
-async def get_models_config(request: Request, user=Depends(get_admin_user)):
-    return {
-        "DEFAULT_MODELS": request.app.state.config.DEFAULT_MODELS,
-        "MODEL_ORDER_LIST": request.app.state.config.MODEL_ORDER_LIST,
-    }
-
-
-@router.post("/models", response_model=ModelsConfigForm)
-async def set_models_config(
-    request: Request, form_data: ModelsConfigForm, user=Depends(get_admin_user)
-):
-    request.app.state.config.DEFAULT_MODELS = form_data.DEFAULT_MODELS
-    request.app.state.config.MODEL_ORDER_LIST = form_data.MODEL_ORDER_LIST
-    return {
-        "DEFAULT_MODELS": request.app.state.config.DEFAULT_MODELS,
-        "MODEL_ORDER_LIST": request.app.state.config.MODEL_ORDER_LIST,
-    }
-
-
-class PromptSuggestion(BaseModel):
-    title: list[str]
-    content: str
-
-
-class SetDefaultSuggestionsForm(BaseModel):
-    suggestions: list[PromptSuggestion]
-
-
-@router.post("/suggestions", response_model=list[PromptSuggestion])
-async def set_default_suggestions(
-    request: Request,
-    form_data: SetDefaultSuggestionsForm,
-    user=Depends(get_admin_user),
-):
-    data = form_data.model_dump()
-    request.app.state.config.DEFAULT_PROMPT_SUGGESTIONS = data["suggestions"]
-    return request.app.state.config.DEFAULT_PROMPT_SUGGESTIONS
-
-
-############################
-# SetBanners
-############################
-
-
-class SetBannersForm(BaseModel):
-    banners: list[BannerModel]
-
-
-@router.post("/banners", response_model=list[BannerModel])
-async def set_banners(
-    request: Request,
-    form_data: SetBannersForm,
-    user=Depends(get_admin_user),
-):
-    data = form_data.model_dump()
-    request.app.state.config.BANNERS = data["banners"]
-    return request.app.state.config.BANNERS
-
-
-@router.get("/banners", response_model=list[BannerModel])
-async def get_banners(
-    request: Request,
-    user=Depends(get_verified_user),
-):
-    return request.app.state.config.BANNERS

backend/open_webui/apps/webui/routers/files.py

@@ -1,349 +0,0 @@
-import logging
-import os
-import uuid
-from pathlib import Path
-from typing import Optional
-from pydantic import BaseModel
-import mimetypes
-
-from open_webui.storage.provider import Storage
-
-from open_webui.apps.webui.models.files import (
-    FileForm,
-    FileModel,
-    FileModelResponse,
-    Files,
-)
-from open_webui.apps.retrieval.main 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
-from fastapi.responses import FileResponse, StreamingResponse
-
-
-from open_webui.utils.utils import get_admin_user, get_verified_user
-
-log = logging.getLogger(__name__)
-log.setLevel(SRC_LOG_LEVELS["MODELS"])
-
-
-router = APIRouter()
-
-############################
-# Upload File
-############################
-
-
-@router.post("/", response_model=FileModelResponse)
-def upload_file(file: UploadFile = File(...), user=Depends(get_verified_user)):
-    log.info(f"file.content_type: {file.content_type}")
-    try:
-        unsanitized_filename = file.filename
-        filename = os.path.basename(unsanitized_filename)
-
-        # replace filename with uuid
-        id = str(uuid.uuid4())
-        name = filename
-        filename = f"{id}_{filename}"
-        contents, file_path = Storage.upload_file(file.file, filename)
-
-        file_item = Files.insert_new_file(
-            user.id,
-            FileForm(
-                **{
-                    "id": id,
-                    "filename": name,
-                    "path": file_path,
-                    "meta": {
-                        "name": name,
-                        "content_type": file.content_type,
-                        "size": len(contents),
-                    },
-                }
-            ),
-        )
-
-        try:
-            process_file(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),
-                }
-            )
-
-        if file_item:
-            return file_item
-        else:
-            raise HTTPException(
-                status_code=status.HTTP_400_BAD_REQUEST,
-                detail=ERROR_MESSAGES.DEFAULT("Error uploading file"),
-            )
-
-    except Exception as e:
-        log.exception(e)
-        raise HTTPException(
-            status_code=status.HTTP_400_BAD_REQUEST,
-            detail=ERROR_MESSAGES.DEFAULT(e),
-        )
-
-
-############################
-# List Files
-############################
-
-
-@router.get("/", response_model=list[FileModelResponse])
-async def list_files(user=Depends(get_verified_user)):
-    if user.role == "admin":
-        files = Files.get_files()
-    else:
-        files = Files.get_files_by_user_id(user.id)
-    return files
-
-
-############################
-# Delete All Files
-############################
-
-
-@router.delete("/all")
-async def delete_all_files(user=Depends(get_admin_user)):
-    result = Files.delete_all_files()
-    if result:
-        try:
-            Storage.delete_all_files()
-        except Exception as e:
-            log.exception(e)
-            log.error(f"Error deleting files")
-            raise HTTPException(
-                status_code=status.HTTP_400_BAD_REQUEST,
-                detail=ERROR_MESSAGES.DEFAULT("Error deleting files"),
-            )
-        return {"message": "All files deleted successfully"}
-    else:
-        raise HTTPException(
-            status_code=status.HTTP_400_BAD_REQUEST,
-            detail=ERROR_MESSAGES.DEFAULT("Error deleting files"),
-        )
-
-
-############################
-# Get File By Id
-############################
-
-
-@router.get("/{id}", response_model=Optional[FileModel])
-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"):
-        return file
-    else:
-        raise HTTPException(
-            status_code=status.HTTP_404_NOT_FOUND,
-            detail=ERROR_MESSAGES.NOT_FOUND,
-        )
-
-
-############################
-# Get File Data Content By Id
-############################
-
-
-@router.get("/{id}/data/content")
-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"):
-        return {"content": file.data.get("content", "")}
-    else:
-        raise HTTPException(
-            status_code=status.HTTP_404_NOT_FOUND,
-            detail=ERROR_MESSAGES.NOT_FOUND,
-        )
-
-
-############################
-# Update File Data Content By Id
-############################
-
-
-class ContentForm(BaseModel):
-    content: str
-
-
-@router.post("/{id}/data/content/update")
-async def update_file_data_content_by_id(
-    id: str, form_data: ContentForm, user=Depends(get_verified_user)
-):
-    file = Files.get_file_by_id(id)
-
-    if file and (file.user_id == user.id or user.role == "admin"):
-        try:
-            process_file(ProcessFileForm(file_id=id, content=form_data.content))
-            file = Files.get_file_by_id(id=id)
-        except Exception as e:
-            log.exception(e)
-            log.error(f"Error processing file: {file.id}")
-
-        return {"content": file.data.get("content", "")}
-    else:
-        raise HTTPException(
-            status_code=status.HTTP_404_NOT_FOUND,
-            detail=ERROR_MESSAGES.NOT_FOUND,
-        )
-
-
-############################
-# Get File Content By Id
-############################
-
-
-@router.get("/{id}/content")
-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"):
-        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}")
-                headers = {
-                    "Content-Disposition": f'attachment; filename="{file.meta.get("name", file.filename)}"'
-                }
-                return FileResponse(file_path, headers=headers)
-            else:
-                raise HTTPException(
-                    status_code=status.HTTP_404_NOT_FOUND,
-                    detail=ERROR_MESSAGES.NOT_FOUND,
-                )
-        except Exception as e:
-            log.exception(e)
-            log.error(f"Error getting file content")
-            raise HTTPException(
-                status_code=status.HTTP_400_BAD_REQUEST,
-                detail=ERROR_MESSAGES.DEFAULT("Error getting file content"),
-            )
-    else:
-        raise HTTPException(
-            status_code=status.HTTP_404_NOT_FOUND,
-            detail=ERROR_MESSAGES.NOT_FOUND,
-        )
-
-
-@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"):
-        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}")
-                return FileResponse(file_path)
-            else:
-                raise HTTPException(
-                    status_code=status.HTTP_404_NOT_FOUND,
-                    detail=ERROR_MESSAGES.NOT_FOUND,
-                )
-        except Exception as e:
-            log.exception(e)
-            log.error(f"Error getting file content")
-            raise HTTPException(
-                status_code=status.HTTP_400_BAD_REQUEST,
-                detail=ERROR_MESSAGES.DEFAULT("Error getting file content"),
-            )
-    else:
-        raise HTTPException(
-            status_code=status.HTTP_404_NOT_FOUND,
-            detail=ERROR_MESSAGES.NOT_FOUND,
-        )
-
-
-@router.get("/{id}/content/{file_name}")
-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"):
-        file_path = file.path
-        if file_path:
-            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}")
-                headers = {
-                    "Content-Disposition": f'attachment; filename="{file.meta.get("name", file.filename)}"'
-                }
-                return FileResponse(file_path, headers=headers)
-            else:
-                raise HTTPException(
-                    status_code=status.HTTP_404_NOT_FOUND,
-                    detail=ERROR_MESSAGES.NOT_FOUND,
-                )
-        else:
-            # File path doesn’t exist, return the content as .txt if possible
-            file_content = file.content.get("content", "")
-            file_name = file.filename
-
-            # Create a generator that encodes the file content
-            def generator():
-                yield file_content.encode("utf-8")
-
-            return StreamingResponse(
-                generator(),
-                media_type="text/plain",
-                headers={"Content-Disposition": f"attachment; filename={file_name}"},
-            )
-    else:
-        raise HTTPException(
-            status_code=status.HTTP_404_NOT_FOUND,
-            detail=ERROR_MESSAGES.NOT_FOUND,
-        )
-
-
-############################
-# Delete File By Id
-############################
-
-
-@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"):
-        result = Files.delete_file_by_id(id)
-        if result:
-            try:
-                Storage.delete_file(file.filename)
-            except Exception as e:
-                log.exception(e)
-                log.error(f"Error deleting files")
-                raise HTTPException(
-                    status_code=status.HTTP_400_BAD_REQUEST,
-                    detail=ERROR_MESSAGES.DEFAULT("Error deleting files"),
-                )
-            return {"message": "File deleted successfully"}
-        else:
-            raise HTTPException(
-                status_code=status.HTTP_400_BAD_REQUEST,
-                detail=ERROR_MESSAGES.DEFAULT("Error deleting file"),
-            )
-    else:
-        raise HTTPException(
-            status_code=status.HTTP_404_NOT_FOUND,
-            detail=ERROR_MESSAGES.NOT_FOUND,
-        )

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 :/"
@@ -113,6 +114,7 @@ class TASKS(str, Enum):
     TAGS_GENERATION = "tags_generation"
     EMOJI_GENERATION = "emoji_generation"
     QUERY_GENERATION = "query_generation"
+    IMAGE_PROMPT_GENERATION = "image_prompt_generation"
     AUTOCOMPLETE_GENERATION = "autocomplete_generation"
     FUNCTION_CALLING = "function_calling"
     MOA_RESPONSE_GENERATION = "moa_response_generation"

backend/open_webui/env.py

@@ -53,15 +53,20 @@ if USE_CUDA.lower() == "true":
 else:
     DEVICE_TYPE = "cpu"
 
+try:
+    import torch
+
+    if torch.backends.mps.is_available() and torch.backends.mps.is_built():
+        DEVICE_TYPE = "mps"
+except Exception:
+    pass
 
 ####################################
 # 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"
@@ -71,6 +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",
@@ -85,6 +91,7 @@ log_sources = [
     "RAG",
     "WEBHOOK",
     "SOCKET",
+    "OAUTH",
 ]
 
 SRC_LOG_LEVELS = {}
@@ -92,21 +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_URL = os.environ.get("WEBUI_URL", "http://localhost:3000")
-
 WEBUI_FAVICON_URL = "https://openwebui.com/favicon.png"
 
+TRUSTED_SIGNATURE_KEY = os.environ.get("TRUSTED_SIGNATURE_KEY", "")
 
 ####################################
 # ENV (dev,test,prod)
@@ -124,7 +129,6 @@ else:
     except Exception:
         PACKAGE_DATA = {"version": "0.0.0"}
 
-
 VERSION = PACKAGE_DATA["version"]
 
 
@@ -155,7 +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)
 
@@ -186,7 +189,6 @@ for version in soup.find_all("h2"):
 
     changelog_json[version_number] = version_data
 
-
 CHANGELOG = changelog_json
 
 ####################################
@@ -203,7 +205,6 @@ ENABLE_FORWARD_USER_INFO_HEADERS = (
     os.environ.get("ENABLE_FORWARD_USER_INFO_HEADERS", "False").lower() == "true"
 )
 
-
 ####################################
 # WEBUI_BUILD_HASH
 ####################################
@@ -238,7 +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"))
@@ -250,7 +250,6 @@ if FROM_INIT_PY:
         os.getenv("FRONTEND_BUILD_DIR", OPEN_WEBUI_DIR / "frontend")
     ).resolve()
 
-
 ####################################
 # Database
 ####################################
@@ -269,6 +268,8 @@ DATABASE_URL = os.environ.get("DATABASE_URL", f"sqlite:///{DATA_DIR}/webui.db")
 if "postgres://" in DATABASE_URL:
     DATABASE_URL = DATABASE_URL.replace("postgres://", "postgresql://")
 
+DATABASE_SCHEMA = os.environ.get("DATABASE_SCHEMA", None)
+
 DATABASE_POOL_SIZE = os.environ.get("DATABASE_POOL_SIZE", 0)
 
 if DATABASE_POOL_SIZE == "":
@@ -313,11 +314,31 @@ 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"
+)
+
 ####################################
 # 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)
@@ -333,6 +354,10 @@ BYPASS_MODEL_ACCESS_CONTROL = (
     os.environ.get("BYPASS_MODEL_ACCESS_CONTROL", "False").lower() == "true"
 )
 
+WEBUI_AUTH_SIGNOUT_REDIRECT_URL = os.environ.get(
+    "WEBUI_AUTH_SIGNOUT_REDIRECT_URL", None
+)
+
 ####################################
 # WEBUI_SECRET_KEY
 ####################################
@@ -344,14 +369,22 @@ 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 == "":
@@ -364,6 +397,11 @@ 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", "")
 
@@ -375,22 +413,147 @@ else:
     except Exception:
         AIOHTTP_CLIENT_TIMEOUT = 300
 
-AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST = os.environ.get(
-    "AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST", "5"
+
+AIOHTTP_CLIENT_SESSION_SSL = (
+    os.environ.get("AIOHTTP_CLIENT_SESSION_SSL", "True").lower() == "true"
 )
 
-if AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST == "":
-    AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST = None
+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_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
+
+
+AIOHTTP_CLIENT_SESSION_TOOL_SERVER_SSL = (
+    os.environ.get("AIOHTTP_CLIENT_SESSION_TOOL_SERVER_SSL", "True").lower() == "true"
+)
+
+
+####################################
+# SENTENCE TRANSFORMERS
+####################################
+
+
+SENTENCE_TRANSFORMERS_BACKEND = os.environ.get("SENTENCE_TRANSFORMERS_BACKEND", "")
+if SENTENCE_TRANSFORMERS_BACKEND == "":
+    SENTENCE_TRANSFORMERS_BACKEND = "torch"
+
+
+SENTENCE_TRANSFORMERS_MODEL_KWARGS = os.environ.get(
+    "SENTENCE_TRANSFORMERS_MODEL_KWARGS", ""
+)
+if SENTENCE_TRANSFORMERS_MODEL_KWARGS == "":
+    SENTENCE_TRANSFORMERS_MODEL_KWARGS = None
+else:
+    try:
+        SENTENCE_TRANSFORMERS_MODEL_KWARGS = json.loads(
+            SENTENCE_TRANSFORMERS_MODEL_KWARGS
+        )
+    except Exception:
+        SENTENCE_TRANSFORMERS_MODEL_KWARGS = None
+
+
+SENTENCE_TRANSFORMERS_CROSS_ENCODER_BACKEND = os.environ.get(
+    "SENTENCE_TRANSFORMERS_CROSS_ENCODER_BACKEND", ""
+)
+if SENTENCE_TRANSFORMERS_CROSS_ENCODER_BACKEND == "":
+    SENTENCE_TRANSFORMERS_CROSS_ENCODER_BACKEND = "torch"
+
+
+SENTENCE_TRANSFORMERS_CROSS_ENCODER_MODEL_KWARGS = os.environ.get(
+    "SENTENCE_TRANSFORMERS_CROSS_ENCODER_MODEL_KWARGS", ""
+)
+if SENTENCE_TRANSFORMERS_CROSS_ENCODER_MODEL_KWARGS == "":
+    SENTENCE_TRANSFORMERS_CROSS_ENCODER_MODEL_KWARGS = None
+else:
+    try:
+        SENTENCE_TRANSFORMERS_CROSS_ENCODER_MODEL_KWARGS = json.loads(
+            SENTENCE_TRANSFORMERS_CROSS_ENCODER_MODEL_KWARGS
+        )
+    except Exception:
+        SENTENCE_TRANSFORMERS_CROSS_ENCODER_MODEL_KWARGS = None
 
 ####################################
 # OFFLINE_MODE
 ####################################
 
 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

@@ -0,0 +1,322 @@
+import logging
+import sys
+import inspect
+import json
+import asyncio
+
+from pydantic import BaseModel
+from typing import AsyncGenerator, Generator, Iterator
+from fastapi import (
+    Depends,
+    FastAPI,
+    File,
+    Form,
+    HTTPException,
+    Request,
+    UploadFile,
+    status,
+)
+from starlette.responses import Response, StreamingResponse
+
+
+from open_webui.socket.main import (
+    get_event_call,
+    get_event_emitter,
+)
+
+
+from open_webui.models.functions import Functions
+from open_webui.models.models import Models
+
+from open_webui.utils.plugin import load_function_module_by_id
+from open_webui.utils.tools import get_tools
+from open_webui.utils.access_control import has_access
+
+from open_webui.env import SRC_LOG_LEVELS, GLOBAL_LOG_LEVEL
+
+from open_webui.utils.misc import (
+    add_or_update_system_message,
+    get_last_user_message,
+    prepend_to_first_user_message_content,
+    openai_chat_chunk_message_template,
+    openai_chat_completion_message_template,
+)
+from open_webui.utils.payload import (
+    apply_model_params_to_body_openai,
+    apply_model_system_prompt_to_body,
+)
+
+
+logging.basicConfig(stream=sys.stdout, level=GLOBAL_LOG_LEVEL)
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["MAIN"])
+
+
+def get_function_module_by_id(request: Request, pipe_id: str):
+    # Check if function is already loaded
+    if pipe_id not in request.app.state.FUNCTIONS:
+        function_module, _, _ = load_function_module_by_id(pipe_id)
+        request.app.state.FUNCTIONS[pipe_id] = function_module
+    else:
+        function_module = request.app.state.FUNCTIONS[pipe_id]
+
+    if hasattr(function_module, "valves") and hasattr(function_module, "Valves"):
+        valves = Functions.get_function_valves_by_id(pipe_id)
+        function_module.valves = function_module.Valves(**(valves if valves else {}))
+    return function_module
+
+
+async def get_function_models(request):
+    pipes = Functions.get_functions_by_type("pipe", active_only=True)
+    pipe_models = []
+
+    for pipe in pipes:
+        function_module = get_function_module_by_id(request, pipe.id)
+
+        # Check if function is a manifold
+        if hasattr(function_module, "pipes"):
+            sub_pipes = []
+
+            # Handle pipes being a list, sync function, or async function
+            try:
+                if callable(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:
+                log.exception(e)
+                sub_pipes = []
+
+            log.debug(
+                f"get_function_models: function '{pipe.id}' is a manifold of {sub_pipes}"
+            )
+
+            for p in sub_pipes:
+                sub_pipe_id = f'{pipe.id}.{p["id"]}'
+                sub_pipe_name = p["name"]
+
+                if hasattr(function_module, "name"):
+                    sub_pipe_name = f"{function_module.name}{sub_pipe_name}"
+
+                pipe_flag = {"type": pipe.type}
+
+                pipe_models.append(
+                    {
+                        "id": sub_pipe_id,
+                        "name": sub_pipe_name,
+                        "object": "model",
+                        "created": pipe.created_at,
+                        "owned_by": "openai",
+                        "pipe": pipe_flag,
+                    }
+                )
+        else:
+            pipe_flag = {"type": "pipe"}
+
+            log.debug(
+                f"get_function_models: function '{pipe.id}' is a single pipe {{ 'id': {pipe.id}, 'name': {pipe.name} }}"
+            )
+
+            pipe_models.append(
+                {
+                    "id": pipe.id,
+                    "name": pipe.name,
+                    "object": "model",
+                    "created": pipe.created_at,
+                    "owned_by": "openai",
+                    "pipe": pipe_flag,
+                }
+            )
+
+    return pipe_models
+
+
+async def generate_function_chat_completion(
+    request, form_data, user, models: dict = {}
+):
+    async def execute_pipe(pipe, params):
+        if inspect.iscoroutinefunction(pipe):
+            return await pipe(**params)
+        else:
+            return pipe(**params)
+
+    async def get_message_content(res: str | Generator | AsyncGenerator) -> str:
+        if isinstance(res, str):
+            return res
+        if isinstance(res, Generator):
+            return "".join(map(str, res))
+        if isinstance(res, AsyncGenerator):
+            return "".join([str(stream) async for stream in res])
+
+    def process_line(form_data: dict, line):
+        if isinstance(line, BaseModel):
+            line = line.model_dump_json()
+            line = f"data: {line}"
+        if isinstance(line, dict):
+            line = f"data: {json.dumps(line)}"
+
+        try:
+            line = line.decode("utf-8")
+        except Exception:
+            pass
+
+        if line.startswith("data:"):
+            return f"{line}\n\n"
+        else:
+            line = openai_chat_chunk_message_template(form_data["model"], line)
+            return f"data: {json.dumps(line)}\n\n"
+
+    def get_pipe_id(form_data: dict) -> str:
+        pipe_id = form_data["model"]
+        if "." in pipe_id:
+            pipe_id, _ = pipe_id.split(".", 1)
+        return pipe_id
+
+    def get_function_params(function_module, form_data, user, extra_params=None):
+        if extra_params is None:
+            extra_params = {}
+
+        pipe_id = get_pipe_id(form_data)
+
+        # Get the signature of the function
+        sig = inspect.signature(function_module.pipe)
+        params = {"body": form_data} | {
+            k: v for k, v in extra_params.items() if k in sig.parameters
+        }
+
+        if "__user__" in params and hasattr(function_module, "UserValves"):
+            user_valves = Functions.get_user_valves_by_id_and_user_id(pipe_id, user.id)
+            try:
+                params["__user__"]["valves"] = function_module.UserValves(**user_valves)
+            except Exception as e:
+                log.exception(e)
+                params["__user__"]["valves"] = function_module.UserValves()
+
+        return params
+
+    model_id = form_data.get("model")
+    model_info = Models.get_model_by_id(model_id)
+
+    metadata = form_data.pop("metadata", {})
+
+    files = metadata.get("files", [])
+    tool_ids = metadata.get("tool_ids", [])
+    # Check if tool_ids is None
+    if tool_ids is None:
+        tool_ids = []
+
+    __event_emitter__ = None
+    __event_call__ = None
+    __task__ = None
+    __task_body__ = None
+
+    if metadata:
+        if all(k in metadata for k in ("session_id", "chat_id", "message_id")):
+            __event_emitter__ = get_event_emitter(metadata)
+            __event_call__ = get_event_call(metadata)
+        __task__ = metadata.get("task", None)
+        __task_body__ = metadata.get("task_body", None)
+
+    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,
+        "__user__": {
+            "id": user.id,
+            "email": user.email,
+            "name": user.name,
+            "role": user.role,
+        },
+        "__metadata__": metadata,
+        "__request__": request,
+    }
+    extra_params["__tools__"] = get_tools(
+        request,
+        tool_ids,
+        user,
+        {
+            **extra_params,
+            "__model__": models.get(form_data["model"], None),
+            "__messages__": form_data["messages"],
+            "__files__": files,
+        },
+    )
+
+    if model_info:
+        if model_info.base_model_id:
+            form_data["model"] = model_info.base_model_id
+
+        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, metadata, user)
+
+    pipe_id = get_pipe_id(form_data)
+    function_module = get_function_module_by_id(request, pipe_id)
+
+    pipe = function_module.pipe
+    params = get_function_params(function_module, form_data, user, extra_params)
+
+    if form_data.get("stream", False):
+
+        async def stream_content():
+            try:
+                res = await execute_pipe(pipe, params)
+
+                # Directly return if the response is a StreamingResponse
+                if isinstance(res, StreamingResponse):
+                    async for data in res.body_iterator:
+                        yield data
+                    return
+                if isinstance(res, dict):
+                    yield f"data: {json.dumps(res)}\n\n"
+                    return
+
+            except Exception as e:
+                log.error(f"Error: {e}")
+                yield f"data: {json.dumps({'error': {'detail':str(e)}})}\n\n"
+                return
+
+            if isinstance(res, str):
+                message = openai_chat_chunk_message_template(form_data["model"], res)
+                yield f"data: {json.dumps(message)}\n\n"
+
+            if isinstance(res, Iterator):
+                for line in res:
+                    yield process_line(form_data, line)
+
+            if isinstance(res, AsyncGenerator):
+                async for line in res:
+                    yield process_line(form_data, line)
+
+            if isinstance(res, str) or isinstance(res, Generator):
+                finish_message = openai_chat_chunk_message_template(
+                    form_data["model"], ""
+                )
+                finish_message["choices"][0]["finish_reason"] = "stop"
+                yield f"data: {json.dumps(finish_message)}\n\n"
+                yield "data: [DONE]"
+
+        return StreamingResponse(stream_content(), media_type="text/event-stream")
+    else:
+        try:
+            res = await execute_pipe(pipe, params)
+
+        except Exception as e:
+            log.error(f"Error: {e}")
+            return {"error": {"detail": str(e)}}
+
+        if isinstance(res, StreamingResponse) or isinstance(res, dict):
+            return res
+        if isinstance(res, BaseModel):
+            return res.model_dump()
+
+        message = await get_message_content(res)
+        return openai_chat_completion_message_template(form_data["model"], message)

backend/open_webui/internal/db.py

@@ -3,10 +3,11 @@ import logging
 from contextlib import contextmanager
 from typing import Any, Optional
 
-from open_webui.apps.webui.internal.wrappers import register_connection
+from open_webui.internal.wrappers import register_connection
 from open_webui.env import (
     OPEN_WEBUI_DIR,
     DATABASE_URL,
+    DATABASE_SCHEMA,
     SRC_LOG_LEVELS,
     DATABASE_POOL_MAX_OVERFLOW,
     DATABASE_POOL_RECYCLE,
@@ -14,7 +15,7 @@ from open_webui.env import (
     DATABASE_POOL_TIMEOUT,
 )
 from peewee_migrate import Router
-from sqlalchemy import Dialect, create_engine, types
+from sqlalchemy import Dialect, create_engine, MetaData, types
 from sqlalchemy.ext.declarative import declarative_base
 from sqlalchemy.orm import scoped_session, sessionmaker
 from sqlalchemy.pool import QueuePool, NullPool
@@ -54,7 +55,7 @@ def handle_peewee_migration(DATABASE_URL):
     try:
         # Replace the postgresql:// with postgres:// to handle the peewee migration
         db = register_connection(DATABASE_URL.replace("postgresql://", "postgres://"))
-        migrate_dir = OPEN_WEBUI_DIR / "apps" / "webui" / "internal" / "migrations"
+        migrate_dir = OPEN_WEBUI_DIR / "internal" / "migrations"
         router = Router(db, logger=log, migrate_dir=migrate_dir)
         router.run()
         db.close()
@@ -99,7 +100,8 @@ else:
 SessionLocal = sessionmaker(
     autocommit=False, autoflush=False, bind=engine, expire_on_commit=False
 )
-Base = declarative_base()
+metadata_obj = MetaData(schema=DATABASE_SCHEMA)
+Base = declarative_base(metadata=metadata_obj)
 Session = scoped_session(SessionLocal)
 
 

backend/open_webui/migrations/env.py

@@ -1,7 +1,7 @@
 from logging.config import fileConfig
 
 from alembic import context
-from open_webui.apps.webui.models.auths import Auth
+from open_webui.models.auths import Auth
 from open_webui.env import DATABASE_URL
 from sqlalchemy import engine_from_config, pool
 

backend/open_webui/migrations/script.py.mako

@@ -9,7 +9,7 @@ from typing import Sequence, Union
 
 from alembic import op
 import sqlalchemy as sa
-import open_webui.apps.webui.internal.db
+import open_webui.internal.db
 ${imports if imports else ""}
 
 # revision identifiers, used by Alembic.

backend/open_webui/migrations/versions/3781e22d8b01_update_message_table.py

@@ -0,0 +1,70 @@
+"""Update message & channel tables
+
+Revision ID: 3781e22d8b01
+Revises: 7826ab40b532
+Create Date: 2024-12-30 03:00:00.000000
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+
+revision = "3781e22d8b01"
+down_revision = "7826ab40b532"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # Add 'type' column to the 'channel' table
+    op.add_column(
+        "channel",
+        sa.Column(
+            "type",
+            sa.Text(),
+            nullable=True,
+        ),
+    )
+
+    # Add 'parent_id' column to the 'message' table for threads
+    op.add_column(
+        "message",
+        sa.Column("parent_id", sa.Text(), nullable=True),
+    )
+
+    op.create_table(
+        "message_reaction",
+        sa.Column(
+            "id", sa.Text(), nullable=False, primary_key=True, unique=True
+        ),  # Unique reaction ID
+        sa.Column("user_id", sa.Text(), nullable=False),  # User who reacted
+        sa.Column(
+            "message_id", sa.Text(), nullable=False
+        ),  # Message that was reacted to
+        sa.Column(
+            "name", sa.Text(), nullable=False
+        ),  # Reaction name (e.g. "thumbs_up")
+        sa.Column(
+            "created_at", sa.BigInteger(), nullable=True
+        ),  # Timestamp of when the reaction was added
+    )
+
+    op.create_table(
+        "channel_member",
+        sa.Column(
+            "id", sa.Text(), nullable=False, primary_key=True, unique=True
+        ),  # Record ID for the membership row
+        sa.Column("channel_id", sa.Text(), nullable=False),  # Associated channel
+        sa.Column("user_id", sa.Text(), nullable=False),  # Associated user
+        sa.Column(
+            "created_at", sa.BigInteger(), nullable=True
+        ),  # Timestamp of when the user joined the channel
+    )
+
+
+def downgrade():
+    # Revert 'type' column addition to the 'channel' table
+    op.drop_column("channel", "type")
+    op.drop_column("message", "parent_id")
+    op.drop_table("message_reaction")
+    op.drop_table("channel_member")

backend/open_webui/migrations/versions/57c599a3cb57_add_channel_table.py

@@ -0,0 +1,48 @@
+"""Add channel table
+
+Revision ID: 57c599a3cb57
+Revises: 922e7a387820
+Create Date: 2024-12-22 03:00:00.000000
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+
+revision = "57c599a3cb57"
+down_revision = "922e7a387820"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    op.create_table(
+        "channel",
+        sa.Column("id", sa.Text(), nullable=False, primary_key=True, unique=True),
+        sa.Column("user_id", sa.Text()),
+        sa.Column("name", sa.Text()),
+        sa.Column("description", sa.Text(), nullable=True),
+        sa.Column("data", sa.JSON(), nullable=True),
+        sa.Column("meta", sa.JSON(), nullable=True),
+        sa.Column("access_control", sa.JSON(), nullable=True),
+        sa.Column("created_at", sa.BigInteger(), nullable=True),
+        sa.Column("updated_at", sa.BigInteger(), nullable=True),
+    )
+
+    op.create_table(
+        "message",
+        sa.Column("id", sa.Text(), nullable=False, primary_key=True, unique=True),
+        sa.Column("user_id", sa.Text()),
+        sa.Column("channel_id", sa.Text(), nullable=True),
+        sa.Column("content", sa.Text()),
+        sa.Column("data", sa.JSON(), nullable=True),
+        sa.Column("meta", sa.JSON(), nullable=True),
+        sa.Column("created_at", sa.BigInteger(), nullable=True),
+        sa.Column("updated_at", sa.BigInteger(), nullable=True),
+    )
+
+
+def downgrade():
+    op.drop_table("channel")
+
+    op.drop_table("message")

backend/open_webui/migrations/versions/7826ab40b532_update_file_table.py

@@ -0,0 +1,26 @@
+"""Update file table
+
+Revision ID: 7826ab40b532
+Revises: 57c599a3cb57
+Create Date: 2024-12-23 03:00:00.000000
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+
+revision = "7826ab40b532"
+down_revision = "57c599a3cb57"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    op.add_column(
+        "file",
+        sa.Column("access_control", sa.JSON(), nullable=True),
+    )
+
+
+def downgrade():
+    op.drop_column("file", "access_control")

backend/open_webui/migrations/versions/7e5b5dc7342b_init.py

@@ -11,8 +11,8 @@ from typing import Sequence, Union
 import sqlalchemy as sa
 from alembic import op
 
-import open_webui.apps.webui.internal.db
-from open_webui.apps.webui.internal.db import JSONField
+import open_webui.internal.db
+from open_webui.internal.db import JSONField
 from open_webui.migrations.util import get_existing_tables
 
 # revision identifiers, used by Alembic.

backend/open_webui/migrations/versions/9f0c9cd09105_add_note_table.py

@@ -0,0 +1,33 @@
+"""Add note table
+
+Revision ID: 9f0c9cd09105
+Revises: 3781e22d8b01
+Create Date: 2025-05-03 03:00:00.000000
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+
+revision = "9f0c9cd09105"
+down_revision = "3781e22d8b01"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    op.create_table(
+        "note",
+        sa.Column("id", sa.Text(), nullable=False, primary_key=True, unique=True),
+        sa.Column("user_id", sa.Text(), nullable=True),
+        sa.Column("title", sa.Text(), nullable=True),
+        sa.Column("data", sa.JSON(), nullable=True),
+        sa.Column("meta", sa.JSON(), nullable=True),
+        sa.Column("access_control", sa.JSON(), nullable=True),
+        sa.Column("created_at", sa.BigInteger(), nullable=True),
+        sa.Column("updated_at", sa.BigInteger(), nullable=True),
+    )
+
+
+def downgrade():
+    op.drop_table("note")

backend/open_webui/models/auths.py

@@ -2,12 +2,12 @@ import logging
 import uuid
 from typing import Optional
 
-from open_webui.apps.webui.internal.db import Base, get_db
-from open_webui.apps.webui.models.users import UserModel, Users
+from open_webui.internal.db import Base, get_db
+from open_webui.models.users import UserModel, Users
 from open_webui.env import SRC_LOG_LEVELS
 from pydantic import BaseModel
 from sqlalchemy import Boolean, Column, String, Text
-from open_webui.utils.utils import verify_password
+from open_webui.utils.auth import verify_password
 
 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["MODELS"])

backend/open_webui/models/channels.py

@@ -0,0 +1,136 @@
+import json
+import time
+import uuid
+from typing import Optional
+
+from open_webui.internal.db import Base, get_db
+from open_webui.utils.access_control import has_access
+
+from pydantic import BaseModel, ConfigDict
+from sqlalchemy import BigInteger, Boolean, Column, String, Text, JSON
+from sqlalchemy import or_, func, select, and_, text
+from sqlalchemy.sql import exists
+
+####################
+# Channel DB Schema
+####################
+
+
+class Channel(Base):
+    __tablename__ = "channel"
+
+    id = Column(Text, primary_key=True)
+    user_id = Column(Text)
+    type = Column(Text, nullable=True)
+
+    name = Column(Text)
+    description = Column(Text, nullable=True)
+
+    data = Column(JSON, nullable=True)
+    meta = Column(JSON, nullable=True)
+    access_control = Column(JSON, nullable=True)
+
+    created_at = Column(BigInteger)
+    updated_at = Column(BigInteger)
+
+
+class ChannelModel(BaseModel):
+    model_config = ConfigDict(from_attributes=True)
+
+    id: str
+    user_id: str
+    type: Optional[str] = None
+
+    name: str
+    description: Optional[str] = None
+
+    data: Optional[dict] = None
+    meta: Optional[dict] = None
+    access_control: Optional[dict] = None
+
+    created_at: int  # timestamp in epoch
+    updated_at: int  # timestamp in epoch
+
+
+####################
+# Forms
+####################
+
+
+class ChannelForm(BaseModel):
+    name: str
+    description: Optional[str] = None
+    data: Optional[dict] = None
+    meta: Optional[dict] = None
+    access_control: Optional[dict] = None
+
+
+class ChannelTable:
+    def insert_new_channel(
+        self, type: Optional[str], form_data: ChannelForm, user_id: str
+    ) -> Optional[ChannelModel]:
+        with get_db() as db:
+            channel = ChannelModel(
+                **{
+                    **form_data.model_dump(),
+                    "type": type,
+                    "name": form_data.name.lower(),
+                    "id": str(uuid.uuid4()),
+                    "user_id": user_id,
+                    "created_at": int(time.time_ns()),
+                    "updated_at": int(time.time_ns()),
+                }
+            )
+
+            new_channel = Channel(**channel.model_dump())
+
+            db.add(new_channel)
+            db.commit()
+            return channel
+
+    def get_channels(self) -> list[ChannelModel]:
+        with get_db() as db:
+            channels = db.query(Channel).all()
+            return [ChannelModel.model_validate(channel) for channel in channels]
+
+    def get_channels_by_user_id(
+        self, user_id: str, permission: str = "read"
+    ) -> list[ChannelModel]:
+        channels = self.get_channels()
+        return [
+            channel
+            for channel in channels
+            if channel.user_id == user_id
+            or has_access(user_id, permission, channel.access_control)
+        ]
+
+    def get_channel_by_id(self, id: str) -> Optional[ChannelModel]:
+        with get_db() as db:
+            channel = db.query(Channel).filter(Channel.id == id).first()
+            return ChannelModel.model_validate(channel) if channel else None
+
+    def update_channel_by_id(
+        self, id: str, form_data: ChannelForm
+    ) -> Optional[ChannelModel]:
+        with get_db() as db:
+            channel = db.query(Channel).filter(Channel.id == id).first()
+            if not channel:
+                return None
+
+            channel.name = form_data.name
+            channel.data = form_data.data
+            channel.meta = form_data.meta
+            channel.access_control = form_data.access_control
+            channel.updated_at = int(time.time_ns())
+
+            db.commit()
+            return ChannelModel.model_validate(channel) if channel else None
+
+    def delete_channel_by_id(self, id: str):
+        with get_db() as db:
+            db.query(Channel).filter(Channel.id == id).delete()
+            db.commit()
+            return True
+
+
+Channels = ChannelTable()

backend/open_webui/models/chats.py

@@ -1,11 +1,12 @@
+import logging
 import json
 import time
 import uuid
 from typing import Optional
 
-from open_webui.apps.webui.internal.db import Base, get_db
-from open_webui.apps.webui.models.tags import TagModel, Tag, Tags
-
+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"
@@ -168,6 +172,100 @@ class ChatTable:
         except Exception:
             return None
 
+    def update_chat_title_by_id(self, id: str, title: str) -> Optional[ChatModel]:
+        chat = self.get_chat_by_id(id)
+        if chat is None:
+            return None
+
+        chat = chat.chat
+        chat["title"] = title
+
+        return self.update_chat_by_id(id, chat)
+
+    def update_chat_tags_by_id(
+        self, id: str, tags: list[str], user
+    ) -> Optional[ChatModel]:
+        chat = self.get_chat_by_id(id)
+        if chat is None:
+            return None
+
+        self.delete_all_tags_by_id_and_user_id(id, user.id)
+
+        for tag in chat.meta.get("tags", []):
+            if self.count_chats_by_tag_name_and_user_id(tag, user.id) == 0:
+                Tags.delete_tag_by_name_and_user_id(tag, user.id)
+
+        for tag_name in tags:
+            if tag_name.lower() == "none":
+                continue
+
+            self.add_chat_tag_by_id_and_user_id_and_tag_name(id, user.id, tag_name)
+        return self.get_chat_by_id(id)
+
+    def get_chat_title_by_id(self, id: str) -> Optional[str]:
+        chat = self.get_chat_by_id(id)
+        if chat is None:
+            return None
+
+        return chat.chat.get("title", "New Chat")
+
+    def get_messages_by_chat_id(self, id: str) -> Optional[dict]:
+        chat = self.get_chat_by_id(id)
+        if chat is None:
+            return None
+
+        return chat.chat.get("history", {}).get("messages", {}) or {}
+
+    def get_message_by_id_and_message_id(
+        self, id: str, message_id: str
+    ) -> Optional[dict]:
+        chat = self.get_chat_by_id(id)
+        if chat is None:
+            return None
+
+        return chat.chat.get("history", {}).get("messages", {}).get(message_id, {})
+
+    def upsert_message_to_chat_by_id_and_message_id(
+        self, id: str, message_id: str, message: dict
+    ) -> Optional[ChatModel]:
+        chat = self.get_chat_by_id(id)
+        if chat is None:
+            return None
+
+        chat = chat.chat
+        history = chat.get("history", {})
+
+        if message_id in history.get("messages", {}):
+            history["messages"][message_id] = {
+                **history["messages"][message_id],
+                **message,
+            }
+        else:
+            history["messages"][message_id] = message
+
+        history["currentId"] = message_id
+
+        chat["history"] = history
+        return self.update_chat_by_id(id, chat)
+
+    def add_message_status_to_chat_by_id_and_message_id(
+        self, id: str, message_id: str, status: dict
+    ) -> Optional[ChatModel]:
+        chat = self.get_chat_by_id(id)
+        if chat is None:
+            return None
+
+        chat = chat.chat
+        history = chat.get("history", {})
+
+        if message_id in history.get("messages", {}):
+            status_history = history["messages"][message_id].get("statusHistory", [])
+            status_history.append(status)
+            history["messages"][message_id]["statusHistory"] = status_history
+
+        chat["history"] = history
+        return self.update_chat_by_id(id, chat)
+
     def insert_shared_chat_by_chat_id(self, chat_id: str) -> Optional[ChatModel]:
         with get_db() as db:
             # Get the existing chat to share
@@ -299,7 +397,7 @@ class ChatTable:
         limit: int = 50,
     ) -> list[ChatModel]:
         with get_db() as db:
-            query = db.query(Chat).filter_by(user_id=user_id).filter_by(folder_id=None)
+            query = db.query(Chat).filter_by(user_id=user_id)
             if not include_archived:
                 query = query.filter_by(archived=False)
 
@@ -338,7 +436,7 @@ class ChatTable:
 
             all_chats = query.all()
 
-            # result has to be destrctured from sqlalchemy `row` and mapped to a dict since the `ChatModel`is not the returned dataclass.
+            # result has to be destructured from sqlalchemy `row` and mapped to a dict since the `ChatModel`is not the returned dataclass.
             return [
                 ChatTitleIdResponse.model_validate(
                     {
@@ -375,6 +473,8 @@ class ChatTable:
     def get_chat_by_share_id(self, id: str) -> Optional[ChatModel]:
         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 checking if a chat with the share_id exists
                 chat = db.query(Chat).filter_by(share_id=id).first()
 
                 if chat:
@@ -574,7 +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]
@@ -635,7 +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(
@@ -656,7 +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(
@@ -714,7 +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

@@ -3,8 +3,8 @@ import time
 import uuid
 from typing import Optional
 
-from open_webui.apps.webui.internal.db import Base, get_db
-from open_webui.apps.webui.models.chats import Chats
+from open_webui.internal.db import Base, get_db
+from open_webui.models.chats import Chats
 
 from open_webui.env import SRC_LOG_LEVELS
 from pydantic import BaseModel, ConfigDict
@@ -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

@@ -2,7 +2,7 @@ import logging
 import time
 from typing import Optional
 
-from open_webui.apps.webui.internal.db import Base, JSONField, get_db
+from open_webui.internal.db import Base, JSONField, get_db
 from open_webui.env import SRC_LOG_LEVELS
 from pydantic import BaseModel, ConfigDict
 from sqlalchemy import BigInteger, Column, String, Text, JSON
@@ -27,6 +27,8 @@ class File(Base):
     data = Column(JSON, nullable=True)
     meta = Column(JSON, nullable=True)
 
+    access_control = Column(JSON, nullable=True)
+
     created_at = Column(BigInteger)
     updated_at = Column(BigInteger)
 
@@ -44,6 +46,8 @@ class FileModel(BaseModel):
     data: Optional[dict] = None
     meta: Optional[dict] = None
 
+    access_control: Optional[dict] = None
+
     created_at: Optional[int]  # timestamp in epoch
     updated_at: Optional[int]  # timestamp in epoch
 
@@ -90,6 +94,7 @@ class FileForm(BaseModel):
     path: str
     data: dict = {}
     meta: dict = {}
+    access_control: Optional[dict] = None
 
 
 class FilesTable:
@@ -114,7 +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

@@ -3,12 +3,14 @@ import time
 import uuid
 from typing import Optional
 
-from open_webui.apps.webui.internal.db import Base, get_db
-from open_webui.apps.webui.models.chats import Chats
+from open_webui.internal.db import Base, get_db
+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

@@ -2,8 +2,8 @@ import logging
 import time
 from typing import Optional
 
-from open_webui.apps.webui.internal.db import Base, JSONField, get_db
-from open_webui.apps.webui.models.users import Users
+from open_webui.internal.db import Base, JSONField, get_db
+from open_webui.models.users import Users
 from open_webui.env import SRC_LOG_LEVELS
 from pydantic import BaseModel, ConfigDict
 from sqlalchemy import BigInteger, Boolean, Column, String, Text
@@ -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/groups.py

@@ -4,10 +4,10 @@ import time
 from typing import Optional
 import uuid
 
-from open_webui.apps.webui.internal.db import Base, get_db
+from open_webui.internal.db import Base, get_db
 from open_webui.env import SRC_LOG_LEVELS
 
-from open_webui.apps.webui.models.files import FileMetadataResponse
+from open_webui.models.files import FileMetadataResponse
 
 
 from pydantic import BaseModel, ConfigDict
@@ -80,12 +80,11 @@ class GroupResponse(BaseModel):
 class GroupForm(BaseModel):
     name: str
     description: str
+    permissions: Optional[dict] = None
 
 
 class GroupUpdateForm(GroupForm):
-    permissions: Optional[dict] = None
     user_ids: Optional[list[str]] = None
-    admin_ids: Optional[list[str]] = None
 
 
 class GroupTable:
@@ -95,7 +94,7 @@ class GroupTable:
         with get_db() as db:
             group = GroupModel(
                 **{
-                    **form_data.model_dump(),
+                    **form_data.model_dump(exclude_none=True),
                     "id": str(uuid.uuid4()),
                     "user_id": user_id,
                     "created_at": int(time.time()),
@@ -146,6 +145,13 @@ class GroupTable:
         except Exception:
             return None
 
+    def get_group_user_ids_by_id(self, id: str) -> Optional[str]:
+        group = self.get_group_by_id(id)
+        if group:
+            return group.user_ids
+        else:
+            return None
+
     def update_group_by_id(
         self, id: str, form_data: GroupUpdateForm, overwrite: bool = False
     ) -> Optional[GroupModel]:
@@ -182,5 +188,24 @@ class GroupTable:
             except Exception:
                 return False
 
+    def remove_user_from_all_groups(self, user_id: str) -> bool:
+        with get_db() as db:
+            try:
+                groups = self.get_groups_by_member_id(user_id)
+
+                for group in groups:
+                    group.user_ids.remove(user_id)
+                    db.query(Group).filter_by(id=group.id).update(
+                        {
+                            "user_ids": group.user_ids,
+                            "updated_at": int(time.time()),
+                        }
+                    )
+                    db.commit()
+
+                return True
+            except Exception:
+                return False
+
 
 Groups = GroupTable()

backend/open_webui/models/knowledge.py

@@ -4,11 +4,11 @@ import time
 from typing import Optional
 import uuid
 
-from open_webui.apps.webui.internal.db import Base, get_db
+from open_webui.internal.db import Base, get_db
 from open_webui.env import SRC_LOG_LEVELS
 
-from open_webui.apps.webui.models.files import FileMetadataResponse
-from open_webui.apps.webui.models.users import Users, UserResponse
+from open_webui.models.files import FileMetadataResponse
+from open_webui.models.users import Users, UserResponse
 
 
 from pydantic import BaseModel, ConfigDict

backend/open_webui/models/memories.py

@@ -2,7 +2,7 @@ import time
 import uuid
 from typing import Optional
 
-from open_webui.apps.webui.internal.db import Base, get_db
+from open_webui.internal.db import Base, get_db
 from pydantic import BaseModel, ConfigDict
 from sqlalchemy import BigInteger, Column, String, Text
 
@@ -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/messages.py

@@ -0,0 +1,279 @@
+import json
+import time
+import uuid
+from typing import Optional
+
+from open_webui.internal.db import Base, get_db
+from open_webui.models.tags import TagModel, Tag, Tags
+
+
+from pydantic import BaseModel, ConfigDict
+from sqlalchemy import BigInteger, Boolean, Column, String, Text, JSON
+from sqlalchemy import or_, func, select, and_, text
+from sqlalchemy.sql import exists
+
+####################
+# Message DB Schema
+####################
+
+
+class MessageReaction(Base):
+    __tablename__ = "message_reaction"
+    id = Column(Text, primary_key=True)
+    user_id = Column(Text)
+    message_id = Column(Text)
+    name = Column(Text)
+    created_at = Column(BigInteger)
+
+
+class MessageReactionModel(BaseModel):
+    model_config = ConfigDict(from_attributes=True)
+
+    id: str
+    user_id: str
+    message_id: str
+    name: str
+    created_at: int  # timestamp in epoch
+
+
+class Message(Base):
+    __tablename__ = "message"
+    id = Column(Text, primary_key=True)
+
+    user_id = Column(Text)
+    channel_id = Column(Text, nullable=True)
+
+    parent_id = Column(Text, nullable=True)
+
+    content = Column(Text)
+    data = Column(JSON, nullable=True)
+    meta = Column(JSON, nullable=True)
+
+    created_at = Column(BigInteger)  # time_ns
+    updated_at = Column(BigInteger)  # time_ns
+
+
+class MessageModel(BaseModel):
+    model_config = ConfigDict(from_attributes=True)
+
+    id: str
+    user_id: str
+    channel_id: Optional[str] = None
+
+    parent_id: Optional[str] = None
+
+    content: str
+    data: Optional[dict] = None
+    meta: Optional[dict] = None
+
+    created_at: int  # timestamp in epoch
+    updated_at: int  # timestamp in epoch
+
+
+####################
+# Forms
+####################
+
+
+class MessageForm(BaseModel):
+    content: str
+    parent_id: Optional[str] = None
+    data: Optional[dict] = None
+    meta: Optional[dict] = None
+
+
+class Reactions(BaseModel):
+    name: str
+    user_ids: list[str]
+    count: int
+
+
+class MessageResponse(MessageModel):
+    latest_reply_at: Optional[int]
+    reply_count: int
+    reactions: list[Reactions]
+
+
+class MessageTable:
+    def insert_new_message(
+        self, form_data: MessageForm, channel_id: str, user_id: str
+    ) -> Optional[MessageModel]:
+        with get_db() as db:
+            id = str(uuid.uuid4())
+
+            ts = int(time.time_ns())
+            message = MessageModel(
+                **{
+                    "id": id,
+                    "user_id": user_id,
+                    "channel_id": channel_id,
+                    "parent_id": form_data.parent_id,
+                    "content": form_data.content,
+                    "data": form_data.data,
+                    "meta": form_data.meta,
+                    "created_at": ts,
+                    "updated_at": ts,
+                }
+            )
+
+            result = Message(**message.model_dump())
+            db.add(result)
+            db.commit()
+            db.refresh(result)
+            return MessageModel.model_validate(result) if result else None
+
+    def get_message_by_id(self, id: str) -> Optional[MessageResponse]:
+        with get_db() as db:
+            message = db.get(Message, id)
+            if not message:
+                return None
+
+            reactions = self.get_reactions_by_message_id(id)
+            replies = self.get_replies_by_message_id(id)
+
+            return MessageResponse(
+                **{
+                    **MessageModel.model_validate(message).model_dump(),
+                    "latest_reply_at": replies[0].created_at if replies else None,
+                    "reply_count": len(replies),
+                    "reactions": reactions,
+                }
+            )
+
+    def get_replies_by_message_id(self, id: str) -> list[MessageModel]:
+        with get_db() as db:
+            all_messages = (
+                db.query(Message)
+                .filter_by(parent_id=id)
+                .order_by(Message.created_at.desc())
+                .all()
+            )
+            return [MessageModel.model_validate(message) for message in all_messages]
+
+    def get_reply_user_ids_by_message_id(self, id: str) -> list[str]:
+        with get_db() as db:
+            return [
+                message.user_id
+                for message in db.query(Message).filter_by(parent_id=id).all()
+            ]
+
+    def get_messages_by_channel_id(
+        self, channel_id: str, skip: int = 0, limit: int = 50
+    ) -> list[MessageModel]:
+        with get_db() as db:
+            all_messages = (
+                db.query(Message)
+                .filter_by(channel_id=channel_id, parent_id=None)
+                .order_by(Message.created_at.desc())
+                .offset(skip)
+                .limit(limit)
+                .all()
+            )
+            return [MessageModel.model_validate(message) for message in all_messages]
+
+    def get_messages_by_parent_id(
+        self, channel_id: str, parent_id: str, skip: int = 0, limit: int = 50
+    ) -> list[MessageModel]:
+        with get_db() as db:
+            message = db.get(Message, parent_id)
+
+            if not message:
+                return []
+
+            all_messages = (
+                db.query(Message)
+                .filter_by(channel_id=channel_id, parent_id=parent_id)
+                .order_by(Message.created_at.desc())
+                .offset(skip)
+                .limit(limit)
+                .all()
+            )
+
+            # If length of all_messages is less than limit, then add the parent message
+            if len(all_messages) < limit:
+                all_messages.append(message)
+
+            return [MessageModel.model_validate(message) for message in all_messages]
+
+    def update_message_by_id(
+        self, id: str, form_data: MessageForm
+    ) -> Optional[MessageModel]:
+        with get_db() as db:
+            message = db.get(Message, id)
+            message.content = form_data.content
+            message.data = form_data.data
+            message.meta = form_data.meta
+            message.updated_at = int(time.time_ns())
+            db.commit()
+            db.refresh(message)
+            return MessageModel.model_validate(message) if message else None
+
+    def add_reaction_to_message(
+        self, id: str, user_id: str, name: str
+    ) -> Optional[MessageReactionModel]:
+        with get_db() as db:
+            reaction_id = str(uuid.uuid4())
+            reaction = MessageReactionModel(
+                id=reaction_id,
+                user_id=user_id,
+                message_id=id,
+                name=name,
+                created_at=int(time.time_ns()),
+            )
+            result = MessageReaction(**reaction.model_dump())
+            db.add(result)
+            db.commit()
+            db.refresh(result)
+            return MessageReactionModel.model_validate(result) if result else None
+
+    def get_reactions_by_message_id(self, id: str) -> list[Reactions]:
+        with get_db() as db:
+            all_reactions = db.query(MessageReaction).filter_by(message_id=id).all()
+
+            reactions = {}
+            for reaction in all_reactions:
+                if reaction.name not in reactions:
+                    reactions[reaction.name] = {
+                        "name": reaction.name,
+                        "user_ids": [],
+                        "count": 0,
+                    }
+                reactions[reaction.name]["user_ids"].append(reaction.user_id)
+                reactions[reaction.name]["count"] += 1
+
+            return [Reactions(**reaction) for reaction in reactions.values()]
+
+    def remove_reaction_by_id_and_user_id_and_name(
+        self, id: str, user_id: str, name: str
+    ) -> bool:
+        with get_db() as db:
+            db.query(MessageReaction).filter_by(
+                message_id=id, user_id=user_id, name=name
+            ).delete()
+            db.commit()
+            return True
+
+    def delete_reactions_by_id(self, id: str) -> bool:
+        with get_db() as db:
+            db.query(MessageReaction).filter_by(message_id=id).delete()
+            db.commit()
+            return True
+
+    def delete_replies_by_id(self, id: str) -> bool:
+        with get_db() as db:
+            db.query(Message).filter_by(parent_id=id).delete()
+            db.commit()
+            return True
+
+    def delete_message_by_id(self, id: str) -> bool:
+        with get_db() as db:
+            db.query(Message).filter_by(id=id).delete()
+
+            # Delete all reactions to this message
+            db.query(MessageReaction).filter_by(message_id=id).delete()
+
+            db.commit()
+            return True
+
+
+Messages = MessageTable()

backend/open_webui/models/models.py

@@ -2,10 +2,10 @@ import logging
 import time
 from typing import Optional
 
-from open_webui.apps.webui.internal.db import Base, JSONField, get_db
+from open_webui.internal.db import Base, JSONField, get_db
 from open_webui.env import SRC_LOG_LEVELS
 
-from open_webui.apps.webui.models.users import Users, UserResponse
+from open_webui.models.users import Users, UserResponse
 
 
 from pydantic import BaseModel, ConfigDict
@@ -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/notes.py

@@ -0,0 +1,135 @@
+import json
+import time
+import uuid
+from typing import Optional
+
+from open_webui.internal.db import Base, get_db
+from open_webui.utils.access_control import has_access
+from open_webui.models.users import Users, UserResponse
+
+
+from pydantic import BaseModel, ConfigDict
+from sqlalchemy import BigInteger, Boolean, Column, String, Text, JSON
+from sqlalchemy import or_, func, select, and_, text
+from sqlalchemy.sql import exists
+
+####################
+# Note DB Schema
+####################
+
+
+class Note(Base):
+    __tablename__ = "note"
+
+    id = Column(Text, primary_key=True)
+    user_id = Column(Text)
+
+    title = Column(Text)
+    data = Column(JSON, nullable=True)
+    meta = Column(JSON, nullable=True)
+
+    access_control = Column(JSON, nullable=True)
+
+    created_at = Column(BigInteger)
+    updated_at = Column(BigInteger)
+
+
+class NoteModel(BaseModel):
+    model_config = ConfigDict(from_attributes=True)
+
+    id: str
+    user_id: str
+
+    title: str
+    data: Optional[dict] = None
+    meta: Optional[dict] = None
+
+    access_control: Optional[dict] = None
+
+    created_at: int  # timestamp in epoch
+    updated_at: int  # timestamp in epoch
+
+
+####################
+# Forms
+####################
+
+
+class NoteForm(BaseModel):
+    title: str
+    data: Optional[dict] = None
+    meta: Optional[dict] = None
+    access_control: Optional[dict] = None
+
+
+class NoteUserResponse(NoteModel):
+    user: Optional[UserResponse] = None
+
+
+class NoteTable:
+    def insert_new_note(
+        self,
+        form_data: NoteForm,
+        user_id: str,
+    ) -> Optional[NoteModel]:
+        with get_db() as db:
+            note = NoteModel(
+                **{
+                    "id": str(uuid.uuid4()),
+                    "user_id": user_id,
+                    **form_data.model_dump(),
+                    "created_at": int(time.time_ns()),
+                    "updated_at": int(time.time_ns()),
+                }
+            )
+
+            new_note = Note(**note.model_dump())
+
+            db.add(new_note)
+            db.commit()
+            return note
+
+    def get_notes(self) -> list[NoteModel]:
+        with get_db() as db:
+            notes = db.query(Note).order_by(Note.updated_at.desc()).all()
+            return [NoteModel.model_validate(note) for note in notes]
+
+    def get_notes_by_user_id(
+        self, user_id: str, permission: str = "write"
+    ) -> list[NoteModel]:
+        notes = self.get_notes()
+        return [
+            note
+            for note in notes
+            if note.user_id == user_id
+            or has_access(user_id, permission, note.access_control)
+        ]
+
+    def get_note_by_id(self, id: str) -> Optional[NoteModel]:
+        with get_db() as db:
+            note = db.query(Note).filter(Note.id == id).first()
+            return NoteModel.model_validate(note) if note else None
+
+    def update_note_by_id(self, id: str, form_data: NoteForm) -> Optional[NoteModel]:
+        with get_db() as db:
+            note = db.query(Note).filter(Note.id == id).first()
+            if not note:
+                return None
+
+            note.title = form_data.title
+            note.data = form_data.data
+            note.meta = form_data.meta
+            note.access_control = form_data.access_control
+            note.updated_at = int(time.time_ns())
+
+            db.commit()
+            return NoteModel.model_validate(note) if note else None
+
+    def delete_note_by_id(self, id: str):
+        with get_db() as db:
+            db.query(Note).filter(Note.id == id).delete()
+            db.commit()
+            return True
+
+
+Notes = NoteTable()

backend/open_webui/models/prompts.py

@@ -1,8 +1,8 @@
 import time
 from typing import Optional
 
-from open_webui.apps.webui.internal.db import Base, get_db
-from open_webui.apps.webui.models.users import Users, UserResponse
+from open_webui.internal.db import Base, get_db
+from open_webui.models.users import Users, UserResponse
 
 from pydantic import BaseModel, ConfigDict
 from sqlalchemy import BigInteger, Column, String, Text, JSON

backend/open_webui/models/tags.py

@@ -3,7 +3,7 @@ import time
 import uuid
 from typing import Optional
 
-from open_webui.apps.webui.internal.db import Base, get_db
+from open_webui.internal.db import Base, get_db
 
 
 from open_webui.env import SRC_LOG_LEVELS
@@ -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

@@ -2,8 +2,8 @@ import logging
 import time
 from typing import Optional
 
-from open_webui.apps.webui.internal.db import Base, JSONField, get_db
-from open_webui.apps.webui.models.users import Users, UserResponse
+from open_webui.internal.db import Base, JSONField, get_db
+from open_webui.models.users import Users, UserResponse
 from open_webui.env import SRC_LOG_LEVELS
 from pydantic import BaseModel, ConfigDict
 from sqlalchemy import BigInteger, Column, String, Text, JSON
@@ -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

@@ -1,10 +1,17 @@
 import time
 from typing import Optional
 
-from open_webui.apps.webui.internal.db import Base, JSONField, get_db
-from open_webui.apps.webui.models.chats import Chats
+from open_webui.internal.db import Base, JSONField, get_db
+
+
+from open_webui.models.chats import Chats
+from open_webui.models.groups import Groups
+
+
 from pydantic import BaseModel, ConfigDict
 from sqlalchemy import BigInteger, Column, String, Text
+from sqlalchemy import or_
+
 
 ####################
 # User DB Schema
@@ -62,6 +69,11 @@ class UserModel(BaseModel):
 ####################
 
 
+class UserListResponse(BaseModel):
+    users: list[UserModel]
+    total: int
+
+
 class UserResponse(BaseModel):
     id: str
     name: str
@@ -70,6 +82,13 @@ class UserResponse(BaseModel):
     profile_image_url: str
 
 
+class UserNameResponse(BaseModel):
+    id: str
+    name: str
+    role: str
+    profile_image_url: str
+
+
 class UserRoleUpdateForm(BaseModel):
     id: str
     role: str
@@ -147,13 +166,79 @@ class UsersTable:
         except Exception:
             return None
 
-    def get_users(self, skip: int = 0, limit: int = 50) -> list[UserModel]:
+    def get_users(
+        self,
+        filter: Optional[dict] = None,
+        skip: Optional[int] = None,
+        limit: Optional[int] = None,
+    ) -> UserListResponse:
         with get_db() as db:
-            users = (
-                db.query(User)
-                # .offset(skip).limit(limit)
-                .all()
-            )
+            query = db.query(User)
+
+            if filter:
+                query_key = filter.get("query")
+                if query_key:
+                    query = query.filter(
+                        or_(
+                            User.name.ilike(f"%{query_key}%"),
+                            User.email.ilike(f"%{query_key}%"),
+                        )
+                    )
+
+                order_by = filter.get("order_by")
+                direction = filter.get("direction")
+
+                if order_by == "name":
+                    if direction == "asc":
+                        query = query.order_by(User.name.asc())
+                    else:
+                        query = query.order_by(User.name.desc())
+                elif order_by == "email":
+                    if direction == "asc":
+                        query = query.order_by(User.email.asc())
+                    else:
+                        query = query.order_by(User.email.desc())
+
+                elif order_by == "created_at":
+                    if direction == "asc":
+                        query = query.order_by(User.created_at.asc())
+                    else:
+                        query = query.order_by(User.created_at.desc())
+
+                elif order_by == "last_active_at":
+                    if direction == "asc":
+                        query = query.order_by(User.last_active_at.asc())
+                    else:
+                        query = query.order_by(User.last_active_at.desc())
+
+                elif order_by == "updated_at":
+                    if direction == "asc":
+                        query = query.order_by(User.updated_at.asc())
+                    else:
+                        query = query.order_by(User.updated_at.desc())
+                elif order_by == "role":
+                    if direction == "asc":
+                        query = query.order_by(User.role.asc())
+                    else:
+                        query = query.order_by(User.role.desc())
+
+            else:
+                query = query.order_by(User.created_at.desc())
+
+            if skip:
+                query = query.offset(skip)
+            if limit:
+                query = query.limit(limit)
+
+            users = query.all()
+            return {
+                "users": [UserModel.model_validate(user) for user in users],
+                "total": db.query(User).count(),
+            }
+
+    def get_users_by_user_ids(self, user_ids: list[str]) -> list[UserModel]:
+        with get_db() as db:
+            users = db.query(User).filter(User.id.in_(user_ids)).all()
             return [UserModel.model_validate(user) for user in users]
 
     def get_num_users(self) -> Optional[int]:
@@ -168,6 +253,22 @@ class UsersTable:
         except Exception:
             return None
 
+    def get_user_webhook_url_by_id(self, id: str) -> Optional[str]:
+        try:
+            with get_db() as db:
+                user = db.query(User).filter_by(id=id).first()
+
+                if user.settings is None:
+                    return None
+                else:
+                    return (
+                        user.settings.get("ui", {})
+                        .get("notifications", {})
+                        .get("webhook_url", None)
+                    )
+        except Exception:
+            return None
+
     def update_user_role_by_id(self, id: str, role: str) -> Optional[UserModel]:
         try:
             with get_db() as db:
@@ -231,11 +332,31 @@ 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
+            Groups.remove_user_from_all_groups(id)
+
             # Delete User Chats
             result = Chats.delete_chats_by_user_id(id)
-
             if result:
                 with get_db() as db:
                     # Delete User
@@ -265,5 +386,18 @@ class UsersTable:
         except Exception:
             return None
 
+    def get_valid_user_ids(self, user_ids: list[str]) -> list[str]:
+        with get_db() as db:
+            users = db.query(User).filter(User.id.in_(user_ids)).all()
+            return [user.id for user in users]
+
+    def get_super_admin_user(self) -> Optional[UserModel]:
+        with get_db() as db:
+            user = db.query(User).filter_by(role="admin").first()
+            if user:
+                return UserModel.model_validate(user)
+            else:
+                return None
+
 
 Users = UsersTable()

backend/open_webui/retrieval/loaders/external.py

@@ -0,0 +1,53 @@
+import requests
+import logging
+from typing import Iterator, List, 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 ExternalLoader(BaseLoader):
+    def __init__(
+        self,
+        web_paths: Union[str, List[str]],
+        external_url: str,
+        external_api_key: str,
+        continue_on_failure: bool = True,
+        **kwargs,
+    ) -> None:
+        self.external_url = external_url
+        self.external_api_key = external_api_key
+        self.urls = web_paths if isinstance(web_paths, list) else [web_paths]
+        self.continue_on_failure = continue_on_failure
+
+    def lazy_load(self) -> Iterator[Document]:
+        batch_size = 20
+        for i in range(0, len(self.urls), batch_size):
+            urls = self.urls[i : i + batch_size]
+            try:
+                response = requests.post(
+                    self.external_url,
+                    headers={
+                        "User-Agent": "Open WebUI (https://github.com/open-webui/open-webui) RAG Bot",
+                        "Authorization": f"Bearer {self.external_api_key}",
+                    },
+                    json={
+                        "urls": urls,
+                    },
+                )
+                response.raise_for_status()
+                results = response.json()
+                for result in results:
+                    yield Document(
+                        page_content=result.get("page_content", ""),
+                        metadata=result.get("metadata", {}),
+                    )
+            except Exception as e:
+                if self.continue_on_failure:
+                    log.error(f"Error extracting content from batch {urls}: {e}")
+                else:
+                    raise e

backend/open_webui/retrieval/loaders/main.py

@@ -0,0 +1,301 @@
+import requests
+import logging
+import ftfy
+import sys
+
+from langchain_community.document_loaders import (
+    AzureAIDocumentIntelligenceLoader,
+    BSHTMLLoader,
+    CSVLoader,
+    Docx2txtLoader,
+    OutlookMessageLoader,
+    PyPDFLoader,
+    TextLoader,
+    UnstructuredEPubLoader,
+    UnstructuredExcelLoader,
+    UnstructuredMarkdownLoader,
+    UnstructuredPowerPointLoader,
+    UnstructuredRSTLoader,
+    UnstructuredXMLLoader,
+    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)
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
+known_source_ext = [
+    "go",
+    "py",
+    "java",
+    "sh",
+    "bat",
+    "ps1",
+    "cmd",
+    "js",
+    "ts",
+    "css",
+    "cpp",
+    "hpp",
+    "h",
+    "c",
+    "cs",
+    "sql",
+    "log",
+    "ini",
+    "pl",
+    "pm",
+    "r",
+    "dart",
+    "dockerfile",
+    "env",
+    "php",
+    "hs",
+    "hsc",
+    "lua",
+    "nginxconf",
+    "conf",
+    "m",
+    "mm",
+    "plsql",
+    "perl",
+    "rb",
+    "rs",
+    "db2",
+    "scala",
+    "bash",
+    "swift",
+    "vue",
+    "svelte",
+    "msg",
+    "ex",
+    "exs",
+    "erl",
+    "tsx",
+    "jsx",
+    "hs",
+    "lhs",
+    "json",
+]
+
+
+class TikaLoader:
+    def __init__(self, url, file_path, mime_type=None, extract_images=None):
+        self.url = url
+        self.file_path = file_path
+        self.mime_type = mime_type
+
+        self.extract_images = extract_images
+
+    def load(self) -> list[Document]:
+        with open(self.file_path, "rb") as f:
+            data = f.read()
+
+        if self.mime_type is not None:
+            headers = {"Content-Type": self.mime_type}
+        else:
+            headers = {}
+
+        if self.extract_images == True:
+            headers["X-Tika-PDFextractInlineImages"] = "true"
+
+        endpoint = self.url
+        if not endpoint.endswith("/"):
+            endpoint += "/"
+        endpoint += "tika/text"
+
+        r = requests.put(endpoint, data=data, headers=headers)
+
+        if r.ok:
+            raw_metadata = r.json()
+            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"]
+
+            log.debug("Tika extracted text: %s", text)
+
+            return [Document(page_content=text, metadata=headers)]
+        else:
+            raise Exception(f"Error calling Tika: {r.reason}")
+
+
+class DoclingLoader:
+    def __init__(
+        self, url, file_path=None, mime_type=None, ocr_engine=None, ocr_lang=None
+    ):
+        self.url = url.rstrip("/")
+        self.file_path = file_path
+        self.mime_type = mime_type
+        self.ocr_engine = ocr_engine
+        self.ocr_lang = ocr_lang
+
+    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",
+            }
+
+            if self.ocr_engine and self.ocr_lang:
+                params["ocr_engine"] = self.ocr_engine
+                params["ocr_lang"] = [
+                    lang.strip() for lang in self.ocr_lang.split(",") if lang.strip()
+                ]
+
+            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
+        self.kwargs = kwargs
+
+    def load(
+        self, filename: str, file_content_type: str, file_path: str
+    ) -> list[Document]:
+        loader = self._get_loader(filename, file_content_type, file_path)
+        docs = loader.load()
+
+        return [
+            Document(
+                page_content=ftfy.fix_text(doc.page_content), metadata=doc.metadata
+            )
+            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 self._is_text_file(file_ext, file_content_type):
+                loader = TextLoader(file_path, autodetect_encoding=True)
+            else:
+                loader = TikaLoader(
+                    url=self.kwargs.get("TIKA_SERVER_URL"),
+                    file_path=file_path,
+                    mime_type=file_content_type,
+                    extract_images=self.kwargs.get("PDF_EXTRACT_IMAGES"),
+                )
+        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,
+                    ocr_engine=self.kwargs.get("DOCLING_OCR_ENGINE"),
+                    ocr_lang=self.kwargs.get("DOCLING_OCR_LANG"),
+                )
+        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, autodetect_encoding=True)
+            elif file_ext == "rst":
+                loader = UnstructuredRSTLoader(file_path, mode="elements")
+            elif file_ext == "xml":
+                loader = UnstructuredXMLLoader(file_path)
+            elif file_ext in ["htm", "html"]:
+                loader = BSHTMLLoader(file_path, open_encoding="unicode_escape")
+            elif file_ext == "md":
+                loader = TextLoader(file_path, autodetect_encoding=True)
+            elif file_content_type == "application/epub+zip":
+                loader = UnstructuredEPubLoader(file_path)
+            elif (
+                file_content_type
+                == "application/vnd.openxmlformats-officedocument.wordprocessingml.document"
+                or file_ext == "docx"
+            ):
+                loader = Docx2txtLoader(file_path)
+            elif file_content_type in [
+                "application/vnd.ms-excel",
+                "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
+            ] or file_ext in ["xls", "xlsx"]:
+                loader = UnstructuredExcelLoader(file_path)
+            elif file_content_type in [
+                "application/vnd.ms-powerpoint",
+                "application/vnd.openxmlformats-officedocument.presentationml.presentation",
+            ] or file_ext in ["ppt", "pptx"]:
+                loader = UnstructuredPowerPointLoader(file_path)
+            elif file_ext == "msg":
+                loader = OutlookMessageLoader(file_path)
+            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)
+
+        return loader

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

@@ -62,12 +62,17 @@ class YoutubeLoader:
         _video_id = _parse_video_id(video_id)
         self.video_id = _video_id if _video_id is not None else video_id
         self._metadata = {"source": video_id}
-        self.language = language
         self.proxy_url = proxy_url
+
+        # Ensure language is a list
         if isinstance(language, str):
             self.language = [language]
         else:
-            self.language = language
+            self.language = list(language)
+
+        # Add English as fallback if not already in the list
+        if "en" not in self.language:
+            self.language.append("en")
 
     def load(self) -> List[Document]:
         """Load YouTube transcripts into `Document` objects."""
@@ -101,17 +106,31 @@ class YoutubeLoader:
             log.exception("Loading YouTube transcript failed")
             return []
 
-        try:
-            transcript = transcript_list.find_transcript(self.language)
-        except NoTranscriptFound:
-            transcript = transcript_list.find_transcript(["en"])
+        # Try each language in order of priority
+        for lang in self.language:
+            try:
+                transcript = transcript_list.find_transcript([lang])
+                log.debug(f"Found transcript for language '{lang}'")
+                transcript_pieces: List[Dict[str, Any]] = transcript.fetch()
+                transcript_text = " ".join(
+                    map(
+                        lambda transcript_piece: transcript_piece.text.strip(" "),
+                        transcript_pieces,
+                    )
+                )
+                return [Document(page_content=transcript_text, metadata=self._metadata)]
+            except NoTranscriptFound:
+                log.debug(f"No transcript found for language '{lang}'")
+                continue
+            except Exception as e:
+                log.info(f"Error finding transcript for language '{lang}'")
+                raise e
 
-        transcript_pieces: List[Dict[str, Any]] = transcript.fetch()
-
-        transcript = " ".join(
-            map(
-                lambda transcript_piece: transcript_piece["text"].strip(" "),
-                transcript_pieces,
-            )
+        # If we get here, all languages failed
+        languages_tried = ", ".join(self.language)
+        log.warning(
+            f"No transcript found for any of the specified languages: {languages_tried}. Verify if the video has transcripts, add more languages if needed."
+        )
+        raise NoTranscriptFound(
+            f"No transcript found for any supported language. Verify if the video has transcripts, add more languages if needed."
         )
-        return [Document(page_content=transcript, metadata=self._metadata)]

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

@@ -0,0 +1,837 @@
+import logging
+import os
+from typing import Optional, Union
+
+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.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"])
+
+
+from typing import Any
+
+from langchain_core.callbacks import CallbackManagerForRetrieverRun
+from langchain_core.retrievers import BaseRetriever
+
+
+class VectorSearchRetriever(BaseRetriever):
+    collection_name: Any
+    embedding_function: Any
+    top_k: int
+
+    def _get_relevant_documents(
+        self,
+        query: str,
+        *,
+        run_manager: CallbackManagerForRetrieverRun,
+    ) -> list[Document]:
+        result = VECTOR_DB_CLIENT.search(
+            collection_name=self.collection_name,
+            vectors=[self.embedding_function(query, RAG_EMBEDDING_QUERY_PREFIX)],
+            limit=self.top_k,
+        )
+
+        ids = result.ids[0]
+        metadatas = result.metadatas[0]
+        documents = result.documents[0]
+
+        results = []
+        for idx in range(len(ids)):
+            results.append(
+                Document(
+                    metadata=metadatas[idx],
+                    page_content=documents[idx],
+                )
+            )
+        return results
+
+
+def query_doc(
+    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],
+            limit=k,
+        )
+
+        if result:
+            log.info(f"query_doc:result {result.ids} {result.metadatas}")
+
+        return result
+    except Exception as 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:
+        log.debug(f"query_doc_with_hybrid_search:doc {collection_name}")
+        bm25_retriever = BM25Retriever.from_texts(
+            texts=collection_result.documents[0],
+            metadatas=collection_result.metadatas[0],
+        )
+        bm25_retriever.k = k
+
+        vector_search_retriever = VectorSearchRetriever(
+            collection_name=collection_name,
+            embedding_function=embedding_function,
+            top_k=k,
+        )
+
+        ensemble_retriever = EnsembleRetriever(
+            retrievers=[bm25_retriever, vector_search_retriever], weights=[0.5, 0.5]
+        )
+        compressor = RerankCompressor(
+            embedding_function=embedding_function,
+            top_n=k_reranker,
+            reranking_function=reranking_function,
+            r_score=r,
+        )
+
+        compression_retriever = ContextualCompressionRetriever(
+            base_compressor=compressor, base_retriever=ensemble_retriever
+        )
+
+        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": [distances],
+            "documents": [documents],
+            "metadatas": [metadatas],
+        }
+
+        log.info(
+            "query_doc_with_hybrid_search:result "
+            + f'{result["metadatas"]} {result["distances"]}'
+        )
+        return result
+    except Exception as e:
+        log.exception(f"Error querying doc {collection_name} with hybrid search: {e}")
+        raise e
+
+
+def merge_get_results(get_results: list[dict]) -> dict:
+    # Initialize lists to store combined data
+    combined_documents = []
+    combined_metadatas = []
+    combined_ids = []
+
+    for data in get_results:
+        combined_documents.extend(data["documents"][0])
+        combined_metadatas.extend(data["metadatas"][0])
+        combined_ids.extend(data["ids"][0])
+
+    # Create the output dictionary
+    result = {
+        "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.sha256(
+                    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],
+    embedding_function,
+    k: int,
+) -> dict:
+    results = []
+    error = False
+
+    def process_query_collection(collection_name, query_embedding):
+        try:
+            if collection_name:
+                result = query_doc(
+                    collection_name=collection_name,
+                    k=k,
+                    query_embedding=query_embedding,
+                )
+                if result is not None:
+                    return result.model_dump(), None
+            return None, None
+        except Exception as e:
+            log.exception(f"Error when querying the collection: {e}")
+            return None, e
+
+    # Generate all query embeddings (in one call)
+    query_embeddings = embedding_function(queries, prefix=RAG_EMBEDDING_QUERY_PREFIX)
+    log.debug(
+        f"query_collection: processing {len(queries)} queries across {len(collection_names)} collections"
+    )
+
+    with ThreadPoolExecutor() as executor:
+        future_results = []
+        for query_embedding in query_embeddings:
+            for collection_name in collection_names:
+                result = executor.submit(
+                    process_query_collection, collection_name, query_embedding
+                )
+                future_results.append(result)
+        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:
+        log.warning("All collection queries failed. No results returned.")
+
+    return merge_and_sort_query_results(results, k=k)
+
+
+def query_collection_with_hybrid_search(
+    collection_names: list[str],
+    queries: list[str],
+    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:
+            log.debug(
+                f"query_collection_with_hybrid_search:VECTOR_DB_CLIENT.get:collection {collection_name}"
+            )
+            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
+
+    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."
+        )
+
+    return merge_and_sort_query_results(results, k=k)
+
+
+def get_embedding_function(
+    embedding_engine,
+    embedding_model,
+    embedding_function,
+    url,
+    key,
+    embedding_batch_size,
+):
+    if embedding_engine == "":
+        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, 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, 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],
+                            prefix=prefix,
+                            user=user,
+                        )
+                    )
+                return embeddings
+            else:
+                return func(query, prefix, user)
+
+        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} {full_context}"
+    )
+
+    extracted_collections = []
+    relevant_contexts = []
+
+    for file in files:
+
+        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")}]],
+            }
+        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"):
+                    collection_names = file.get("collection_names", [])
+                else:
+                    collection_names.append(file["id"])
+            elif file.get("collection_name"):
+                collection_names.append(file["collection_name"])
+            elif file.get("id"):
+                if file.get("legacy"):
+                    collection_names.append(f"{file['id']}")
+                else:
+                    collection_names.append(f"file-{file['id']}")
+
+            collection_names = set(collection_names).difference(extracted_collections)
+            if not collection_names:
+                log.debug(f"skipping {file} as it has already been extracted")
+                continue
+
+            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,
+                            )
+                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 = []
+    for context in relevant_contexts:
+        try:
+            if "documents" in context:
+                if "metadatas" in context:
+                    source = {
+                        "source": context["file"],
+                        "document": context["documents"][0],
+                        "metadata": context["metadatas"][0],
+                    }
+                    if "distances" in context and context["distances"]:
+                        source["distances"] = context["distances"][0]
+
+                    sources.append(source)
+        except Exception as e:
+            log.exception(e)
+
+    return sources
+
+
+def get_model_path(model: str, update_model: bool = False):
+    # Construct huggingface_hub kwargs with local_files_only to return the snapshot path
+    cache_dir = os.getenv("SENTENCE_TRANSFORMERS_HOME")
+
+    local_files_only = not update_model
+
+    if OFFLINE_MODE:
+        local_files_only = True
+
+    snapshot_kwargs = {
+        "cache_dir": cache_dir,
+        "local_files_only": local_files_only,
+    }
+
+    log.debug(f"model: {model}")
+    log.debug(f"snapshot_kwargs: {snapshot_kwargs}")
+
+    # Inspiration from upstream sentence_transformers
+    if (
+        os.path.exists(model)
+        or ("\\" in model or model.count("/") > 1)
+        and local_files_only
+    ):
+        # If fully qualified path exists, return input, else set repo_id
+        return model
+    elif "/" not in model:
+        # Set valid repo_id for model short-name
+        model = "sentence-transformers" + "/" + model
+
+    snapshot_kwargs["repo_id"] = model
+
+    # Attempt to query the huggingface_hub library to determine the local path and/or to update
+    try:
+        model_repo_path = snapshot_download(**snapshot_kwargs)
+        log.debug(f"model_repo_path: {model_repo_path}")
+        return model_repo_path
+    except Exception as e:
+        log.exception(f"Cannot determine model snapshot path: {e}")
+        return model
+
+
+def generate_openai_batch_embeddings(
+    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=json_data,
+        )
+        r.raise_for_status()
+        data = r.json()
+        if "data" in data:
+            return [elem["embedding"] for elem in data["data"]]
+        else:
+            raise "Something went wrong :/"
+    except Exception as 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 = "",
+    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=json_data,
+        )
+        r.raise_for_status()
+        data = r.json()
+
+        if "embeddings" in data:
+            return data["embeddings"]
+        else:
+            raise "Something went wrong :/"
+    except Exception as e:
+        log.exception(f"Error generating ollama batch embeddings: {e}")
+        return None
+
+
+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,
+                    "prefix": prefix,
+                    "user": user,
+                }
+            )
+        else:
+            embeddings = generate_ollama_batch_embeddings(
+                **{
+                    "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, prefix, user
+            )
+        else:
+            embeddings = generate_openai_batch_embeddings(
+                model, [text], url, key, prefix, user
+            )
+        return embeddings[0] if isinstance(text, str) else embeddings
+
+
+import operator
+from typing import Optional, Sequence
+
+from langchain_core.callbacks import Callbacks
+from langchain_core.documents import BaseDocumentCompressor, Document
+
+
+class RerankCompressor(BaseDocumentCompressor):
+    embedding_function: Any
+    top_n: int
+    reranking_function: Any
+    r_score: float
+
+    class Config:
+        extra = "forbid"
+        arbitrary_types_allowed = True
+
+    def compress_documents(
+        self,
+        documents: Sequence[Document],
+        query: str,
+        callbacks: Optional[Callbacks] = None,
+    ) -> Sequence[Document]:
+        reranking = self.reranking_function is not None
+
+        if reranking:
+            scores = self.reranking_function.predict(
+                [(query, doc.page_content) for doc in documents]
+            )
+        else:
+            from sentence_transformers import util
+
+            query_embedding = self.embedding_function(query, RAG_EMBEDDING_QUERY_PREFIX)
+            document_embedding = self.embedding_function(
+                [doc.page_content for doc in documents], RAG_EMBEDDING_CONTENT_PREFIX
+            )
+            scores = util.cos_sim(query_embedding, document_embedding)[0]
+
+        docs_with_scores = list(zip(documents, scores.tolist()))
+        if self.r_score:
+            docs_with_scores = [
+                (d, s) for d, s in docs_with_scores if s >= self.r_score
+            ]
+
+        result = sorted(docs_with_scores, key=operator.itemgetter(1), reverse=True)
+        final_results = []
+        for doc, doc_score in result[: self.top_n]:
+            metadata = doc.metadata
+            metadata["score"] = doc_score
+            doc = Document(
+                page_content=doc.page_content,
+                metadata=metadata,
+            )
+            final_results.append(doc)
+        return final_results

backend/open_webui/retrieval/vector/connector.py

@@ -0,0 +1,30 @@
+from open_webui.config import VECTOR_DB
+
+if VECTOR_DB == "milvus":
+    from open_webui.retrieval.vector.dbs.milvus import MilvusClient
+
+    VECTOR_DB_CLIENT = MilvusClient()
+elif VECTOR_DB == "qdrant":
+    from open_webui.retrieval.vector.dbs.qdrant import QdrantClient
+
+    VECTOR_DB_CLIENT = QdrantClient()
+elif VECTOR_DB == "opensearch":
+    from open_webui.retrieval.vector.dbs.opensearch import OpenSearchClient
+
+    VECTOR_DB_CLIENT = OpenSearchClient()
+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()
+elif VECTOR_DB == "pinecone":
+    from open_webui.retrieval.vector.dbs.pinecone import PineconeClient
+
+    VECTOR_DB_CLIENT = PineconeClient()
+else:
+    from open_webui.retrieval.vector.dbs.chroma import ChromaClient
+
+    VECTOR_DB_CLIENT = ChromaClient()

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

@@ -1,10 +1,16 @@
 import chromadb
+import logging
 from chromadb import Settings
 from chromadb.utils.batch_utils import create_batches
 
 from typing import Optional
 
-from open_webui.apps.retrieval.vector.main import VectorItem, SearchResult, GetResult
+from open_webui.retrieval.vector.main import (
+    VectorDBBase,
+    VectorItem,
+    SearchResult,
+    GetResult,
+)
 from open_webui.config import (
     CHROMA_DATA_PATH,
     CHROMA_HTTP_HOST,
@@ -16,9 +22,13 @@ 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:
+class ChromaClient(VectorDBBase):
     def __init__(self):
         settings_dict = {
             "allow_reset": True,
@@ -51,8 +61,8 @@ class ChromaClient:
 
     def has_collection(self, collection_name: str) -> bool:
         # Check if the collection exists based on the collection name.
-        collections = self.client.list_collections()
-        return collection_name in [collection.name for collection in collections]
+        collection_names = self.client.list_collections()
+        return collection_name in collection_names
 
     def delete_collection(self, collection_name: str):
         # Delete the collection based on the collection name.
@@ -70,10 +80,16 @@ 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 +118,7 @@ class ChromaClient:
                     }
                 )
             return None
-        except Exception as e:
-            print(e)
+        except:
             return None
 
     def get(self, collection_name: str) -> Optional[GetResult]:
@@ -162,12 +177,19 @@ 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 +1,300 @@
+from elasticsearch import Elasticsearch, BadRequestError
+from typing import Optional
+import ssl
+from elasticsearch.helpers import bulk, scan
+from open_webui.retrieval.vector.main import (
+    VectorDBBase,
+    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(VectorDBBase):
+    """
+    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

@@ -1,19 +1,33 @@
 from pymilvus import MilvusClient as Client
 from pymilvus import FieldSchema, DataType
 import json
-
+import logging
 from typing import Optional
 
-from open_webui.apps.retrieval.vector.main import VectorItem, SearchResult, GetResult
+from open_webui.retrieval.vector.main import (
+    VectorDBBase,
+    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:
+class MilvusClient(VectorDBBase):
     def __init__(self):
         self.collection_prefix = "open_webui"
-        self.client = Client(uri=MILVUS_URI)
+        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 = []
@@ -55,7 +69,10 @@ 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"))
 
@@ -163,7 +180,7 @@ 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
@@ -190,10 +207,12 @@ 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

@@ -0,0 +1,263 @@
+from opensearchpy import OpenSearch
+from opensearchpy.helpers import bulk
+from typing import Optional
+
+from open_webui.retrieval.vector.main import (
+    VectorDBBase,
+    VectorItem,
+    SearchResult,
+    GetResult,
+)
+from open_webui.config import (
+    OPENSEARCH_URI,
+    OPENSEARCH_SSL,
+    OPENSEARCH_CERT_VERIFY,
+    OPENSEARCH_USERNAME,
+    OPENSEARCH_PASSWORD,
+)
+
+
+class OpenSearchClient(VectorDBBase):
+    def __init__(self):
+        self.index_prefix = "open_webui"
+        self.client = OpenSearch(
+            hosts=[OPENSEARCH_URI],
+            use_ssl=OPENSEARCH_SSL,
+            verify_certs=OPENSEARCH_CERT_VERIFY,
+            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 = []
+
+        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])
+
+    def _result_to_search_result(self, result) -> SearchResult:
+        if not result["hits"]["hits"]:
+            return None
+
+        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],
+        )
+
+    def _create_index(self, collection_name: str, dimension: int):
+        body = {
+            "settings": {"index": {"knn": True}},
+            "mappings": {
+                "properties": {
+                    "id": {"type": "keyword"},
+                    "vector": {
+                        "type": "knn_vector",
+                        "dimension": dimension,  # Adjust based on your vector dimensions
+                        "index": True,
+                        "similarity": "faiss",
+                        "method": {
+                            "name": "hnsw",
+                            "space_type": "innerproduct",  # Use inner product to approximate cosine similarity
+                            "engine": "faiss",
+                            "parameters": {
+                                "ef_construction": 128,
+                                "m": 16,
+                            },
+                        },
+                    },
+                    "text": {"type": "text"},
+                    "metadata": {"type": "object"},
+                }
+            },
+        }
+        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, 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=self._get_index_name(collection_name))
+
+    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=self._get_index_name(collection_name))
+
+    def search(
+        self, collection_name: str, vectors: list[list[float | int]], limit: int
+    ) -> Optional[SearchResult]:
+        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"],
+        }
+
+        for field, value in filter.items():
+            query_body["query"]["bool"]["filter"].append(
+                {"match": {"metadata." + str(field): value}}
+            )
+
+        size = limit if limit else 10
+
+        try:
+            result = self.client.search(
+                index=self._get_index_name(collection_name),
+                body=query_body,
+                size=size,
+            )
+
+            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=self._get_index_name(collection_name), body=query
+        )
+        return self._result_to_get_result(result)
+
+    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 = [
+                {
+                    "_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
+            ]
+            bulk(self.client, actions)
+
+    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 = [
+                {
+                    "_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
+            ]
+            bulk(self.client, 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}_*")
+        for index in indices:
+            self.client.indices.delete(index=index)

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

@@ -1,13 +1,16 @@
 from typing import Optional, List, Dict, Any
+import logging
 from sqlalchemy import (
     cast,
     column,
     create_engine,
     Column,
     Integer,
+    MetaData,
     select,
     text,
     Text,
+    Table,
     values,
 )
 from sqlalchemy.sql import true
@@ -17,13 +20,24 @@ from sqlalchemy.orm import declarative_base, scoped_session, sessionmaker
 from sqlalchemy.dialects.postgresql import JSONB, array
 from pgvector.sqlalchemy import Vector
 from sqlalchemy.ext.mutable import MutableDict
+from sqlalchemy.exc import NoSuchTableError
 
-from open_webui.apps.retrieval.vector.main import VectorItem, SearchResult, GetResult
-from open_webui.config import PGVECTOR_DB_URL
+from open_webui.retrieval.vector.main import (
+    VectorDBBase,
+    VectorItem,
+    SearchResult,
+    GetResult,
+)
+from open_webui.config import PGVECTOR_DB_URL, PGVECTOR_INITIALIZE_MAX_VECTOR_LENGTH
 
-VECTOR_LENGTH = 1536
+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"
@@ -35,12 +49,12 @@ class DocumentChunk(Base):
     vmetadata = Column(MutableDict.as_mutable(JSONB), nullable=True)
 
 
-class PgvectorClient:
+class PgvectorClient(VectorDBBase):
     def __init__(self) -> None:
 
         # if no pgvector uri, use the existing database connection
         if not PGVECTOR_DB_URL:
-            from open_webui.apps.webui.internal.db import Session
+            from open_webui.internal.db import Session
 
             self.session = Session
         else:
@@ -56,6 +70,9 @@ class PgvectorClient:
             # Ensure the pgvector extension is available
             self.session.execute(text("CREATE EXTENSION IF NOT EXISTS vector;"))
 
+            # Check vector length consistency
+            self.check_vector_length()
+
             # Create the tables if they do not exist
             # Base.metadata.create_all requires a bind (engine or connection)
             # Get the connection from the session
@@ -76,12 +93,47 @@ 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:
+        """
+        Check if the VECTOR_LENGTH matches the existing vector column dimension in the database.
+        Raises an exception if there is a mismatch.
+        """
+        metadata = MetaData()
+        try:
+            # Attempt to reflect the 'document_chunk' table
+            document_chunk_table = Table(
+                "document_chunk", metadata, autoload_with=self.session.bind
+            )
+        except NoSuchTableError:
+            # Table does not exist; no action needed
+            return
+
+        # Proceed to check the vector column
+        if "vector" in document_chunk_table.columns:
+            vector_column = document_chunk_table.columns["vector"]
+            vector_type = vector_column.type
+            if isinstance(vector_type, Vector):
+                db_vector_length = vector_type.dim
+                if db_vector_length != VECTOR_LENGTH:
+                    raise Exception(
+                        f"VECTOR_LENGTH {VECTOR_LENGTH} does not match existing vector column dimension {db_vector_length}. "
+                        "Cannot change vector size after initialization without migrating the data."
+                    )
+            else:
+                raise Exception(
+                    "The 'vector' column exists but is not of type 'Vector'."
+                )
+        else:
+            raise Exception(
+                "The 'vector' column does not exist in the 'document_chunk' table."
+            )
+
     def adjust_vector_length(self, vector: List[float]) -> List[float]:
         # Adjust vector to have length VECTOR_LENGTH
         current_length = len(vector)
@@ -89,9 +141,8 @@ class PgvectorClient:
             # Pad the vector with zeros
             vector += [0.0] * (VECTOR_LENGTH - current_length)
         elif current_length > VECTOR_LENGTH:
-            raise Exception(
-                f"Vector length {current_length} not supported. Max length must be <= {VECTOR_LENGTH}"
-            )
+            # Truncate the vector to VECTOR_LENGTH
+            vector = vector[:VECTOR_LENGTH]
         return vector
 
     def insert(self, collection_name: str, items: List[VectorItem]) -> None:
@@ -109,12 +160,12 @@ 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:
@@ -143,10 +194,12 @@ 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(
@@ -229,7 +282,9 @@ 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)
 
@@ -237,7 +292,7 @@ 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(
@@ -269,7 +324,7 @@ 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(
@@ -293,7 +348,7 @@ 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(
@@ -315,22 +370,22 @@ 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:
@@ -346,9 +401,9 @@ 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/pinecone.py

@@ -0,0 +1,412 @@
+from typing import Optional, List, Dict, Any, Union
+import logging
+from pinecone import Pinecone, ServerlessSpec
+
+from open_webui.retrieval.vector.main import (
+    VectorDBBase,
+    VectorItem,
+    SearchResult,
+    GetResult,
+)
+from open_webui.config import (
+    PINECONE_API_KEY,
+    PINECONE_ENVIRONMENT,
+    PINECONE_INDEX_NAME,
+    PINECONE_DIMENSION,
+    PINECONE_METRIC,
+    PINECONE_CLOUD,
+)
+from open_webui.env import SRC_LOG_LEVELS
+
+NO_LIMIT = 10000  # Reasonable limit to avoid overwhelming the system
+BATCH_SIZE = 100  # Recommended batch size for Pinecone operations
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
+
+class PineconeClient(VectorDBBase):
+    def __init__(self):
+        self.collection_prefix = "open-webui"
+
+        # Validate required configuration
+        self._validate_config()
+
+        # Store configuration values
+        self.api_key = PINECONE_API_KEY
+        self.environment = PINECONE_ENVIRONMENT
+        self.index_name = PINECONE_INDEX_NAME
+        self.dimension = PINECONE_DIMENSION
+        self.metric = PINECONE_METRIC
+        self.cloud = PINECONE_CLOUD
+
+        # Initialize Pinecone client
+        self.client = Pinecone(api_key=self.api_key)
+
+        # Create index if it doesn't exist
+        self._initialize_index()
+
+    def _validate_config(self) -> None:
+        """Validate that all required configuration variables are set."""
+        missing_vars = []
+        if not PINECONE_API_KEY:
+            missing_vars.append("PINECONE_API_KEY")
+        if not PINECONE_ENVIRONMENT:
+            missing_vars.append("PINECONE_ENVIRONMENT")
+        if not PINECONE_INDEX_NAME:
+            missing_vars.append("PINECONE_INDEX_NAME")
+        if not PINECONE_DIMENSION:
+            missing_vars.append("PINECONE_DIMENSION")
+        if not PINECONE_CLOUD:
+            missing_vars.append("PINECONE_CLOUD")
+
+        if missing_vars:
+            raise ValueError(
+                f"Required configuration missing: {', '.join(missing_vars)}"
+            )
+
+    def _initialize_index(self) -> None:
+        """Initialize the Pinecone index."""
+        try:
+            # Check if index exists
+            if self.index_name not in self.client.list_indexes().names():
+                log.info(f"Creating Pinecone index '{self.index_name}'...")
+                self.client.create_index(
+                    name=self.index_name,
+                    dimension=self.dimension,
+                    metric=self.metric,
+                    spec=ServerlessSpec(cloud=self.cloud, region=self.environment),
+                )
+                log.info(f"Successfully created Pinecone index '{self.index_name}'")
+            else:
+                log.info(f"Using existing Pinecone index '{self.index_name}'")
+
+            # Connect to the index
+            self.index = self.client.Index(self.index_name)
+
+        except Exception as e:
+            log.error(f"Failed to initialize Pinecone index: {e}")
+            raise RuntimeError(f"Failed to initialize Pinecone index: {e}")
+
+    def _create_points(
+        self, items: List[VectorItem], collection_name_with_prefix: str
+    ) -> List[Dict[str, Any]]:
+        """Convert VectorItem objects to Pinecone point format."""
+        points = []
+        for item in items:
+            # Start with any existing metadata or an empty dict
+            metadata = item.get("metadata", {}).copy() if item.get("metadata") else {}
+
+            # Add text to metadata if available
+            if "text" in item:
+                metadata["text"] = item["text"]
+
+            # Always add collection_name to metadata for filtering
+            metadata["collection_name"] = collection_name_with_prefix
+
+            point = {
+                "id": item["id"],
+                "values": item["vector"],
+                "metadata": metadata,
+            }
+            points.append(point)
+        return points
+
+    def _get_collection_name_with_prefix(self, collection_name: str) -> str:
+        """Get the collection name with prefix."""
+        return f"{self.collection_prefix}_{collection_name}"
+
+    def _normalize_distance(self, score: float) -> float:
+        """Normalize distance score based on the metric used."""
+        if self.metric.lower() == "cosine":
+            # Cosine similarity ranges from -1 to 1, normalize to 0 to 1
+            return (score + 1.0) / 2.0
+        elif self.metric.lower() in ["euclidean", "dotproduct"]:
+            # These are already suitable for ranking (smaller is better for Euclidean)
+            return score
+        else:
+            # For other metrics, use as is
+            return score
+
+    def _result_to_get_result(self, matches: list) -> GetResult:
+        """Convert Pinecone matches to GetResult format."""
+        ids = []
+        documents = []
+        metadatas = []
+
+        for match in matches:
+            metadata = match.get("metadata", {})
+            ids.append(match["id"])
+            documents.append(metadata.get("text", ""))
+            metadatas.append(metadata)
+
+        return GetResult(
+            **{
+                "ids": [ids],
+                "documents": [documents],
+                "metadatas": [metadatas],
+            }
+        )
+
+    def has_collection(self, collection_name: str) -> bool:
+        """Check if a collection exists by searching for at least one item."""
+        collection_name_with_prefix = self._get_collection_name_with_prefix(
+            collection_name
+        )
+
+        try:
+            # Search for at least 1 item with this collection name in metadata
+            response = self.index.query(
+                vector=[0.0] * self.dimension,  # dummy vector
+                top_k=1,
+                filter={"collection_name": collection_name_with_prefix},
+                include_metadata=False,
+            )
+            return len(response.matches) > 0
+        except Exception as e:
+            log.exception(
+                f"Error checking collection '{collection_name_with_prefix}': {e}"
+            )
+            return False
+
+    def delete_collection(self, collection_name: str) -> None:
+        """Delete a collection by removing all vectors with the collection name in metadata."""
+        collection_name_with_prefix = self._get_collection_name_with_prefix(
+            collection_name
+        )
+        try:
+            self.index.delete(filter={"collection_name": collection_name_with_prefix})
+            log.info(
+                f"Collection '{collection_name_with_prefix}' deleted (all vectors removed)."
+            )
+        except Exception as e:
+            log.warning(
+                f"Failed to delete collection '{collection_name_with_prefix}': {e}"
+            )
+            raise
+
+    def insert(self, collection_name: str, items: List[VectorItem]) -> None:
+        """Insert vectors into a collection."""
+        if not items:
+            log.warning("No items to insert")
+            return
+
+        collection_name_with_prefix = self._get_collection_name_with_prefix(
+            collection_name
+        )
+        points = self._create_points(items, collection_name_with_prefix)
+
+        # Insert in batches for better performance and reliability
+        for i in range(0, len(points), BATCH_SIZE):
+            batch = points[i : i + BATCH_SIZE]
+            try:
+                self.index.upsert(vectors=batch)
+                log.debug(
+                    f"Inserted batch of {len(batch)} vectors into '{collection_name_with_prefix}'"
+                )
+            except Exception as e:
+                log.error(
+                    f"Error inserting batch into '{collection_name_with_prefix}': {e}"
+                )
+                raise
+
+        log.info(
+            f"Successfully inserted {len(items)} vectors into '{collection_name_with_prefix}'"
+        )
+
+    def upsert(self, collection_name: str, items: List[VectorItem]) -> None:
+        """Upsert (insert or update) vectors into a collection."""
+        if not items:
+            log.warning("No items to upsert")
+            return
+
+        collection_name_with_prefix = self._get_collection_name_with_prefix(
+            collection_name
+        )
+        points = self._create_points(items, collection_name_with_prefix)
+
+        # Upsert in batches
+        for i in range(0, len(points), BATCH_SIZE):
+            batch = points[i : i + BATCH_SIZE]
+            try:
+                self.index.upsert(vectors=batch)
+                log.debug(
+                    f"Upserted batch of {len(batch)} vectors into '{collection_name_with_prefix}'"
+                )
+            except Exception as e:
+                log.error(
+                    f"Error upserting batch into '{collection_name_with_prefix}': {e}"
+                )
+                raise
+
+        log.info(
+            f"Successfully upserted {len(items)} vectors into '{collection_name_with_prefix}'"
+        )
+
+    def search(
+        self, collection_name: str, vectors: List[List[Union[float, int]]], limit: int
+    ) -> Optional[SearchResult]:
+        """Search for similar vectors in a collection."""
+        if not vectors or not vectors[0]:
+            log.warning("No vectors provided for search")
+            return None
+
+        collection_name_with_prefix = self._get_collection_name_with_prefix(
+            collection_name
+        )
+
+        if limit is None or limit <= 0:
+            limit = NO_LIMIT
+
+        try:
+            # Search using the first vector (assuming this is the intended behavior)
+            query_vector = vectors[0]
+
+            # Perform the search
+            query_response = self.index.query(
+                vector=query_vector,
+                top_k=limit,
+                include_metadata=True,
+                filter={"collection_name": collection_name_with_prefix},
+            )
+
+            if not query_response.matches:
+                # Return empty result if no matches
+                return SearchResult(
+                    ids=[[]],
+                    documents=[[]],
+                    metadatas=[[]],
+                    distances=[[]],
+                )
+
+            # Convert to GetResult format
+            get_result = self._result_to_get_result(query_response.matches)
+
+            # Calculate normalized distances based on metric
+            distances = [
+                [
+                    self._normalize_distance(match.score)
+                    for match in query_response.matches
+                ]
+            ]
+
+            return SearchResult(
+                ids=get_result.ids,
+                documents=get_result.documents,
+                metadatas=get_result.metadatas,
+                distances=distances,
+            )
+        except Exception as e:
+            log.error(f"Error searching in '{collection_name_with_prefix}': {e}")
+            return None
+
+    def query(
+        self, collection_name: str, filter: Dict, limit: Optional[int] = None
+    ) -> Optional[GetResult]:
+        """Query vectors by metadata filter."""
+        collection_name_with_prefix = self._get_collection_name_with_prefix(
+            collection_name
+        )
+
+        if limit is None or limit <= 0:
+            limit = NO_LIMIT
+
+        try:
+            # Create a zero vector for the dimension as Pinecone requires a vector
+            zero_vector = [0.0] * self.dimension
+
+            # Combine user filter with collection_name
+            pinecone_filter = {"collection_name": collection_name_with_prefix}
+            if filter:
+                pinecone_filter.update(filter)
+
+            # Perform metadata-only query
+            query_response = self.index.query(
+                vector=zero_vector,
+                filter=pinecone_filter,
+                top_k=limit,
+                include_metadata=True,
+            )
+
+            return self._result_to_get_result(query_response.matches)
+
+        except Exception as e:
+            log.error(f"Error querying collection '{collection_name}': {e}")
+            return None
+
+    def get(self, collection_name: str) -> Optional[GetResult]:
+        """Get all vectors in a collection."""
+        collection_name_with_prefix = self._get_collection_name_with_prefix(
+            collection_name
+        )
+
+        try:
+            # Use a zero vector for fetching all entries
+            zero_vector = [0.0] * self.dimension
+
+            # Add filter to only get vectors for this collection
+            query_response = self.index.query(
+                vector=zero_vector,
+                top_k=NO_LIMIT,
+                include_metadata=True,
+                filter={"collection_name": collection_name_with_prefix},
+            )
+
+            return self._result_to_get_result(query_response.matches)
+
+        except Exception as e:
+            log.error(f"Error getting collection '{collection_name}': {e}")
+            return None
+
+    def delete(
+        self,
+        collection_name: str,
+        ids: Optional[List[str]] = None,
+        filter: Optional[Dict] = None,
+    ) -> None:
+        """Delete vectors by IDs or filter."""
+        collection_name_with_prefix = self._get_collection_name_with_prefix(
+            collection_name
+        )
+
+        try:
+            if ids:
+                # Delete by IDs (in batches for large deletions)
+                for i in range(0, len(ids), BATCH_SIZE):
+                    batch_ids = ids[i : i + BATCH_SIZE]
+                    # Note: When deleting by ID, we can't filter by collection_name
+                    # This is a limitation of Pinecone - be careful with ID uniqueness
+                    self.index.delete(ids=batch_ids)
+                    log.debug(
+                        f"Deleted batch of {len(batch_ids)} vectors by ID from '{collection_name_with_prefix}'"
+                    )
+                log.info(
+                    f"Successfully deleted {len(ids)} vectors by ID from '{collection_name_with_prefix}'"
+                )
+
+            elif filter:
+                # Combine user filter with collection_name
+                pinecone_filter = {"collection_name": collection_name_with_prefix}
+                if filter:
+                    pinecone_filter.update(filter)
+                # Delete by metadata filter
+                self.index.delete(filter=pinecone_filter)
+                log.info(
+                    f"Successfully deleted vectors by filter from '{collection_name_with_prefix}'"
+                )
+
+            else:
+                log.warning("No ids or filter provided for delete operation")
+
+        except Exception as e:
+            log.error(f"Error deleting from collection '{collection_name}': {e}")
+            raise
+
+    def reset(self) -> None:
+        """Reset the database by deleting all collections."""
+        try:
+            self.index.delete(delete_all=True)
+            log.info("All vectors successfully deleted from the index.")
+        except Exception as e:
+            log.error(f"Failed to reset Pinecone index: {e}")
+            raise