Merge pull request #5565 from open-webui/dev

0.3.23

.github/workflows/integration-test.yml

@@ -157,7 +157,7 @@ jobs:
           GLOBAL_LOG_LEVEL: debug
         run: |
           cd backend
-          uvicorn main:app --port "8080" --forwarded-allow-ips '*' &
+          uvicorn open_webui.main:app --port "8080" --forwarded-allow-ips '*' &
           UVICORN_PID=$!
           # Wait up to 40 seconds for the server to start
           for i in {1..40}; do
@@ -184,7 +184,7 @@ jobs:
           DATABASE_URL: postgresql://postgres:postgres@localhost:5432/postgres
         run: |
           cd backend
-          uvicorn main:app --port "8081" --forwarded-allow-ips '*' &
+          uvicorn open_webui.main:app --port "8081" --forwarded-allow-ips '*' &
           UVICORN_PID=$!
           # Wait up to 20 seconds for the server to start
           for i in {1..20}; do
@@ -230,7 +230,7 @@ jobs:
 #          DATABASE_URL: mysql://root:mysql@localhost:3306/mysql
 #        run: |
 #          cd backend
-#          uvicorn main:app --port "8083" --forwarded-allow-ips '*' &
+#          uvicorn open_webui.main:app --port "8083" --forwarded-allow-ips '*' &
 #          UVICORN_PID=$!
 #          # Wait up to 20 seconds for the server to start
 #          for i in {1..20}; do

CHANGELOG.md

@@ -5,6 +5,117 @@ 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).
 
+## [0.3.23] - 2024-09-21
+
+### Added
+
+- **🚀 WebSocket Redis Support**: Enhanced load balancing capabilities for multiple instance setups, promoting better performance and reliability in WebUI.
+- **🔧 Adjustable Chat Controls**: Introduced width-adjustable chat controls, enabling a personalized and more comfortable user interface.
+- **🌎 i18n Updates**: Improved and updated the Chinese translations.
+
+### Fixed
+
+- **🌐 Task Model Unloading Issue**: Modified task handling to use the Ollama /api/chat endpoint instead of OpenAI compatible endpoint, ensuring models stay loaded and ready with custom parameters, thus minimizing delays in task execution.
+- **📝 Title Generation Fix for OpenAI Compatible APIs**: Resolved an issue preventing the generation of titles, enhancing consistency and reliability when using multiple API providers.
+- **🗃️ RAG Duplicate Collection Issue**: Fixed a bug causing repeated processing of the same uploaded file. Now utilizes indexed files to prevent unnecessary duplications, optimizing resource usage.
+- **🖼️ Image Generation Enhancement**: Refactored OpenAI image generation endpoint to be asynchronous, preventing the WebUI from becoming unresponsive during processing, thus enhancing user experience.
+- **🔓 Downgrade Authlib**: Reverted Authlib to version 1.3.1 to address and resolve issues concerning OAuth functionality.
+
+### Changed
+
+- **🔍 Improved Message Interaction**: Enhanced the message node interface to allow for easier focus redirection with a simple click, streamlining user interaction.
+- **✨ Styling Refactor**: Updated WebUI styling for a cleaner, more modern look, enhancing user experience across the platform.
+
+## [0.3.22] - 2024-09-19
+
+### Added
+
+- **⭐ Chat Overview**: Introducing a node-based interactive messages diagram for improved visualization of conversation flows.
+- **🔗 Multiple Vector DB Support**: Now supports multiple vector databases, including the newly added Milvus support. Community contributions for additional database support are highly encouraged!
+- **📡 Experimental Non-Stream Chat Completion**: Experimental feature allowing the use of OpenAI o1 models, which do not support streaming, ensuring more versatile model deployment.
+- **🔍 Experimental Colbert-AI Reranker Integration**: Added support for "jinaai/jina-colbert-v2" as a reranker, enhancing search relevance and accuracy. Note: it may not function at all on low-spec computers.
+- **🕸️ ENABLE_WEBSOCKET_SUPPORT**: Added environment variable for instances to ignore websocket upgrades, stabilizing connections on platforms with websocket issues.
+- **🔊 Azure Speech Service Integration**: Added support for Azure Speech services for Text-to-Speech (TTS).
+- **🎚️ Customizable Playback Speed**: Playback speed control is now available in Call mode settings, allowing users to adjust audio playback speed to their preferences.
+- **🧠 Enhanced Error Messaging**: System now displays helpful error messages directly to users during chat completion issues.
+- **📂 Save Model as Transparent PNG**: Model profile images are now saved as PNGs, supporting transparency and improving visual integration.
+- **📱 iPhone Compatibility Adjustments**: Added padding to accommodate the iPhone navigation bar, improving UI display on these devices.
+- **🔗 Secure Response Headers**: Implemented security response headers, bolstering web application security.
+- **🔧 Enhanced AUTOMATIC1111 Settings**: Users can now configure 'CFG Scale', 'Sampler', and 'Scheduler' parameters directly in the admin settings, enhancing workflow flexibility without source code modifications.
+- **🌍 i18n Updates**: Enhanced translations for Chinese, Ukrainian, Russian, and French, fostering a better localized experience.
+
+### Fixed
+
+- **🛠️ Chat Message Deletion**: Resolved issues with chat message deletion, ensuring a smoother user interaction and system stability.
+- **🔢 Ordered List Numbering**: Fixed the incorrect ordering in lists.
+
+### Changed
+
+- **🎨 Transparent Icon Handling**: Allowed model icons to be displayed on transparent backgrounds, improving UI aesthetics.
+- **📝 Improved RAG Template**: Enhanced Retrieval-Augmented Generation template, optimizing context handling and error checking for more precise operation.
+
+## [0.3.21] - 2024-09-08
+
+### Added
+
+- **📊 Document Count Display**: Now displays the total number of documents directly within the dashboard.
+- **🚀 Ollama Embed API Endpoint**: Enabled /api/embed endpoint proxy support.
+
+### Fixed
+
+- **🐳 Docker Launch Issue**: Resolved the problem preventing Open-WebUI from launching correctly when using Docker.
+
+### Changed
+
+- **🔍 Enhanced Search Prompts**: Improved the search query generation prompts for better accuracy and user interaction, enhancing the overall search experience.
+
+## [0.3.20] - 2024-09-07
+
+### Added
+
+- **🌐 Translation Update**: Updated Catalan translations to improve user experience for Catalan speakers.
+
+### Fixed
+
+- **📄 PDF Download**: Resolved a configuration issue with fonts directory, ensuring PDFs are now downloaded with the correct formatting.
+- **🛠️ Installation of Tools & Functions Requirements**: Fixed a bug where necessary requirements for tools and functions were not properly installing.
+- **🔗 Inline Image Link Rendering**: Enabled rendering of images directly from links in chat.
+- **📞 Post-Call User Interface Cleanup**: Adjusted UI behavior to automatically close chat controls after a voice call ends, reducing screen clutter.
+- **🎙️ Microphone Deactivation Post-Call**: Addressed an issue where the microphone remained active after calls.
+- **✍️ Markdown Spacing Correction**: Corrected spacing in Markdown rendering, ensuring text appears neatly and as expected.
+- **🔄 Message Re-rendering**: Fixed an issue causing all response messages to re-render with each new message, now improving chat performance.
+
+### Changed
+
+- **🌐 Refined Web Search Integration**: Deprecated the Search Query Generation Prompt threshold; introduced a toggle button for "Enable Web Search Query Generation" allowing users to opt-in to using web search more judiciously.
+- **📝 Default Prompt Templates Update**: Emptied environment variable templates for search and title generation now default to the Open WebUI default prompt templates, simplifying configuration efforts.
+
+## [0.3.19] - 2024-09-05
+
+### Added
+
+- **🌐 Translation Update**: Improved Chinese translations.
+
+### Fixed
+
+- **📂 DATA_DIR Overriding**: Fixed an issue to avoid overriding DATA_DIR, preventing errors when directories are set identically, ensuring smoother operation and data management.
+- **🛠️ Frontmatter Extraction**: Fixed the extraction process for frontmatter in tools and functions.
+
+### Changed
+
+- **🎨 UI Styling**: Refined the user interface styling for enhanced visual coherence and user experience.
+
+## [0.3.18] - 2024-09-04
+
+### Added
+
+- **🛠️ Direct Database Execution for Tools & Functions**: Enhanced the execution of Python files for tools and functions, now directly loading from the database for a more streamlined backend process.
+
+### Fixed
+
+- **🔄 Automatic Rewrite of Import Statements in Tools & Functions**: Tool and function scripts that import 'utils', 'apps', 'main', 'config' will now automatically rename these with 'open_webui.', ensuring compatibility and consistency across different modules.
+- **🎨 Styling Adjustments**: Minor fixes in the visual styling to improve user experience and interface consistency.
+
 ## [0.3.17] - 2024-09-04
 
 ### Added

Dockerfile

@@ -74,6 +74,10 @@ ENV RAG_EMBEDDING_MODEL="$USE_EMBEDDING_MODEL_DOCKER" \
 
 ## Hugging Face download cache ##
 ENV HF_HOME="/app/backend/data/cache/embedding/models"
+
+## Torch Extensions ##
+# ENV TORCH_EXTENSIONS_DIR="/.cache/torch_extensions"
+
 #### Other models ##########################################################
 
 WORKDIR /app/backend
@@ -96,7 +100,7 @@ RUN chown -R $UID:$GID /app $HOME
 RUN if [ "$USE_OLLAMA" = "true" ]; then \
     apt-get update && \
     # Install pandoc and netcat
-    apt-get install -y --no-install-recommends pandoc netcat-openbsd curl && \
+    apt-get install -y --no-install-recommends git build-essential pandoc netcat-openbsd curl && \
     apt-get install -y --no-install-recommends gcc python3-dev && \
     # for RAG OCR
     apt-get install -y --no-install-recommends ffmpeg libsm6 libxext6 && \
@@ -109,7 +113,7 @@ RUN if [ "$USE_OLLAMA" = "true" ]; then \
     else \
     apt-get update && \
     # Install pandoc, netcat and gcc
-    apt-get install -y --no-install-recommends pandoc gcc netcat-openbsd curl jq && \
+    apt-get install -y --no-install-recommends git build-essential pandoc gcc netcat-openbsd curl jq && \
     apt-get install -y --no-install-recommends gcc python3-dev && \
     # for RAG OCR
     apt-get install -y --no-install-recommends ffmpeg libsm6 libxext6 && \
@@ -157,5 +161,6 @@ USER $UID:$GID
 
 ARG BUILD_HASH
 ENV WEBUI_BUILD_VERSION=${BUILD_HASH}
+ENV DOCKER true
 
 CMD [ "bash", "start.sh"]

backend/.gitignore

@@ -8,9 +8,5 @@ _test
 Pipfile
 !/data
 /data/*
-!/data/litellm
-/data/litellm/*
-!data/litellm/config.yaml
-
-!data/config.json
+/open_webui/data/*
 .webui_secret_key

backend/open_webui/__init__.py

@@ -16,6 +16,7 @@ def serve(
     host: str = "0.0.0.0",
     port: int = 8080,
 ):
+    os.environ["FROM_INIT_PY"] = "true"
     if os.getenv("WEBUI_SECRET_KEY") is None:
         typer.echo(
             "Loading WEBUI_SECRET_KEY from file, not provided as an environment variable."
@@ -38,6 +39,19 @@ def serve(
                 "/usr/local/lib/python3.11/site-packages/nvidia/cudnn/lib",
             ]
         )
+        try:
+            import torch
+
+            assert torch.cuda.is_available(), "CUDA not available"
+            typer.echo("CUDA seems to be working")
+        except Exception as e:
+            typer.echo(
+                "Error when testing CUDA but USE_CUDA_DOCKER is true. "
+                "Resetting USE_CUDA_DOCKER to false and removing "
+                f"LD_LIBRARY_PATH modifications: {e}"
+            )
+            os.environ["USE_CUDA_DOCKER"] = "false"
+            os.environ["LD_LIBRARY_PATH"] = ":".join(LD_LIBRARY_PATH)
     import open_webui.main  # we need set environment variables before importing main
 
     uvicorn.run(open_webui.main.app, host=host, port=port, forwarded_allow_ips="*")

backend/open_webui/apps/audio/main.py

@@ -19,16 +19,18 @@ from open_webui.config import (
     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,
-    DEVICE_TYPE,
     WHISPER_MODEL,
     WHISPER_MODEL_AUTO_UPDATE,
     WHISPER_MODEL_DIR,
     AppConfig,
 )
+
 from open_webui.constants import ERROR_MESSAGES
-from open_webui.env import SRC_LOG_LEVELS
+from open_webui.env import SRC_LOG_LEVELS, DEVICE_TYPE
 from fastapi import Depends, FastAPI, File, HTTPException, Request, UploadFile, status
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import FileResponse
@@ -62,6 +64,9 @@ 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.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}")
@@ -78,6 +83,8 @@ class TTSConfigForm(BaseModel):
     MODEL: str
     VOICE: str
     SPLIT_ON: str
+    AZURE_SPEECH_REGION: str
+    AZURE_SPEECH_OUTPUT_FORMAT: str
 
 
 class STTConfigForm(BaseModel):
@@ -130,6 +137,8 @@ async def get_audio_config(user=Depends(get_admin_user)):
             "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,
@@ -151,6 +160,10 @@ async def update_audio_config(
     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
@@ -166,6 +179,8 @@ async def update_audio_config(
             "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,
@@ -301,6 +316,42 @@ async def speech(request: Request, user=Depends(get_verified_user)):
                 detail=error_detail,
             )
 
+    elif app.state.config.TTS_ENGINE == "azure":
+        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")
+
+        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>"""
+
+        response = requests.post(url, headers=headers, data=data)
+
+        if response.status_code == 200:
+            with open(file_path, "wb") as f:
+                f.write(response.content)
+            return FileResponse(file_path)
+        else:
+            log.error(f"Error synthesizing speech - {response.reason}")
+            raise HTTPException(
+                status_code=500, detail=f"Error synthesizing speech - {response.reason}"
+            )
+
 
 @app.post("/transcriptions")
 def transcribe(
@@ -309,7 +360,7 @@ def transcribe(
 ):
     log.info(f"file.content_type: {file.content_type}")
 
-    if file.content_type not in ["audio/mpeg", "audio/wav"]:
+    if file.content_type not in ["audio/mpeg", "audio/wav", "audio/ogg"]:
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=ERROR_MESSAGES.FILE_NOT_SUPPORTED,
@@ -443,7 +494,7 @@ def get_available_models() -> list[dict]:
 
         try:
             response = requests.get(
-                "https://api.elevenlabs.io/v1/models", headers=headers
+                "https://api.elevenlabs.io/v1/models", headers=headers, timeout=5
             )
             response.raise_for_status()
             models = response.json()
@@ -478,6 +529,21 @@ def get_available_voices() -> dict:
         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
 

backend/open_webui/apps/images/main.py

@@ -17,6 +17,9 @@ from open_webui.apps.images.utils.comfyui import (
 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,
@@ -65,6 +68,9 @@ 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
@@ -85,6 +91,9 @@ async def get_config(request: Request, user=Depends(get_admin_user)):
         "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,
@@ -102,6 +111,9 @@ class OpenAIConfigForm(BaseModel):
 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):
@@ -133,6 +145,22 @@ async def update_config(form_data: ConfigForm, user=Depends(get_admin_user)):
         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
@@ -147,6 +175,9 @@ async def update_config(form_data: ConfigForm, user=Depends(get_admin_user)):
         "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,
@@ -439,7 +470,9 @@ async def image_generations(
                 "response_format": "b64_json",
             }
 
-            r = requests.post(
+            # 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,
@@ -524,6 +557,15 @@ async def image_generations(
             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,

backend/open_webui/apps/ollama/main.py

@@ -28,11 +28,15 @@ from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import StreamingResponse
 from pydantic import BaseModel, ConfigDict
 from starlette.background import BackgroundTask
+
+
 from open_webui.utils.misc import (
+    calculate_sha256,
+)
+from open_webui.utils.payload import (
     apply_model_params_to_body_ollama,
     apply_model_params_to_body_openai,
     apply_model_system_prompt_to_body,
-    calculate_sha256,
 )
 from open_webui.utils.utils import get_admin_user, get_verified_user
 
@@ -539,6 +543,57 @@ class GenerateEmbeddingsForm(BaseModel):
     keep_alive: Optional[Union[int, str]] = None
 
 
+@app.post("/api/embed")
+@app.post("/api/embed/{url_idx}")
+async def generate_embeddings(
+    form_data: GenerateEmbeddingsForm,
+    url_idx: Optional[int] = None,
+    user=Depends(get_verified_user),
+):
+    if url_idx is None:
+        model = form_data.model
+
+        if ":" not in model:
+            model = f"{model}:latest"
+
+        if model in app.state.MODELS:
+            url_idx = random.choice(app.state.MODELS[model]["urls"])
+        else:
+            raise HTTPException(
+                status_code=400,
+                detail=ERROR_MESSAGES.MODEL_NOT_FOUND(form_data.model),
+            )
+
+    url = app.state.config.OLLAMA_BASE_URLS[url_idx]
+    log.info(f"url: {url}")
+
+    r = requests.request(
+        method="POST",
+        url=f"{url}/api/embed",
+        headers={"Content-Type": "application/json"},
+        data=form_data.model_dump_json(exclude_none=True).encode(),
+    )
+    try:
+        r.raise_for_status()
+
+        return r.json()
+    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"Ollama: {res['error']}"
+            except Exception:
+                error_detail = f"Ollama: {e}"
+
+        raise HTTPException(
+            status_code=r.status_code if r else 500,
+            detail=error_detail,
+        )
+
+
 @app.post("/api/embeddings")
 @app.post("/api/embeddings/{url_idx}")
 async def generate_embeddings(
@@ -761,7 +816,10 @@ async def generate_chat_completion(
     log.debug(payload)
 
     return await post_streaming_url(
-        f"{url}/api/chat", json.dumps(payload), content_type="application/x-ndjson"
+        f"{url}/api/chat",
+        json.dumps(payload),
+        stream=form_data.stream,
+        content_type="application/x-ndjson",
     )
 
 

backend/open_webui/apps/openai/main.py

@@ -26,10 +26,13 @@ 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.misc import (
+
+
+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
 
 log = logging.getLogger(__name__)
@@ -217,7 +220,17 @@ def merge_models_lists(model_lists):
                     for model in models
                     if "api.openai.com"
                     not in app.state.config.OPENAI_API_BASE_URLS[idx]
-                    or "gpt" in model["id"]
+                    or not any(
+                        name in model["id"]
+                        for name in [
+                            "babbage",
+                            "dall-e",
+                            "davinci",
+                            "embedding",
+                            "tts",
+                            "whisper",
+                        ]
+                    )
                 ]
             )
 
@@ -320,10 +333,24 @@ async def get_models(url_idx: Optional[int] = None, user=Depends(get_verified_us
             r.raise_for_status()
 
             response_data = r.json()
+
             if "api.openai.com" in url:
-                response_data["data"] = list(
-                    filter(lambda model: "gpt" in model["id"], response_data["data"])
-                )
+                # Filter the response data
+                response_data["data"] = [
+                    model
+                    for model in response_data["data"]
+                    if not any(
+                        name in model["id"]
+                        for name in [
+                            "babbage",
+                            "dall-e",
+                            "davinci",
+                            "embedding",
+                            "tts",
+                            "whisper",
+                        ]
+                    )
+                ]
 
             return response_data
         except Exception as e:
@@ -378,14 +405,19 @@ async def generate_chat_completion(
             "role": user.role,
         }
 
+    url = app.state.config.OPENAI_API_BASE_URLS[idx]
+    key = app.state.config.OPENAI_API_KEYS[idx]
+
+    # Change max_completion_tokens to max_tokens (Backward compatible)
+    if "api.openai.com" not in url and not payload["model"].lower().startswith("o1-"):
+        if "max_completion_tokens" in payload:
+            payload["max_tokens"] = payload.pop("max_completion_tokens")
+
     # Convert the modified body back to JSON
     payload = json.dumps(payload)
 
     log.debug(payload)
 
-    url = app.state.config.OPENAI_API_BASE_URLS[idx]
-    key = app.state.config.OPENAI_API_KEYS[idx]
-
     headers = {}
     headers["Authorization"] = f"Bearer {key}"
     headers["Content-Type"] = "application/json"
@@ -396,6 +428,7 @@ async def generate_chat_completion(
     r = None
     session = None
     streaming = False
+    response = None
 
     try:
         session = aiohttp.ClientSession(
@@ -408,8 +441,6 @@ async def generate_chat_completion(
             headers=headers,
         )
 
-        r.raise_for_status()
-
         # Check if response is SSE
         if "text/event-stream" in r.headers.get("Content-Type", ""):
             streaming = True
@@ -422,19 +453,23 @@ async def generate_chat_completion(
                 ),
             )
         else:
-            response_data = await r.json()
-            return response_data
+            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 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}"
+        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:

backend/open_webui/apps/rag/main.py

@@ -10,13 +10,21 @@ from datetime import datetime
 from pathlib import Path
 from typing import Iterator, Optional, Sequence, Union
 
+
+import numpy as np
+import torch
 import requests
 import validators
+
+from fastapi import Depends, FastAPI, File, Form, HTTPException, UploadFile, status
+from fastapi.middleware.cors import CORSMiddleware
+from pydantic import BaseModel
+
+from open_webui.apps.rag.search.main import SearchResult
 from open_webui.apps.rag.search.brave import search_brave
 from open_webui.apps.rag.search.duckduckgo import search_duckduckgo
 from open_webui.apps.rag.search.google_pse import search_google_pse
 from open_webui.apps.rag.search.jina_search import search_jina
-from open_webui.apps.rag.search.main import SearchResult
 from open_webui.apps.rag.search.searchapi import search_searchapi
 from open_webui.apps.rag.search.searxng import search_searxng
 from open_webui.apps.rag.search.serper import search_serper
@@ -33,15 +41,12 @@ from open_webui.apps.rag.utils import (
 )
 from open_webui.apps.webui.models.documents import DocumentForm, Documents
 from open_webui.apps.webui.models.files import Files
-from chromadb.utils.batch_utils import create_batches
 from open_webui.config import (
     BRAVE_SEARCH_API_KEY,
-    CHROMA_CLIENT,
     CHUNK_OVERLAP,
     CHUNK_SIZE,
     CONTENT_EXTRACTION_ENGINE,
     CORS_ALLOW_ORIGIN,
-    DEVICE_TYPE,
     DOCS_DIR,
     ENABLE_RAG_HYBRID_SEARCH,
     ENABLE_RAG_LOCAL_WEB_FETCH,
@@ -64,6 +69,7 @@ from open_webui.config import (
     RAG_RERANKING_MODEL,
     RAG_RERANKING_MODEL_AUTO_UPDATE,
     RAG_RERANKING_MODEL_TRUST_REMOTE_CODE,
+    DEFAULT_RAG_TEMPLATE,
     RAG_TEMPLATE,
     RAG_TOP_K,
     RAG_WEB_SEARCH_CONCURRENT_REQUESTS,
@@ -84,9 +90,16 @@ from open_webui.config import (
     AppConfig,
 )
 from open_webui.constants import ERROR_MESSAGES
-from open_webui.env import SRC_LOG_LEVELS
-from fastapi import Depends, FastAPI, File, Form, HTTPException, UploadFile, status
-from fastapi.middleware.cors import CORSMiddleware
+from open_webui.env import SRC_LOG_LEVELS, DEVICE_TYPE, DOCKER
+from open_webui.utils.misc import (
+    calculate_sha256,
+    calculate_sha256_string,
+    extract_folders_after_data_docs,
+    sanitize_filename,
+)
+from open_webui.utils.utils import get_admin_user, get_verified_user
+from open_webui.apps.rag.vector.connector import VECTOR_DB_CLIENT
+
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain_community.document_loaders import (
     BSHTMLLoader,
@@ -105,14 +118,8 @@ from langchain_community.document_loaders import (
     YoutubeLoader,
 )
 from langchain_core.documents import Document
-from pydantic import BaseModel
-from open_webui.utils.misc import (
-    calculate_sha256,
-    calculate_sha256_string,
-    extract_folders_after_data_docs,
-    sanitize_filename,
-)
-from open_webui.utils.utils import get_admin_user, get_verified_user
+from colbert.infra import ColBERTConfig
+from colbert.modeling.checkpoint import Checkpoint
 
 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["RAG"])
@@ -143,13 +150,11 @@ app.state.config.RAG_EMBEDDING_OPENAI_BATCH_SIZE = RAG_EMBEDDING_OPENAI_BATCH_SI
 app.state.config.RAG_RERANKING_MODEL = RAG_RERANKING_MODEL
 app.state.config.RAG_TEMPLATE = RAG_TEMPLATE
 
-
 app.state.config.OPENAI_API_BASE_URL = RAG_OPENAI_API_BASE_URL
 app.state.config.OPENAI_API_KEY = RAG_OPENAI_API_KEY
 
 app.state.config.PDF_EXTRACT_IMAGES = PDF_EXTRACT_IMAGES
 
-
 app.state.config.YOUTUBE_LOADER_LANGUAGE = YOUTUBE_LOADER_LANGUAGE
 app.state.YOUTUBE_LOADER_TRANSLATION = None
 
@@ -175,13 +180,13 @@ app.state.config.RAG_WEB_SEARCH_CONCURRENT_REQUESTS = RAG_WEB_SEARCH_CONCURRENT_
 
 def update_embedding_model(
     embedding_model: str,
-    update_model: bool = False,
+    auto_update: bool = False,
 ):
     if embedding_model and app.state.config.RAG_EMBEDDING_ENGINE == "":
         import sentence_transformers
 
         app.state.sentence_transformer_ef = sentence_transformers.SentenceTransformer(
-            get_model_path(embedding_model, update_model),
+            get_model_path(embedding_model, auto_update),
             device=DEVICE_TYPE,
             trust_remote_code=RAG_EMBEDDING_MODEL_TRUST_REMOTE_CODE,
         )
@@ -191,16 +196,108 @@ def update_embedding_model(
 
 def update_reranking_model(
     reranking_model: str,
-    update_model: bool = False,
+    auto_update: bool = False,
 ):
     if reranking_model:
-        import sentence_transformers
+        if any(model in reranking_model for model in ["jinaai/jina-colbert-v2"]):
 
-        app.state.sentence_transformer_rf = sentence_transformers.CrossEncoder(
-            get_model_path(reranking_model, update_model),
-            device=DEVICE_TYPE,
-            trust_remote_code=RAG_RERANKING_MODEL_TRUST_REMOTE_CODE,
-        )
+            class ColBERT:
+                def __init__(self, name) -> None:
+                    print("ColBERT: Loading model", name)
+                    self.device = "cuda" if torch.cuda.is_available() else "cpu"
+
+                    if DOCKER:
+                        # This is a workaround for the issue with the docker container
+                        # where the torch extension is not loaded properly
+                        # and the following error is thrown:
+                        # /root/.cache/torch_extensions/py311_cpu/segmented_maxsim_cpp/segmented_maxsim_cpp.so: cannot open shared object file: No such file or directory
+
+                        lock_file = "/root/.cache/torch_extensions/py311_cpu/segmented_maxsim_cpp/lock"
+                        if os.path.exists(lock_file):
+                            os.remove(lock_file)
+
+                    self.ckpt = Checkpoint(
+                        name,
+                        colbert_config=ColBERTConfig(model_name=name),
+                    ).to(self.device)
+                    pass
+
+                def calculate_similarity_scores(
+                    self, query_embeddings, document_embeddings
+                ):
+
+                    query_embeddings = query_embeddings.to(self.device)
+                    document_embeddings = document_embeddings.to(self.device)
+
+                    # Validate dimensions to ensure compatibility
+                    if query_embeddings.dim() != 3:
+                        raise ValueError(
+                            f"Expected query embeddings to have 3 dimensions, but got {query_embeddings.dim()}."
+                        )
+                    if document_embeddings.dim() != 3:
+                        raise ValueError(
+                            f"Expected document embeddings to have 3 dimensions, but got {document_embeddings.dim()}."
+                        )
+                    if query_embeddings.size(0) not in [1, document_embeddings.size(0)]:
+                        raise ValueError(
+                            "There should be either one query or queries equal to the number of documents."
+                        )
+
+                    # Transpose the query embeddings to align for matrix multiplication
+                    transposed_query_embeddings = query_embeddings.permute(0, 2, 1)
+                    # Compute similarity scores using batch matrix multiplication
+                    computed_scores = torch.matmul(
+                        document_embeddings, transposed_query_embeddings
+                    )
+                    # Apply max pooling to extract the highest semantic similarity across each document's sequence
+                    maximum_scores = torch.max(computed_scores, dim=1).values
+
+                    # Sum up the maximum scores across features to get the overall document relevance scores
+                    final_scores = maximum_scores.sum(dim=1)
+
+                    normalized_scores = torch.softmax(final_scores, dim=0)
+
+                    return normalized_scores.detach().cpu().numpy().astype(np.float32)
+
+                def predict(self, sentences):
+
+                    query = sentences[0][0]
+                    docs = [i[1] for i in sentences]
+
+                    # Embedding the documents
+                    embedded_docs = self.ckpt.docFromText(docs, bsize=32)[0]
+                    # Embedding the queries
+                    embedded_queries = self.ckpt.queryFromText([query], bsize=32)
+                    embedded_query = embedded_queries[0]
+
+                    # Calculate retrieval scores for the query against all documents
+                    scores = self.calculate_similarity_scores(
+                        embedded_query.unsqueeze(0), embedded_docs
+                    )
+
+                    return scores
+
+            try:
+                app.state.sentence_transformer_rf = ColBERT(
+                    get_model_path(reranking_model, auto_update)
+                )
+            except Exception as e:
+                log.error(f"ColBERT: {e}")
+                app.state.sentence_transformer_rf = None
+                app.state.config.ENABLE_RAG_HYBRID_SEARCH = False
+        else:
+            import sentence_transformers
+
+            try:
+                app.state.sentence_transformer_rf = sentence_transformers.CrossEncoder(
+                    get_model_path(reranking_model, auto_update),
+                    device=DEVICE_TYPE,
+                    trust_remote_code=RAG_RERANKING_MODEL_TRUST_REMOTE_CODE,
+                )
+            except:
+                log.error("CrossEncoder error")
+                app.state.sentence_transformer_rf = None
+                app.state.config.ENABLE_RAG_HYBRID_SEARCH = False
     else:
         app.state.sentence_transformer_rf = None
 
@@ -593,7 +690,7 @@ async def update_query_settings(
     form_data: QuerySettingsForm, user=Depends(get_admin_user)
 ):
     app.state.config.RAG_TEMPLATE = (
-        form_data.template if form_data.template else RAG_TEMPLATE
+        form_data.template if form_data.template != "" else DEFAULT_RAG_TEMPLATE
     )
     app.state.config.TOP_K = form_data.k if form_data.k else 4
     app.state.config.RELEVANCE_THRESHOLD = form_data.r if form_data.r else 0.0
@@ -998,41 +1095,39 @@ def store_docs_in_vector_db(
 
     try:
         if overwrite:
-            for collection in CHROMA_CLIENT.list_collections():
-                if collection_name == collection.name:
-                    log.info(f"deleting existing collection {collection_name}")
-                    CHROMA_CLIENT.delete_collection(name=collection_name)
+            if VECTOR_DB_CLIENT.has_collection(collection_name=collection_name):
+                log.info(f"deleting existing collection {collection_name}")
+                VECTOR_DB_CLIENT.delete_collection(collection_name=collection_name)
 
-        collection = CHROMA_CLIENT.create_collection(name=collection_name)
-
-        embedding_func = get_embedding_function(
-            app.state.config.RAG_EMBEDDING_ENGINE,
-            app.state.config.RAG_EMBEDDING_MODEL,
-            app.state.sentence_transformer_ef,
-            app.state.config.OPENAI_API_KEY,
-            app.state.config.OPENAI_API_BASE_URL,
-            app.state.config.RAG_EMBEDDING_OPENAI_BATCH_SIZE,
-        )
-
-        embedding_texts = list(map(lambda x: x.replace("\n", " "), texts))
-        embeddings = embedding_func(embedding_texts)
-
-        for batch in create_batches(
-            api=CHROMA_CLIENT,
-            ids=[str(uuid.uuid4()) for _ in texts],
-            metadatas=metadatas,
-            embeddings=embeddings,
-            documents=texts,
-        ):
-            collection.add(*batch)
-
-        return True
-    except Exception as e:
-        if e.__class__.__name__ == "UniqueConstraintError":
+        if VECTOR_DB_CLIENT.has_collection(collection_name=collection_name):
+            log.info(f"collection {collection_name} already exists")
             return True
+        else:
+            embedding_function = get_embedding_function(
+                app.state.config.RAG_EMBEDDING_ENGINE,
+                app.state.config.RAG_EMBEDDING_MODEL,
+                app.state.sentence_transformer_ef,
+                app.state.config.OPENAI_API_KEY,
+                app.state.config.OPENAI_API_BASE_URL,
+                app.state.config.RAG_EMBEDDING_OPENAI_BATCH_SIZE,
+            )
 
+            VECTOR_DB_CLIENT.insert(
+                collection_name=collection_name,
+                items=[
+                    {
+                        "id": str(uuid.uuid4()),
+                        "text": text,
+                        "vector": embedding_function(text.replace("\n", " ")),
+                        "metadata": metadatas[idx],
+                    }
+                    for idx, text in enumerate(texts)
+                ],
+            )
+
+            return True
+    except Exception as e:
         log.exception(e)
-
         return False
 
 
@@ -1158,7 +1253,7 @@ def get_loader(filename: str, file_content_type: str, file_path: str):
         elif (
             file_content_type
             == "application/vnd.openxmlformats-officedocument.wordprocessingml.document"
-            or file_ext in ["doc", "docx"]
+            or file_ext == "docx"
         ):
             loader = Docx2txtLoader(file_path)
         elif file_content_type in [
@@ -1396,7 +1491,7 @@ def scan_docs_dir(user=Depends(get_admin_user)):
 
 @app.post("/reset/db")
 def reset_vector_db(user=Depends(get_admin_user)):
-    CHROMA_CLIENT.reset()
+    VECTOR_DB_CLIENT.reset()
 
 
 @app.post("/reset/uploads")
@@ -1437,7 +1532,7 @@ def reset(user=Depends(get_admin_user)) -> bool:
             log.error("Failed to delete %s. Reason: %s" % (file_path, e))
 
     try:
-        CHROMA_CLIENT.reset()
+        VECTOR_DB_CLIENT.reset()
     except Exception as e:
         log.exception(e)
 

backend/open_webui/apps/rag/utils.py

@@ -1,24 +1,68 @@
 import logging
 import os
+import uuid
 from typing import Optional, Union
 
 import requests
-from open_webui.apps.ollama.main import (
-    GenerateEmbeddingsForm,
-    generate_ollama_embeddings,
-)
-from open_webui.config import CHROMA_CLIENT
-from open_webui.env import SRC_LOG_LEVELS
+
 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.ollama.main import (
+    GenerateEmbeddingsForm,
+    generate_ollama_embeddings,
+)
+from open_webui.apps.rag.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: str,
@@ -26,17 +70,18 @@ def query_doc(
     k: int,
 ):
     try:
-        collection = CHROMA_CLIENT.get_collection(name=collection_name)
-        query_embeddings = embedding_function(query)
-
-        result = collection.query(
-            query_embeddings=[query_embeddings],
-            n_results=k,
+        result = VECTOR_DB_CLIENT.search(
+            collection_name=collection_name,
+            vectors=[embedding_function(query)],
+            limit=k,
         )
 
+        print("result", result)
+
         log.info(f"query_doc:result {result}")
         return result
     except Exception as e:
+        print(e)
         raise e
 
 
@@ -47,27 +92,25 @@ def query_doc_with_hybrid_search(
     k: int,
     reranking_function,
     r: float,
-):
+) -> dict:
     try:
-        collection = CHROMA_CLIENT.get_collection(name=collection_name)
-        documents = collection.get()  # get all documents
+        result = VECTOR_DB_CLIENT.get(collection_name=collection_name)
 
         bm25_retriever = BM25Retriever.from_texts(
-            texts=documents.get("documents"),
-            metadatas=documents.get("metadatas"),
+            texts=result.documents[0],
+            metadatas=result.metadatas[0],
         )
         bm25_retriever.k = k
 
-        chroma_retriever = ChromaRetriever(
-            collection=collection,
+        vector_search_retriever = VectorSearchRetriever(
+            collection_name=collection_name,
             embedding_function=embedding_function,
-            top_n=k,
+            top_k=k,
         )
 
         ensemble_retriever = EnsembleRetriever(
-            retrievers=[bm25_retriever, chroma_retriever], weights=[0.5, 0.5]
+            retrievers=[bm25_retriever, vector_search_retriever], weights=[0.5, 0.5]
         )
-
         compressor = RerankCompressor(
             embedding_function=embedding_function,
             top_n=k,
@@ -92,7 +135,9 @@ def query_doc_with_hybrid_search(
         raise e
 
 
-def merge_and_sort_query_results(query_results, k, reverse=False):
+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 = []
@@ -138,7 +183,7 @@ def query_collection(
     query: str,
     embedding_function,
     k: int,
-):
+) -> dict:
     results = []
     for collection_name in collection_names:
         if collection_name:
@@ -149,9 +194,9 @@ def query_collection(
                     k=k,
                     embedding_function=embedding_function,
                 )
-                results.append(result)
-            except Exception:
-                pass
+                results.append(result.model_dump())
+            except Exception as e:
+                log.exception(f"Error when querying the collection: {e}")
         else:
             pass
 
@@ -165,8 +210,9 @@ def query_collection_with_hybrid_search(
     k: int,
     reranking_function,
     r: float,
-):
+) -> dict:
     results = []
+    error = False
     for collection_name in collection_names:
         try:
             result = query_doc_with_hybrid_search(
@@ -178,14 +224,39 @@ def query_collection_with_hybrid_search(
                 r=r,
             )
             results.append(result)
-        except Exception:
-            pass
+        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 rag_template(template: str, context: str, query: str):
-    template = template.replace("[context]", context)
-    template = template.replace("[query]", query)
+    count = template.count("[context]")
+    assert "[context]" in template, "RAG template does not contain '[context]'"
+
+    if "<context>" in context and "</context>" in context:
+        log.debug(
+            "WARNING: Potential prompt injection attack: the RAG "
+            "context contains '<context>' and '</context>'. This might be "
+            "nothing, or the user might be trying to hack something."
+        )
+
+    if "[query]" in context:
+        query_placeholder = f"[query-{str(uuid.uuid4())}]"
+        template = template.replace("[query]", query_placeholder)
+        template = template.replace("[context]", context)
+        template = template.replace(query_placeholder, query)
+    else:
+        template = template.replace("[context]", context)
+        template = template.replace("[query]", query)
     return template
 
 
@@ -262,19 +333,27 @@ def get_rag_context(
             continue
 
         try:
+            context = None
             if file["type"] == "text":
                 context = file["content"]
             else:
                 if hybrid_search:
-                    context = query_collection_with_hybrid_search(
-                        collection_names=collection_names,
-                        query=query,
-                        embedding_function=embedding_function,
-                        k=k,
-                        reranking_function=reranking_function,
-                        r=r,
-                    )
-                else:
+                    try:
+                        context = query_collection_with_hybrid_search(
+                            collection_names=collection_names,
+                            query=query,
+                            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,
                         query=query,
@@ -283,7 +362,6 @@ def get_rag_context(
                     )
         except Exception as e:
             log.exception(e)
-            context = None
 
         if context:
             relevant_contexts.append({**context, "source": file})
@@ -391,51 +469,11 @@ def generate_openai_batch_embeddings(
         return None
 
 
-from typing import Any
-
-from langchain_core.callbacks import CallbackManagerForRetrieverRun
-from langchain_core.retrievers import BaseRetriever
-
-
-class ChromaRetriever(BaseRetriever):
-    collection: Any
-    embedding_function: Any
-    top_n: int
-
-    def _get_relevant_documents(
-        self,
-        query: str,
-        *,
-        run_manager: CallbackManagerForRetrieverRun,
-    ) -> list[Document]:
-        query_embeddings = self.embedding_function(query)
-
-        results = self.collection.query(
-            query_embeddings=[query_embeddings],
-            n_results=self.top_n,
-        )
-
-        ids = results["ids"][0]
-        metadatas = results["metadatas"][0]
-        documents = results["documents"][0]
-
-        results = []
-        for idx in range(len(ids)):
-            results.append(
-                Document(
-                    metadata=metadatas[idx],
-                    page_content=documents[idx],
-                )
-            )
-        return results
-
-
 import operator
 from typing import Optional, Sequence
 
 from langchain_core.callbacks import Callbacks
 from langchain_core.documents import BaseDocumentCompressor, Document
-from langchain_core.pydantic_v1 import Extra
 
 
 class RerankCompressor(BaseDocumentCompressor):
@@ -445,7 +483,7 @@ class RerankCompressor(BaseDocumentCompressor):
     r_score: float
 
     class Config:
-        extra = Extra.forbid
+        extra = "forbid"
         arbitrary_types_allowed = True
 
     def compress_documents(

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

@@ -0,0 +1,10 @@
+from open_webui.apps.rag.vector.dbs.chroma import ChromaClient
+from open_webui.apps.rag.vector.dbs.milvus import MilvusClient
+
+
+from open_webui.config import VECTOR_DB
+
+if VECTOR_DB == "milvus":
+    VECTOR_DB_CLIENT = MilvusClient()
+else:
+    VECTOR_DB_CLIENT = ChromaClient()

backend/open_webui/apps/rag/vector/dbs/chroma.py

@@ -0,0 +1,122 @@
+import chromadb
+from chromadb import Settings
+from chromadb.utils.batch_utils import create_batches
+
+from typing import Optional
+
+from open_webui.apps.rag.vector.main import VectorItem, SearchResult, GetResult
+from open_webui.config import (
+    CHROMA_DATA_PATH,
+    CHROMA_HTTP_HOST,
+    CHROMA_HTTP_PORT,
+    CHROMA_HTTP_HEADERS,
+    CHROMA_HTTP_SSL,
+    CHROMA_TENANT,
+    CHROMA_DATABASE,
+)
+
+
+class ChromaClient:
+    def __init__(self):
+        if CHROMA_HTTP_HOST != "":
+            self.client = chromadb.HttpClient(
+                host=CHROMA_HTTP_HOST,
+                port=CHROMA_HTTP_PORT,
+                headers=CHROMA_HTTP_HEADERS,
+                ssl=CHROMA_HTTP_SSL,
+                tenant=CHROMA_TENANT,
+                database=CHROMA_DATABASE,
+                settings=Settings(allow_reset=True, anonymized_telemetry=False),
+            )
+        else:
+            self.client = chromadb.PersistentClient(
+                path=CHROMA_DATA_PATH,
+                settings=Settings(allow_reset=True, anonymized_telemetry=False),
+                tenant=CHROMA_TENANT,
+                database=CHROMA_DATABASE,
+            )
+
+    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]
+
+    def delete_collection(self, collection_name: str):
+        # Delete the collection based on the collection name.
+        return self.client.delete_collection(name=collection_name)
+
+    def search(
+        self, collection_name: str, vectors: list[list[float | int]], limit: int
+    ) -> Optional[SearchResult]:
+        # Search for the nearest neighbor items based on the vectors and return 'limit' number of results.
+        collection = self.client.get_collection(name=collection_name)
+        if collection:
+            result = collection.query(
+                query_embeddings=vectors,
+                n_results=limit,
+            )
+
+            return SearchResult(
+                **{
+                    "ids": result["ids"],
+                    "distances": result["distances"],
+                    "documents": result["documents"],
+                    "metadatas": result["metadatas"],
+                }
+            )
+        return None
+
+    def get(self, collection_name: str) -> Optional[GetResult]:
+        # Get all the items in the collection.
+        collection = self.client.get_collection(name=collection_name)
+        if collection:
+            result = collection.get()
+            return GetResult(
+                **{
+                    "ids": [result["ids"]],
+                    "documents": [result["documents"]],
+                    "metadatas": [result["metadatas"]],
+                }
+            )
+        return None
+
+    def insert(self, collection_name: str, items: list[VectorItem]):
+        # Insert the items into the collection, if the collection does not exist, it will be created.
+        collection = self.client.get_or_create_collection(name=collection_name)
+
+        ids = [item["id"] for item in items]
+        documents = [item["text"] for item in items]
+        embeddings = [item["vector"] for item in items]
+        metadatas = [item["metadata"] for item in items]
+
+        for batch in create_batches(
+            api=self.client,
+            documents=documents,
+            embeddings=embeddings,
+            ids=ids,
+            metadatas=metadatas,
+        ):
+            collection.add(*batch)
+
+    def upsert(self, collection_name: str, items: list[VectorItem]):
+        # Update the items in the collection, if the items are not present, insert them. If the collection does not exist, it will be created.
+        collection = self.client.get_or_create_collection(name=collection_name)
+
+        ids = [item["id"] for item in items]
+        documents = [item["text"] for item in items]
+        embeddings = [item["vector"] for item in items]
+        metadatas = [item["metadata"] for item in items]
+
+        collection.upsert(
+            ids=ids, documents=documents, embeddings=embeddings, metadatas=metadatas
+        )
+
+    def delete(self, collection_name: str, ids: list[str]):
+        # Delete the items from the collection based on the ids.
+        collection = self.client.get_collection(name=collection_name)
+        if collection:
+            collection.delete(ids=ids)
+
+    def reset(self):
+        # Resets the database. This will delete all collections and item entries.
+        return self.client.reset()

backend/open_webui/apps/rag/vector/dbs/milvus.py

@@ -0,0 +1,205 @@
+from pymilvus import MilvusClient as Client
+from pymilvus import FieldSchema, DataType
+import json
+
+from typing import Optional
+
+from open_webui.apps.rag.vector.main import VectorItem, SearchResult, GetResult
+from open_webui.config import (
+    MILVUS_URI,
+)
+
+
+class MilvusClient:
+    def __init__(self):
+        self.collection_prefix = "open_webui"
+        self.client = Client(uri=MILVUS_URI)
+
+    def _result_to_get_result(self, result) -> GetResult:
+        print(result)
+
+        ids = []
+        documents = []
+        metadatas = []
+
+        for match in result:
+            _ids = []
+            _documents = []
+            _metadatas = []
+
+            for item in match:
+                _ids.append(item.get("id"))
+                _documents.append(item.get("data", {}).get("text"))
+                _metadatas.append(item.get("metadata"))
+
+            ids.append(_ids)
+            documents.append(_documents)
+            metadatas.append(_metadatas)
+
+        return GetResult(
+            **{
+                "ids": ids,
+                "documents": documents,
+                "metadatas": metadatas,
+            }
+        )
+
+    def _result_to_search_result(self, result) -> SearchResult:
+        print(result)
+
+        ids = []
+        distances = []
+        documents = []
+        metadatas = []
+
+        for match in result:
+            _ids = []
+            _distances = []
+            _documents = []
+            _metadatas = []
+
+            for item in match:
+                _ids.append(item.get("id"))
+                _distances.append(item.get("distance"))
+                _documents.append(item.get("entity", {}).get("data", {}).get("text"))
+                _metadatas.append(item.get("entity", {}).get("metadata"))
+
+            ids.append(_ids)
+            distances.append(_distances)
+            documents.append(_documents)
+            metadatas.append(_metadatas)
+
+        return SearchResult(
+            **{
+                "ids": ids,
+                "distances": distances,
+                "documents": documents,
+                "metadatas": metadatas,
+            }
+        )
+
+    def _create_collection(self, collection_name: str, dimension: int):
+        schema = self.client.create_schema(
+            auto_id=False,
+            enable_dynamic_field=True,
+        )
+        schema.add_field(
+            field_name="id",
+            datatype=DataType.VARCHAR,
+            is_primary=True,
+            max_length=65535,
+        )
+        schema.add_field(
+            field_name="vector",
+            datatype=DataType.FLOAT_VECTOR,
+            dim=dimension,
+            description="vector",
+        )
+        schema.add_field(field_name="data", datatype=DataType.JSON, description="data")
+        schema.add_field(
+            field_name="metadata", datatype=DataType.JSON, description="metadata"
+        )
+
+        index_params = self.client.prepare_index_params()
+        index_params.add_index(
+            field_name="vector", index_type="HNSW", metric_type="COSINE", params={}
+        )
+
+        self.client.create_collection(
+            collection_name=f"{self.collection_prefix}_{collection_name}",
+            schema=schema,
+            index_params=index_params,
+        )
+
+    def has_collection(self, collection_name: str) -> bool:
+        # Check if the collection exists based on the collection name.
+        return self.client.has_collection(
+            collection_name=f"{self.collection_prefix}_{collection_name}"
+        )
+
+    def delete_collection(self, collection_name: str):
+        # Delete the collection based on the collection name.
+        return self.client.drop_collection(
+            collection_name=f"{self.collection_prefix}_{collection_name}"
+        )
+
+    def search(
+        self, collection_name: str, vectors: list[list[float | int]], limit: int
+    ) -> Optional[SearchResult]:
+        # Search for the nearest neighbor items based on the vectors and return 'limit' number of results.
+        result = self.client.search(
+            collection_name=f"{self.collection_prefix}_{collection_name}",
+            data=vectors,
+            limit=limit,
+            output_fields=["data", "metadata"],
+        )
+
+        return self._result_to_search_result(result)
+
+    def get(self, collection_name: str) -> Optional[GetResult]:
+        # Get all the items in the collection.
+        result = self.client.query(
+            collection_name=f"{self.collection_prefix}_{collection_name}",
+            filter='id != ""',
+        )
+        return self._result_to_get_result([result])
+
+    def insert(self, collection_name: str, items: list[VectorItem]):
+        # Insert the items into the collection, if the collection does not exist, it will be created.
+        if not self.client.has_collection(
+            collection_name=f"{self.collection_prefix}_{collection_name}"
+        ):
+            self._create_collection(
+                collection_name=collection_name, dimension=len(items[0]["vector"])
+            )
+
+        return self.client.insert(
+            collection_name=f"{self.collection_prefix}_{collection_name}",
+            data=[
+                {
+                    "id": item["id"],
+                    "vector": item["vector"],
+                    "data": {"text": item["text"]},
+                    "metadata": item["metadata"],
+                }
+                for item in items
+            ],
+        )
+
+    def upsert(self, collection_name: str, items: list[VectorItem]):
+        # Update the items in the collection, if the items are not present, insert them. If the collection does not exist, it will be created.
+        if not self.client.has_collection(
+            collection_name=f"{self.collection_prefix}_{collection_name}"
+        ):
+            self._create_collection(
+                collection_name=collection_name, dimension=len(items[0]["vector"])
+            )
+
+        return self.client.upsert(
+            collection_name=f"{self.collection_prefix}_{collection_name}",
+            data=[
+                {
+                    "id": item["id"],
+                    "vector": item["vector"],
+                    "data": {"text": item["text"]},
+                    "metadata": item["metadata"],
+                }
+                for item in items
+            ],
+        )
+
+    def delete(self, collection_name: str, ids: list[str]):
+        # Delete the items from the collection based on the ids.
+
+        return self.client.delete(
+            collection_name=f"{self.collection_prefix}_{collection_name}",
+            ids=ids,
+        )
+
+    def reset(self):
+        # Resets the database. This will delete all collections and item entries.
+
+        collection_names = self.client.list_collections()
+        for collection_name in collection_names:
+            if collection_name.startswith(self.collection_prefix):
+                self.client.drop_collection(collection_name=collection_name)

backend/open_webui/apps/rag/vector/main.py

@@ -0,0 +1,19 @@
+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/socket/main.py

@@ -2,9 +2,38 @@ import asyncio
 
 import socketio
 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
 
-sio = socketio.AsyncServer(cors_allowed_origins=[], async_mode="asgi")
+
+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,
+    )
+
+
 app = socketio.ASGIApp(sio, socketio_path="/ws/socket.io")
 
 # Dictionary to maintain the user pool
@@ -32,7 +61,7 @@ async def connect(sid, environ, auth):
             else:
                 USER_POOL[user.id] = [sid]
 
-            print(f"user {user.name}({user.id}) connected with session ID {sid}")
+            # print(f"user {user.name}({user.id}) connected with session ID {sid}")
 
             await sio.emit("user-count", {"count": len(set(USER_POOL))})
             await sio.emit("usage", {"models": get_models_in_use()})
@@ -40,7 +69,7 @@ async def connect(sid, environ, auth):
 
 @sio.on("user-join")
 async def user_join(sid, data):
-    print("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:
@@ -60,7 +89,7 @@ async def user_join(sid, data):
     else:
         USER_POOL[user.id] = [sid]
 
-    print(f"user {user.name}({user.id}) connected with session ID {sid}")
+    # print(f"user {user.name}({user.id}) connected with session ID {sid}")
 
     await sio.emit("user-count", {"count": len(set(USER_POOL))})
 
@@ -109,7 +138,7 @@ async def remove_after_timeout(sid, model_id):
     try:
         await asyncio.sleep(TIMEOUT_DURATION)
         if model_id in USAGE_POOL:
-            print(USAGE_POOL[model_id]["sids"])
+            # print(USAGE_POOL[model_id]["sids"])
             USAGE_POOL[model_id]["sids"].remove(sid)
             USAGE_POOL[model_id]["sids"] = list(set(USAGE_POOL[model_id]["sids"]))
 
@@ -136,7 +165,8 @@ async def disconnect(sid):
 
         await sio.emit("user-count", {"count": len(USER_POOL)})
     else:
-        print(f"Unknown session ID {sid} disconnected")
+        pass
+        # print(f"Unknown session ID {sid} disconnected")
 
 
 def get_event_emitter(request_info):

backend/open_webui/apps/webui/main.py

@@ -51,11 +51,15 @@ from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import StreamingResponse
 from pydantic import BaseModel
 from open_webui.utils.misc import (
-    apply_model_params_to_body_openai,
-    apply_model_system_prompt_to_body,
     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()
@@ -152,29 +156,33 @@ async def get_pipe_models():
 
         # Check if function is a manifold
         if hasattr(function_module, "pipes"):
-            manifold_pipes = []
+            sub_pipes = []
 
             # Check if pipes is a function or a list
-            if callable(function_module.pipes):
-                manifold_pipes = function_module.pipes()
-            else:
-                manifold_pipes = function_module.pipes
 
-            for p in manifold_pipes:
-                manifold_pipe_id = f'{pipe.id}.{p["id"]}'
-                manifold_pipe_name = p["name"]
+            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 = []
+
+            print(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"):
-                    manifold_pipe_name = f"{function_module.name}{manifold_pipe_name}"
+                    sub_pipe_name = f"{function_module.name}{sub_pipe_name}"
 
                 pipe_flag = {"type": pipe.type}
-                if hasattr(function_module, "ChatValves"):
-                    pipe_flag["valves_spec"] = function_module.ChatValves.schema()
-
                 pipe_models.append(
                     {
-                        "id": manifold_pipe_id,
-                        "name": manifold_pipe_name,
+                        "id": sub_pipe_id,
+                        "name": sub_pipe_name,
                         "object": "model",
                         "created": pipe.created_at,
                         "owned_by": "openai",
@@ -183,8 +191,6 @@ async def get_pipe_models():
                 )
         else:
             pipe_flag = {"type": "pipe"}
-            if hasattr(function_module, "ChatValves"):
-                pipe_flag["valves_spec"] = function_module.ChatValves.schema()
 
             pipe_models.append(
                 {

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

@@ -190,8 +190,8 @@ async def signin(request: Request, response: Response, form_data: SigninForm):
 async def signup(request: Request, response: Response, form_data: SignupForm):
     if (
         not request.app.state.config.ENABLE_SIGNUP
-        and request.app.state.config.ENABLE_LOGIN_FORM
-        and WEBUI_AUTH
+        or not request.app.state.config.ENABLE_LOGIN_FORM
+        or not WEBUI_AUTH
     ):
         raise HTTPException(
             status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.ACCESS_PROHIBITED

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

@@ -8,7 +8,7 @@ from open_webui.apps.webui.models.functions import (
     FunctionResponse,
     Functions,
 )
-from open_webui.apps.webui.utils import load_function_module_by_id
+from open_webui.apps.webui.utils import load_function_module_by_id, replace_imports
 from open_webui.config import CACHE_DIR, FUNCTIONS_DIR
 from open_webui.constants import ERROR_MESSAGES
 from fastapi import APIRouter, Depends, HTTPException, Request, status
@@ -55,13 +55,11 @@ async def create_new_function(
 
     function = Functions.get_function_by_id(form_data.id)
     if function is None:
-        function_path = os.path.join(FUNCTIONS_DIR, f"{form_data.id}.py")
         try:
-            with open(function_path, "w") as function_file:
-                function_file.write(form_data.content)
-
+            form_data.content = replace_imports(form_data.content)
             function_module, function_type, frontmatter = load_function_module_by_id(
-                form_data.id
+                form_data.id,
+                content=form_data.content,
             )
             form_data.meta.manifest = frontmatter
 
@@ -174,13 +172,11 @@ async def toggle_global_by_id(id: str, user=Depends(get_admin_user)):
 async def update_function_by_id(
     request: Request, id: str, form_data: FunctionForm, user=Depends(get_admin_user)
 ):
-    function_path = os.path.join(FUNCTIONS_DIR, f"{id}.py")
-
     try:
-        with open(function_path, "w") as function_file:
-            function_file.write(form_data.content)
-
-        function_module, function_type, frontmatter = load_function_module_by_id(id)
+        form_data.content = replace_imports(form_data.content)
+        function_module, function_type, frontmatter = load_function_module_by_id(
+            id, content=form_data.content
+        )
         form_data.meta.manifest = frontmatter
 
         FUNCTIONS = request.app.state.FUNCTIONS
@@ -222,13 +218,6 @@ async def delete_function_by_id(
         if id in FUNCTIONS:
             del FUNCTIONS[id]
 
-        # delete the function file
-        function_path = os.path.join(FUNCTIONS_DIR, f"{id}.py")
-        try:
-            os.remove(function_path)
-        except Exception:
-            pass
-
     return result
 
 

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

@@ -1,12 +1,13 @@
+from fastapi import APIRouter, Depends, HTTPException, Request
+from pydantic import BaseModel
 import logging
 from typing import Optional
 
 from open_webui.apps.webui.models.memories import Memories, MemoryModel
-from open_webui.config import CHROMA_CLIENT
-from open_webui.env import SRC_LOG_LEVELS
-from fastapi import APIRouter, Depends, HTTPException, Request
-from pydantic import BaseModel
+from open_webui.apps.rag.vector.connector import VECTOR_DB_CLIENT
 from open_webui.utils.utils import get_verified_user
+from open_webui.env import SRC_LOG_LEVELS
+
 
 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["MODELS"])
@@ -49,14 +50,17 @@ async def add_memory(
     user=Depends(get_verified_user),
 ):
     memory = Memories.insert_new_memory(user.id, form_data.content)
-    memory_embedding = request.app.state.EMBEDDING_FUNCTION(memory.content)
 
-    collection = CHROMA_CLIENT.get_or_create_collection(name=f"user-memory-{user.id}")
-    collection.upsert(
-        documents=[memory.content],
-        ids=[memory.id],
-        embeddings=[memory_embedding],
-        metadatas=[{"created_at": memory.created_at}],
+    VECTOR_DB_CLIENT.upsert(
+        collection_name=f"user-memory-{user.id}",
+        items=[
+            {
+                "id": memory.id,
+                "text": memory.content,
+                "vector": request.app.state.EMBEDDING_FUNCTION(memory.content),
+                "metadata": {"created_at": memory.created_at},
+            }
+        ],
     )
 
     return memory
@@ -76,12 +80,10 @@ class QueryMemoryForm(BaseModel):
 async def query_memory(
     request: Request, form_data: QueryMemoryForm, user=Depends(get_verified_user)
 ):
-    query_embedding = request.app.state.EMBEDDING_FUNCTION(form_data.content)
-    collection = CHROMA_CLIENT.get_or_create_collection(name=f"user-memory-{user.id}")
-
-    results = collection.query(
-        query_embeddings=[query_embedding],
-        n_results=form_data.k,  # how many results to return
+    results = VECTOR_DB_CLIENT.search(
+        collection_name=f"user-memory-{user.id}",
+        vectors=[request.app.state.EMBEDDING_FUNCTION(form_data.content)],
+        limit=form_data.k,
     )
 
     return results
@@ -94,17 +96,25 @@ async def query_memory(
 async def reset_memory_from_vector_db(
     request: Request, user=Depends(get_verified_user)
 ):
-    CHROMA_CLIENT.delete_collection(f"user-memory-{user.id}")
-    collection = CHROMA_CLIENT.get_or_create_collection(name=f"user-memory-{user.id}")
+    VECTOR_DB_CLIENT.delete_collection(f"user-memory-{user.id}")
 
     memories = Memories.get_memories_by_user_id(user.id)
-    for memory in memories:
-        memory_embedding = request.app.state.EMBEDDING_FUNCTION(memory.content)
-        collection.upsert(
-            documents=[memory.content],
-            ids=[memory.id],
-            embeddings=[memory_embedding],
-        )
+    VECTOR_DB_CLIENT.upsert(
+        collection_name=f"user-memory-{user.id}",
+        items=[
+            {
+                "id": memory.id,
+                "text": memory.content,
+                "vector": request.app.state.EMBEDDING_FUNCTION(memory.content),
+                "metadata": {
+                    "created_at": memory.created_at,
+                    "updated_at": memory.updated_at,
+                },
+            }
+            for memory in memories
+        ],
+    )
+
     return True
 
 
@@ -119,7 +129,7 @@ async def delete_memory_by_user_id(user=Depends(get_verified_user)):
 
     if result:
         try:
-            CHROMA_CLIENT.delete_collection(f"user-memory-{user.id}")
+            VECTOR_DB_CLIENT.delete_collection(f"user-memory-{user.id}")
         except Exception as e:
             log.error(e)
         return True
@@ -144,16 +154,18 @@ async def update_memory_by_id(
         raise HTTPException(status_code=404, detail="Memory not found")
 
     if form_data.content is not None:
-        memory_embedding = request.app.state.EMBEDDING_FUNCTION(form_data.content)
-        collection = CHROMA_CLIENT.get_or_create_collection(
-            name=f"user-memory-{user.id}"
-        )
-        collection.upsert(
-            documents=[form_data.content],
-            ids=[memory.id],
-            embeddings=[memory_embedding],
-            metadatas=[
-                {"created_at": memory.created_at, "updated_at": memory.updated_at}
+        VECTOR_DB_CLIENT.upsert(
+            collection_name=f"user-memory-{user.id}",
+            items=[
+                {
+                    "id": memory.id,
+                    "text": memory.content,
+                    "vector": request.app.state.EMBEDDING_FUNCTION(memory.content),
+                    "metadata": {
+                        "created_at": memory.created_at,
+                        "updated_at": memory.updated_at,
+                    },
+                }
             ],
         )
 
@@ -170,10 +182,9 @@ async def delete_memory_by_id(memory_id: str, user=Depends(get_verified_user)):
     result = Memories.delete_memory_by_id_and_user_id(memory_id, user.id)
 
     if result:
-        collection = CHROMA_CLIENT.get_or_create_collection(
-            name=f"user-memory-{user.id}"
+        VECTOR_DB_CLIENT.delete(
+            collection_name=f"user-memory-{user.id}", ids=[memory_id]
         )
-        collection.delete(ids=[memory_id])
         return True
 
     return False

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

@@ -18,8 +18,18 @@ router = APIRouter()
 
 
 @router.get("/", response_model=list[ModelResponse])
-async def get_models(user=Depends(get_verified_user)):
-    return Models.get_all_models()
+async def get_models(id: Optional[str] = None, user=Depends(get_verified_user)):
+    if id:
+        model = Models.get_model_by_id(id)
+        if model:
+            return [model]
+        else:
+            raise HTTPException(
+                status_code=status.HTTP_401_UNAUTHORIZED,
+                detail=ERROR_MESSAGES.NOT_FOUND,
+            )
+    else:
+        return Models.get_all_models()
 
 
 ############################
@@ -50,24 +60,6 @@ async def add_new_model(
             )
 
 
-############################
-# GetModelById
-############################
-
-
-@router.get("/", response_model=Optional[ModelModel])
-async def get_model_by_id(id: str, user=Depends(get_verified_user)):
-    model = Models.get_model_by_id(id)
-
-    if model:
-        return model
-    else:
-        raise HTTPException(
-            status_code=status.HTTP_401_UNAUTHORIZED,
-            detail=ERROR_MESSAGES.NOT_FOUND,
-        )
-
-
 ############################
 # UpdateModelById
 ############################

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

@@ -3,7 +3,7 @@ from pathlib import Path
 from typing import Optional
 
 from open_webui.apps.webui.models.tools import ToolForm, ToolModel, ToolResponse, Tools
-from open_webui.apps.webui.utils import load_toolkit_module_by_id
+from open_webui.apps.webui.utils import load_toolkit_module_by_id, replace_imports
 from open_webui.config import CACHE_DIR, DATA_DIR
 from open_webui.constants import ERROR_MESSAGES
 from fastapi import APIRouter, Depends, HTTPException, Request, status
@@ -59,12 +59,11 @@ async def create_new_toolkit(
 
     toolkit = Tools.get_tool_by_id(form_data.id)
     if toolkit is None:
-        toolkit_path = os.path.join(TOOLS_DIR, f"{form_data.id}.py")
         try:
-            with open(toolkit_path, "w") as tool_file:
-                tool_file.write(form_data.content)
-
-            toolkit_module, frontmatter = load_toolkit_module_by_id(form_data.id)
+            form_data.content = replace_imports(form_data.content)
+            toolkit_module, frontmatter = load_toolkit_module_by_id(
+                form_data.id, content=form_data.content
+            )
             form_data.meta.manifest = frontmatter
 
             TOOLS = request.app.state.TOOLS
@@ -126,13 +125,11 @@ async def update_toolkit_by_id(
     form_data: ToolForm,
     user=Depends(get_admin_user),
 ):
-    toolkit_path = os.path.join(TOOLS_DIR, f"{id}.py")
-
     try:
-        with open(toolkit_path, "w") as tool_file:
-            tool_file.write(form_data.content)
-
-        toolkit_module, frontmatter = load_toolkit_module_by_id(id)
+        form_data.content = replace_imports(form_data.content)
+        toolkit_module, frontmatter = load_toolkit_module_by_id(
+            id, content=form_data.content
+        )
         form_data.meta.manifest = frontmatter
 
         TOOLS = request.app.state.TOOLS
@@ -177,10 +174,6 @@ async def delete_toolkit_by_id(request: Request, id: str, user=Depends(get_admin
         if id in TOOLS:
             del TOOLS[id]
 
-        # delete the toolkit file
-        toolkit_path = os.path.join(TOOLS_DIR, f"{id}.py")
-        os.remove(toolkit_path)
-
     return result
 
 

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

@@ -4,6 +4,7 @@ from pathlib import Path
 import black
 import markdown
 from open_webui.config import DATA_DIR, ENABLE_ADMIN_EXPORT
+from open_webui.env import FONTS_DIR
 from open_webui.constants import ERROR_MESSAGES
 from fastapi import APIRouter, Depends, HTTPException, Response, status
 from fpdf import FPDF
@@ -57,14 +58,11 @@ class ChatForm(BaseModel):
 async def download_chat_as_pdf(
     form_data: ChatForm,
 ):
+    global FONTS_DIR
+
     pdf = FPDF()
     pdf.add_page()
 
-    # When running in docker, workdir is /app/backend, so fonts is in /app/backend/static/fonts
-    FONTS_DIR = Path("./static/fonts")
-
-    # Non Docker Installation
-
     # When running using `pip install` the static directory is in the site packages.
     if not FONTS_DIR.exists():
         FONTS_DIR = Path(site.getsitepackages()[0]) / "static/fonts"

backend/open_webui/apps/webui/utils.py

@@ -3,15 +3,17 @@ import re
 import subprocess
 import sys
 from importlib import util
+import types
+import tempfile
 
 from open_webui.apps.webui.models.functions import Functions
 from open_webui.apps.webui.models.tools import Tools
 from open_webui.config import FUNCTIONS_DIR, TOOLS_DIR
 
 
-def extract_frontmatter(file_path):
+def extract_frontmatter(content):
     """
-    Extract frontmatter as a dictionary from the specified file path.
+    Extract frontmatter as a dictionary from the provided content string.
     """
     frontmatter = {}
     frontmatter_started = False
@@ -19,29 +21,25 @@ def extract_frontmatter(file_path):
     frontmatter_pattern = re.compile(r"^\s*([a-z_]+):\s*(.*)\s*$", re.IGNORECASE)
 
     try:
-        with open(file_path, "r", encoding="utf-8") as file:
-            first_line = file.readline()
-            if first_line.strip() != '"""':
-                # The file doesn't start with triple quotes
-                return {}
+        lines = content.splitlines()
+        if len(lines) < 1 or lines[0].strip() != '"""':
+            # The content doesn't start with triple quotes
+            return {}
 
-            frontmatter_started = True
+        frontmatter_started = True
 
-            for line in file:
-                if '"""' in line:
-                    if frontmatter_started:
-                        frontmatter_ended = True
-                        break
+        for line in lines[1:]:
+            if '"""' in line:
+                if frontmatter_started:
+                    frontmatter_ended = True
+                    break
 
-                if frontmatter_started and not frontmatter_ended:
-                    match = frontmatter_pattern.match(line)
-                    if match:
-                        key, value = match.groups()
-                        frontmatter[key.strip()] = value.strip()
+            if frontmatter_started and not frontmatter_ended:
+                match = frontmatter_pattern.match(line)
+                if match:
+                    key, value = match.groups()
+                    frontmatter[key.strip()] = value.strip()
 
-    except FileNotFoundError:
-        print(f"Error: The file {file_path} does not exist.")
-        return {}
     except Exception as e:
         print(f"An error occurred: {e}")
         return {}
@@ -49,55 +47,101 @@ def extract_frontmatter(file_path):
     return frontmatter
 
 
-def load_toolkit_module_by_id(toolkit_id):
-    toolkit_path = os.path.join(TOOLS_DIR, f"{toolkit_id}.py")
+def replace_imports(content):
+    """
+    Replace the import paths in the content.
+    """
+    replacements = {
+        "from utils": "from open_webui.utils",
+        "from apps": "from open_webui.apps",
+        "from main": "from open_webui.main",
+        "from config": "from open_webui.config",
+    }
 
-    if not os.path.exists(toolkit_path):
+    for old, new in replacements.items():
+        content = content.replace(old, new)
+
+    return content
+
+
+def load_toolkit_module_by_id(toolkit_id, content=None):
+
+    if content is None:
         tool = Tools.get_tool_by_id(toolkit_id)
-        if tool:
-            with open(toolkit_path, "w") as file:
-                file.write(tool.content)
-        else:
+        if not tool:
             raise Exception(f"Toolkit not found: {toolkit_id}")
 
-    spec = util.spec_from_file_location(toolkit_id, toolkit_path)
-    module = util.module_from_spec(spec)
-    frontmatter = extract_frontmatter(toolkit_path)
+        content = tool.content
+
+        content = replace_imports(content)
+        Tools.update_tool_by_id(toolkit_id, {"content": content})
+    else:
+        frontmatter = extract_frontmatter(content)
+        # Install required packages found within the frontmatter
+        install_frontmatter_requirements(frontmatter.get("requirements", ""))
+
+    module_name = f"tool_{toolkit_id}"
+    module = types.ModuleType(module_name)
+    sys.modules[module_name] = module
+
+    # Create a temporary file and use it to define `__file__` so
+    # that it works as expected from the module's perspective.
+    temp_file = tempfile.NamedTemporaryFile(delete=False)
 
     try:
-        install_frontmatter_requirements(frontmatter.get("requirements", ""))
-        spec.loader.exec_module(module)
+        with open(temp_file.name, "w", encoding="utf-8") as f:
+            f.write(content)
+        module.__dict__["__file__"] = temp_file.name
+
+        # Executing the modified content in the created module's namespace
+        exec(content, module.__dict__)
+        frontmatter = extract_frontmatter(content)
         print(f"Loaded module: {module.__name__}")
+
+        # Create and return the object if the class 'Tools' is found in the module
         if hasattr(module, "Tools"):
             return module.Tools(), frontmatter
         else:
-            raise Exception("No Tools class found")
+            raise Exception("No Tools class found in the module")
     except Exception as e:
-        print(f"Error loading module: {toolkit_id}")
-        # Move the file to the error folder
-        os.rename(toolkit_path, f"{toolkit_path}.error")
+        print(f"Error loading module: {toolkit_id}: {e}")
+        del sys.modules[module_name]  # Clean up
         raise e
+    finally:
+        os.unlink(temp_file.name)
 
 
-def load_function_module_by_id(function_id):
-    function_path = os.path.join(FUNCTIONS_DIR, f"{function_id}.py")
-
-    if not os.path.exists(function_path):
+def load_function_module_by_id(function_id, content=None):
+    if content is None:
         function = Functions.get_function_by_id(function_id)
-        if function:
-            with open(function_path, "w") as file:
-                file.write(function.content)
-        else:
+        if not function:
             raise Exception(f"Function not found: {function_id}")
+        content = function.content
 
-    spec = util.spec_from_file_location(function_id, function_path)
-    module = util.module_from_spec(spec)
-    frontmatter = extract_frontmatter(function_path)
-
-    try:
+        content = replace_imports(content)
+        Functions.update_function_by_id(function_id, {"content": content})
+    else:
+        frontmatter = extract_frontmatter(content)
         install_frontmatter_requirements(frontmatter.get("requirements", ""))
-        spec.loader.exec_module(module)
+
+    module_name = f"function_{function_id}"
+    module = types.ModuleType(module_name)
+    sys.modules[module_name] = module
+
+    # Create a temporary file and use it to define `__file__` so
+    # that it works as expected from the module's perspective.
+    temp_file = tempfile.NamedTemporaryFile(delete=False)
+    try:
+        with open(temp_file.name, "w", encoding="utf-8") as f:
+            f.write(content)
+        module.__dict__["__file__"] = temp_file.name
+
+        # Execute the modified content in the created module's namespace
+        exec(content, module.__dict__)
+        frontmatter = extract_frontmatter(content)
         print(f"Loaded module: {module.__name__}")
+
+        # Create appropriate object based on available class type in the module
         if hasattr(module, "Pipe"):
             return module.Pipe(), "pipe", frontmatter
         elif hasattr(module, "Filter"):
@@ -105,12 +149,15 @@ def load_function_module_by_id(function_id):
         elif hasattr(module, "Action"):
             return module.Action(), "action", frontmatter
         else:
-            raise Exception("No Function class found")
+            raise Exception("No Function class found in the module")
     except Exception as e:
-        print(f"Error loading module: {function_id}")
-        # Move the file to the error folder
-        os.rename(function_path, f"{function_path}.error")
+        print(f"Error loading module: {function_id}: {e}")
+        del sys.modules[module_name]  # Cleanup by removing the module in case of error
+
+        Functions.update_function_by_id(function_id, {"is_active": False})
         raise e
+    finally:
+        os.unlink(temp_file.name)
 
 
 def install_frontmatter_requirements(requirements):

backend/open_webui/config.py

@@ -11,7 +11,6 @@ import chromadb
 import requests
 import yaml
 from open_webui.apps.webui.internal.db import Base, get_db
-from chromadb import Settings
 from open_webui.env import (
     OPEN_WEBUI_DIR,
     DATA_DIR,
@@ -540,40 +539,6 @@ Path(TOOLS_DIR).mkdir(parents=True, exist_ok=True)
 FUNCTIONS_DIR = os.getenv("FUNCTIONS_DIR", f"{DATA_DIR}/functions")
 Path(FUNCTIONS_DIR).mkdir(parents=True, exist_ok=True)
 
-
-####################################
-# LITELLM_CONFIG
-####################################
-
-
-def create_config_file(file_path):
-    directory = os.path.dirname(file_path)
-
-    # Check if directory exists, if not, create it
-    if not os.path.exists(directory):
-        os.makedirs(directory)
-
-    # Data to write into the YAML file
-    config_data = {
-        "general_settings": {},
-        "litellm_settings": {},
-        "model_list": [],
-        "router_settings": {},
-    }
-
-    # Write data to YAML file
-    with open(file_path, "w") as file:
-        yaml.dump(config_data, file)
-
-
-LITELLM_CONFIG_PATH = f"{DATA_DIR}/litellm/config.yaml"
-
-# if not os.path.exists(LITELLM_CONFIG_PATH):
-#     log.info("Config file doesn't exist. Creating...")
-#     create_config_file(LITELLM_CONFIG_PATH)
-#     log.info("Config file created successfully.")
-
-
 ####################################
 # OLLAMA_BASE_URL
 ####################################
@@ -898,76 +863,37 @@ TASK_MODEL_EXTERNAL = PersistentConfig(
 TITLE_GENERATION_PROMPT_TEMPLATE = PersistentConfig(
     "TITLE_GENERATION_PROMPT_TEMPLATE",
     "task.title.prompt_template",
-    os.environ.get(
-        "TITLE_GENERATION_PROMPT_TEMPLATE",
-        """Create a concise, 3-5 word title with an emoji as a title for the prompt in the given language. Suitable Emojis for the summary can be used to enhance understanding but avoid quotation marks or special formatting. RESPOND ONLY WITH THE TITLE TEXT.
+    os.environ.get("TITLE_GENERATION_PROMPT_TEMPLATE", ""),
+)
 
-Examples of titles:
-📉 Stock Market Trends
-🍪 Perfect Chocolate Chip Recipe
-Evolution of Music Streaming
-Remote Work Productivity Tips
-Artificial Intelligence in Healthcare
-🎮 Video Game Development Insights
-
-Prompt: {{prompt:middletruncate:8000}}""",
-    ),
+ENABLE_SEARCH_QUERY = PersistentConfig(
+    "ENABLE_SEARCH_QUERY",
+    "task.search.enable",
+    os.environ.get("ENABLE_SEARCH_QUERY", "True").lower() == "true",
 )
 
 
 SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE = PersistentConfig(
     "SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE",
     "task.search.prompt_template",
-    os.environ.get(
-        "SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE",
-        """You are tasked with generating web search queries. Give me an appropriate query to answer my question for google search. Answer with only the query. Today is {{CURRENT_DATE}}.
-        
-Question:
-{{prompt:end:4000}}""",
-    ),
+    os.environ.get("SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE", ""),
 )
 
-SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD = PersistentConfig(
-    "SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD",
-    "task.search.prompt_length_threshold",
-    int(
-        os.environ.get(
-            "SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD",
-            100,
-        )
-    ),
-)
 
 TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE = PersistentConfig(
     "TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE",
     "task.tools.prompt_template",
-    os.environ.get(
-        "TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE",
-        """Available Tools: {{TOOLS}}\nReturn an empty string if no tools match the query. If a function tool matches, construct and return a JSON object in the format {\"name\": \"functionName\", \"parameters\": {\"requiredFunctionParamKey\": \"requiredFunctionParamValue\"}} using the appropriate tool and its parameters. Only return the object and limit the response to the JSON object without additional text.""",
-    ),
+    os.environ.get("TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE", ""),
 )
 
 
 ####################################
-# RAG document content extraction
+# Vector Database
 ####################################
 
-CONTENT_EXTRACTION_ENGINE = PersistentConfig(
-    "CONTENT_EXTRACTION_ENGINE",
-    "rag.CONTENT_EXTRACTION_ENGINE",
-    os.environ.get("CONTENT_EXTRACTION_ENGINE", "").lower(),
-)
-
-TIKA_SERVER_URL = PersistentConfig(
-    "TIKA_SERVER_URL",
-    "rag.tika_server_url",
-    os.getenv("TIKA_SERVER_URL", "http://tika:9998"),  # Default for sidecar deployment
-)
-
-####################################
-# RAG
-####################################
+VECTOR_DB = os.environ.get("VECTOR_DB", "chroma")
 
+# Chroma
 CHROMA_DATA_PATH = f"{DATA_DIR}/vector_db"
 CHROMA_TENANT = os.environ.get("CHROMA_TENANT", chromadb.DEFAULT_TENANT)
 CHROMA_DATABASE = os.environ.get("CHROMA_DATABASE", chromadb.DEFAULT_DATABASE)
@@ -984,8 +910,29 @@ else:
 CHROMA_HTTP_SSL = os.environ.get("CHROMA_HTTP_SSL", "false").lower() == "true"
 # this uses the model defined in the Dockerfile ENV variable. If you dont use docker or docker based deployments such as k8s, the default embedding model will be used (sentence-transformers/all-MiniLM-L6-v2)
 
+# Milvus
+
+MILVUS_URI = os.environ.get("MILVUS_URI", f"{DATA_DIR}/vector_db/milvus.db")
+
+####################################
+# RAG
+####################################
+
+# RAG Content Extraction
+CONTENT_EXTRACTION_ENGINE = PersistentConfig(
+    "CONTENT_EXTRACTION_ENGINE",
+    "rag.CONTENT_EXTRACTION_ENGINE",
+    os.environ.get("CONTENT_EXTRACTION_ENGINE", "").lower(),
+)
+
+TIKA_SERVER_URL = PersistentConfig(
+    "TIKA_SERVER_URL",
+    "rag.tika_server_url",
+    os.getenv("TIKA_SERVER_URL", "http://tika:9998"),  # Default for sidecar deployment
+)
+
 RAG_TOP_K = PersistentConfig(
-    "RAG_TOP_K", "rag.top_k", int(os.environ.get("RAG_TOP_K", "5"))
+    "RAG_TOP_K", "rag.top_k", int(os.environ.get("RAG_TOP_K", "3"))
 )
 RAG_RELEVANCE_THRESHOLD = PersistentConfig(
     "RAG_RELEVANCE_THRESHOLD",
@@ -1074,36 +1021,8 @@ RAG_RERANKING_MODEL_TRUST_REMOTE_CODE = (
     os.environ.get("RAG_RERANKING_MODEL_TRUST_REMOTE_CODE", "").lower() == "true"
 )
 
-
-if CHROMA_HTTP_HOST != "":
-    CHROMA_CLIENT = chromadb.HttpClient(
-        host=CHROMA_HTTP_HOST,
-        port=CHROMA_HTTP_PORT,
-        headers=CHROMA_HTTP_HEADERS,
-        ssl=CHROMA_HTTP_SSL,
-        tenant=CHROMA_TENANT,
-        database=CHROMA_DATABASE,
-        settings=Settings(allow_reset=True, anonymized_telemetry=False),
-    )
-else:
-    CHROMA_CLIENT = chromadb.PersistentClient(
-        path=CHROMA_DATA_PATH,
-        settings=Settings(allow_reset=True, anonymized_telemetry=False),
-        tenant=CHROMA_TENANT,
-        database=CHROMA_DATABASE,
-    )
-
-
-# device type embedding models - "cpu" (default), "cuda" (nvidia gpu required) or "mps" (apple silicon) - choosing this right can lead to better performance
-USE_CUDA = os.environ.get("USE_CUDA_DOCKER", "false")
-
-if USE_CUDA.lower() == "true":
-    DEVICE_TYPE = "cuda"
-else:
-    DEVICE_TYPE = "cpu"
-
 CHUNK_SIZE = PersistentConfig(
-    "CHUNK_SIZE", "rag.chunk_size", int(os.environ.get("CHUNK_SIZE", "1500"))
+    "CHUNK_SIZE", "rag.chunk_size", int(os.environ.get("CHUNK_SIZE", "1000"))
 )
 CHUNK_OVERLAP = PersistentConfig(
     "CHUNK_OVERLAP",
@@ -1111,19 +1030,25 @@ CHUNK_OVERLAP = PersistentConfig(
     int(os.environ.get("CHUNK_OVERLAP", "100")),
 )
 
-DEFAULT_RAG_TEMPLATE = """Use the following context as your learned knowledge, inside <context></context> XML tags.
+DEFAULT_RAG_TEMPLATE = """You are given a user query, some textual context and rules, all inside xml tags. You have to answer the query based on the context while respecting the rules.
+
 <context>
-    [context]
+[context]
 </context>
 
-When answer to user:
-- If you don't know, just say that you don't know.
-- If you don't know when you are not sure, ask for clarification.
-Avoid mentioning that you obtained the information from the context.
-And answer according to the language of the user's question.
+<rules>
+- If you don't know, just say so.
+- If you are not sure, ask for clarification.
+- Answer in the same language as the user query.
+- If the context appears unreadable or of poor quality, tell the user then answer as best as you can.
+- If the answer is not in the context but you think you know the answer, explain that to the user then answer with your own knowledge.
+- Answer directly and without using xml tags.
+</rules>
 
-Given the context information, answer the query.
-Query: [query]"""
+<user_query>
+[query]
+</user_query>
+"""
 
 RAG_TEMPLATE = PersistentConfig(
     "RAG_TEMPLATE",
@@ -1293,6 +1218,37 @@ AUTOMATIC1111_API_AUTH = PersistentConfig(
     os.getenv("AUTOMATIC1111_API_AUTH", ""),
 )
 
+AUTOMATIC1111_CFG_SCALE = PersistentConfig(
+    "AUTOMATIC1111_CFG_SCALE",
+    "image_generation.automatic1111.cfg_scale",
+    (
+        float(os.environ.get("AUTOMATIC1111_CFG_SCALE"))
+        if os.environ.get("AUTOMATIC1111_CFG_SCALE")
+        else None
+    ),
+)
+
+
+AUTOMATIC1111_SAMPLER = PersistentConfig(
+    "AUTOMATIC1111_SAMPLERE",
+    "image_generation.automatic1111.sampler",
+    (
+        os.environ.get("AUTOMATIC1111_SAMPLER")
+        if os.environ.get("AUTOMATIC1111_SAMPLER")
+        else None
+    ),
+)
+
+AUTOMATIC1111_SCHEDULER = PersistentConfig(
+    "AUTOMATIC1111_SCHEDULER",
+    "image_generation.automatic1111.scheduler",
+    (
+        os.environ.get("AUTOMATIC1111_SCHEDULER")
+        if os.environ.get("AUTOMATIC1111_SCHEDULER")
+        else None
+    ),
+)
+
 COMFYUI_BASE_URL = PersistentConfig(
     "COMFYUI_BASE_URL",
     "image_generation.comfyui.base_url",
@@ -1516,3 +1472,17 @@ AUDIO_TTS_SPLIT_ON = PersistentConfig(
     "audio.tts.split_on",
     os.getenv("AUDIO_TTS_SPLIT_ON", "punctuation"),
 )
+
+AUDIO_TTS_AZURE_SPEECH_REGION = PersistentConfig(
+    "AUDIO_TTS_AZURE_SPEECH_REGION",
+    "audio.tts.azure.speech_region",
+    os.getenv("AUDIO_TTS_AZURE_SPEECH_REGION", "eastus"),
+)
+
+AUDIO_TTS_AZURE_SPEECH_OUTPUT_FORMAT = PersistentConfig(
+    "AUDIO_TTS_AZURE_SPEECH_OUTPUT_FORMAT",
+    "audio.tts.azure.speech_output_format",
+    os.getenv(
+        "AUDIO_TTS_AZURE_SPEECH_OUTPUT_FORMAT", "audio-24khz-160kbitrate-mono-mp3"
+    ),
+)

backend/open_webui/env.py

@@ -31,6 +31,28 @@ try:
 except ImportError:
     print("dotenv not installed, skipping...")
 
+DOCKER = os.environ.get("DOCKER", "False").lower() == "true"
+
+# device type embedding models - "cpu" (default), "cuda" (nvidia gpu required) or "mps" (apple silicon) - choosing this right can lead to better performance
+USE_CUDA = os.environ.get("USE_CUDA_DOCKER", "false")
+
+if USE_CUDA.lower() == "true":
+    try:
+        import torch
+
+        assert torch.cuda.is_available(), "CUDA not available"
+        DEVICE_TYPE = "cuda"
+    except Exception as e:
+        cuda_error = (
+            "Error when testing CUDA but USE_CUDA_DOCKER is true. "
+            f"Resetting USE_CUDA_DOCKER to false: {e}"
+        )
+        os.environ["USE_CUDA_DOCKER"] = "false"
+        USE_CUDA = "false"
+        DEVICE_TYPE = "cpu"
+else:
+    DEVICE_TYPE = "cpu"
+
 
 ####################################
 # LOGGING
@@ -47,6 +69,9 @@ else:
 log = logging.getLogger(__name__)
 log.info(f"GLOBAL_LOG_LEVEL: {GLOBAL_LOG_LEVEL}")
 
+if "cuda_error" in locals():
+    log.exception(cuda_error)
+
 log_sources = [
     "AUDIO",
     "COMFYUI",
@@ -88,15 +113,9 @@ WEBUI_FAVICON_URL = "https://openwebui.com/favicon.png"
 
 ENV = os.environ.get("ENV", "dev")
 
-PIP_INSTALL = False
-try:
-    importlib.metadata.version("open-webui")
-    PIP_INSTALL = True
-except importlib.metadata.PackageNotFoundError:
-    pass
+FROM_INIT_PY = os.environ.get("FROM_INIT_PY", "False").lower() == "true"
 
-
-if PIP_INSTALL:
+if FROM_INIT_PY:
     PACKAGE_DATA = {"version": importlib.metadata.version("open-webui")}
 else:
     try:
@@ -187,12 +206,12 @@ WEBUI_BUILD_HASH = os.environ.get("WEBUI_BUILD_HASH", "dev-build")
 
 DATA_DIR = Path(os.getenv("DATA_DIR", BACKEND_DIR / "data")).resolve()
 
-if PIP_INSTALL:
+if FROM_INIT_PY:
     NEW_DATA_DIR = Path(os.getenv("DATA_DIR", OPEN_WEBUI_DIR / "data")).resolve()
     NEW_DATA_DIR.mkdir(parents=True, exist_ok=True)
 
     # Check if the data directory exists in the package directory
-    if DATA_DIR.exists():
+    if DATA_DIR.exists() and DATA_DIR != NEW_DATA_DIR:
         log.info(f"Moving {DATA_DIR} to {NEW_DATA_DIR}")
         for item in DATA_DIR.iterdir():
             dest = NEW_DATA_DIR / item.name
@@ -201,11 +220,14 @@ if PIP_INSTALL:
             else:
                 shutil.copy2(item, dest)
 
-    DATA_DIR = OPEN_WEBUI_DIR / "data"
+    DATA_DIR = Path(os.getenv("DATA_DIR", OPEN_WEBUI_DIR / "data"))
 
 
+FONTS_DIR = Path(os.getenv("FONTS_DIR", OPEN_WEBUI_DIR / "static" / "fonts"))
+
 FRONTEND_BUILD_DIR = Path(os.getenv("FRONTEND_BUILD_DIR", BASE_DIR / "build")).resolve()
-if PIP_INSTALL:
+
+if FROM_INIT_PY:
     FRONTEND_BUILD_DIR = Path(
         os.getenv("FRONTEND_BUILD_DIR", OPEN_WEBUI_DIR / "frontend")
     ).resolve()
@@ -276,3 +298,11 @@ WEBUI_SESSION_COOKIE_SECURE = os.environ.get(
 
 if WEBUI_AUTH and WEBUI_SECRET_KEY == "":
     raise ValueError(ERROR_MESSAGES.ENV_VAR_NOT_FOUND)
+
+ENABLE_WEBSOCKET_SUPPORT = (
+    os.environ.get("ENABLE_WEBSOCKET_SUPPORT", "True").lower() == "true"
+)
+
+WEBSOCKET_MANAGER = os.environ.get("WEBSOCKET_MANAGER", "")
+
+WEBSOCKET_REDIS_URL = os.environ.get("WEBSOCKET_REDIS_URL", "redis://localhost:6379/0")

backend/open_webui/main.py

@@ -19,7 +19,9 @@ from open_webui.apps.audio.main import app as audio_app
 from open_webui.apps.images.main import app as images_app
 from open_webui.apps.ollama.main import app as ollama_app
 from open_webui.apps.ollama.main import (
-    generate_openai_chat_completion as generate_ollama_chat_completion,
+    GenerateChatCompletionForm,
+    generate_chat_completion as generate_ollama_chat_completion,
+    generate_openai_chat_completion as generate_ollama_openai_chat_completion,
 )
 from open_webui.apps.ollama.main import get_all_models as get_ollama_models
 from open_webui.apps.openai.main import app as openai_app
@@ -63,8 +65,8 @@ from open_webui.config import (
     MODEL_FILTER_LIST,
     OAUTH_MERGE_ACCOUNTS_BY_EMAIL,
     OAUTH_PROVIDERS,
+    ENABLE_SEARCH_QUERY,
     SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE,
-    SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD,
     STATIC_DIR,
     TASK_MODEL,
     TASK_MODEL_EXTERNAL,
@@ -109,6 +111,7 @@ from starlette.middleware.base import BaseHTTPMiddleware
 from starlette.middleware.sessions import SessionMiddleware
 from starlette.responses import RedirectResponse, Response, StreamingResponse
 
+from open_webui.utils.security_headers import SecurityHeadersMiddleware
 
 from open_webui.utils.misc import (
     add_or_update_system_message,
@@ -134,6 +137,12 @@ from open_webui.utils.utils import (
 )
 from open_webui.utils.webhook import post_webhook
 
+from open_webui.utils.payload import convert_payload_openai_to_ollama
+from open_webui.utils.response import (
+    convert_response_ollama_to_openai,
+    convert_streaming_response_ollama_to_openai,
+)
+
 if SAFE_MODE:
     print("SAFE MODE ENABLED")
     Functions.deactivate_all_functions()
@@ -199,9 +208,7 @@ app.state.config.TITLE_GENERATION_PROMPT_TEMPLATE = TITLE_GENERATION_PROMPT_TEMP
 app.state.config.SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE = (
     SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE
 )
-app.state.config.SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD = (
-    SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD
-)
+app.state.config.ENABLE_SEARCH_QUERY = ENABLE_SEARCH_QUERY
 app.state.config.TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE = (
     TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE
 )
@@ -397,8 +404,13 @@ async def chat_completion_tools_handler(
     specs = [tool["spec"] for tool in tools.values()]
     tools_specs = json.dumps(specs)
 
+    if app.state.config.TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE != "":
+        template = app.state.config.TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE
+    else:
+        template = """Available Tools: {{TOOLS}}\nReturn an empty string if no tools match the query. If a function tool matches, construct and return a JSON object in the format {\"name\": \"functionName\", \"parameters\": {\"requiredFunctionParamKey\": \"requiredFunctionParamValue\"}} using the appropriate tool and its parameters. Only return the object and limit the response to the JSON object without additional text."""
+
     tools_function_calling_prompt = tools_function_calling_generation_template(
-        app.state.config.TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE, tools_specs
+        template, tools_specs
     )
     log.info(f"{tools_function_calling_prompt=}")
     payload = get_tools_function_calling_payload(
@@ -583,8 +595,17 @@ class ChatCompletionMiddleware(BaseHTTPMiddleware):
         if len(contexts) > 0:
             context_string = "/n".join(contexts).strip()
             prompt = get_last_user_message(body["messages"])
+
             if prompt is None:
                 raise Exception("No user message found")
+            if (
+                rag_app.state.config.RELEVANCE_THRESHOLD == 0
+                and context_string.strip() == ""
+            ):
+                log.debug(
+                    f"With a 0 relevancy threshold for RAG, the context cannot be empty"
+                )
+
             # Workaround for Ollama 2.0+ system prompt issue
             # TODO: replace with add_or_update_system_message
             if model["owned_by"] == "ollama":
@@ -777,6 +798,8 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
+app.add_middleware(SecurityHeadersMiddleware)
+
 
 @app.middleware("http")
 async def commit_session_after_request(request: Request, call_next):
@@ -809,6 +832,24 @@ async def update_embedding_function(request: Request, call_next):
     return response
 
 
+@app.middleware("http")
+async def inspect_websocket(request: Request, call_next):
+    if (
+        "/ws/socket.io" in request.url.path
+        and request.query_params.get("transport") == "websocket"
+    ):
+        upgrade = (request.headers.get("Upgrade") or "").lower()
+        connection = (request.headers.get("Connection") or "").lower().split(",")
+        # Check that there's the correct headers for an upgrade, else reject the connection
+        # This is to work around this upstream issue: https://github.com/miguelgrinberg/python-engineio/issues/367
+        if upgrade != "websocket" or "upgrade" not in connection:
+            return JSONResponse(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                content={"detail": "Invalid WebSocket upgrade request"},
+            )
+    return await call_next(request)
+
+
 app.mount("/ws", socket_app)
 
 app.mount("/ollama", ollama_app)
@@ -1015,7 +1056,18 @@ async def generate_chat_completions(form_data: dict, user=Depends(get_verified_u
     if model.get("pipe"):
         return await generate_function_chat_completion(form_data, user=user)
     if model["owned_by"] == "ollama":
-        return await generate_ollama_chat_completion(form_data, user=user)
+        # Using /ollama/api/chat endpoint
+        form_data = convert_payload_openai_to_ollama(form_data)
+        form_data = GenerateChatCompletionForm(**form_data)
+        response = await generate_ollama_chat_completion(form_data=form_data, user=user)
+        if form_data.stream:
+            response.headers["content-type"] = "text/event-stream"
+            return StreamingResponse(
+                convert_streaming_response_ollama_to_openai(response),
+                headers=dict(response.headers),
+            )
+        else:
+            return convert_response_ollama_to_openai(response)
     else:
         return await generate_openai_chat_completion(form_data, user=user)
 
@@ -1312,8 +1364,8 @@ async def get_task_config(user=Depends(get_verified_user)):
         "TASK_MODEL": app.state.config.TASK_MODEL,
         "TASK_MODEL_EXTERNAL": app.state.config.TASK_MODEL_EXTERNAL,
         "TITLE_GENERATION_PROMPT_TEMPLATE": app.state.config.TITLE_GENERATION_PROMPT_TEMPLATE,
+        "ENABLE_SEARCH_QUERY": app.state.config.ENABLE_SEARCH_QUERY,
         "SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE": app.state.config.SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE,
-        "SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD": app.state.config.SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD,
         "TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE": app.state.config.TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE,
     }
 
@@ -1323,7 +1375,7 @@ class TaskConfigForm(BaseModel):
     TASK_MODEL_EXTERNAL: Optional[str]
     TITLE_GENERATION_PROMPT_TEMPLATE: str
     SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE: str
-    SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD: int
+    ENABLE_SEARCH_QUERY: bool
     TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE: str
 
 
@@ -1337,9 +1389,7 @@ async def update_task_config(form_data: TaskConfigForm, user=Depends(get_admin_u
     app.state.config.SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE = (
         form_data.SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE
     )
-    app.state.config.SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD = (
-        form_data.SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD
-    )
+    app.state.config.ENABLE_SEARCH_QUERY = form_data.ENABLE_SEARCH_QUERY
     app.state.config.TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE = (
         form_data.TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE
     )
@@ -1349,7 +1399,7 @@ async def update_task_config(form_data: TaskConfigForm, user=Depends(get_admin_u
         "TASK_MODEL_EXTERNAL": app.state.config.TASK_MODEL_EXTERNAL,
         "TITLE_GENERATION_PROMPT_TEMPLATE": app.state.config.TITLE_GENERATION_PROMPT_TEMPLATE,
         "SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE": app.state.config.SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE,
-        "SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD": app.state.config.SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD,
+        "ENABLE_SEARCH_QUERY": app.state.config.ENABLE_SEARCH_QUERY,
         "TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE": app.state.config.TOOLS_FUNCTION_CALLING_PROMPT_TEMPLATE,
     }
 
@@ -1367,11 +1417,25 @@ async def generate_title(form_data: dict, user=Depends(get_verified_user)):
 
     # Check if the user has a custom task model
     # If the user has a custom task model, use that model
-    model_id = get_task_model_id(model_id)
+    task_model_id = get_task_model_id(model_id)
+    print(task_model_id)
 
-    print(model_id)
+    model = app.state.MODELS[task_model_id]
 
-    template = app.state.config.TITLE_GENERATION_PROMPT_TEMPLATE
+    if app.state.config.TITLE_GENERATION_PROMPT_TEMPLATE != "":
+        template = app.state.config.TITLE_GENERATION_PROMPT_TEMPLATE
+    else:
+        template = """Create a concise, 3-5 word title with an emoji as a title for the prompt in the given language. Suitable Emojis for the summary can be used to enhance understanding but avoid quotation marks or special formatting. RESPOND ONLY WITH THE TITLE TEXT.
+
+Examples of titles:
+📉 Stock Market Trends
+🍪 Perfect Chocolate Chip Recipe
+Evolution of Music Streaming
+Remote Work Productivity Tips
+Artificial Intelligence in Healthcare
+🎮 Video Game Development Insights
+
+Prompt: {{prompt:middletruncate:8000}}"""
 
     content = title_generation_template(
         template,
@@ -1383,16 +1447,22 @@ async def generate_title(form_data: dict, user=Depends(get_verified_user)):
     )
 
     payload = {
-        "model": model_id,
+        "model": task_model_id,
         "messages": [{"role": "user", "content": content}],
         "stream": False,
-        "max_tokens": 50,
+        **(
+            {"max_tokens": 50}
+            if app.state.MODELS[task_model_id]["owned_by"] == "ollama"
+            else {
+                "max_completion_tokens": 50,
+            }
+        ),
         "chat_id": form_data.get("chat_id", None),
         "metadata": {"task": str(TASKS.TITLE_GENERATION)},
     }
-
     log.debug(payload)
 
+    # Handle pipeline filters
     try:
         payload = filter_pipeline(payload, user)
     except Exception as e:
@@ -1406,7 +1476,6 @@ async def generate_title(form_data: dict, user=Depends(get_verified_user)):
                 status_code=status.HTTP_400_BAD_REQUEST,
                 content={"detail": str(e)},
             )
-
     if "chat_id" in payload:
         del payload["chat_id"]
 
@@ -1416,11 +1485,10 @@ async def generate_title(form_data: dict, user=Depends(get_verified_user)):
 @app.post("/api/task/query/completions")
 async def generate_search_query(form_data: dict, user=Depends(get_verified_user)):
     print("generate_search_query")
-
-    if len(form_data["prompt"]) < app.state.config.SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD:
+    if not app.state.config.ENABLE_SEARCH_QUERY:
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
-            detail=f"Skip search query generation for short prompts (< {app.state.config.SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD} characters)",
+            detail=f"Search query generation is disabled",
         )
 
     model_id = form_data["model"]
@@ -1432,26 +1500,46 @@ async def generate_search_query(form_data: dict, user=Depends(get_verified_user)
 
     # Check if the user has a custom task model
     # If the user has a custom task model, use that model
-    model_id = get_task_model_id(model_id)
+    task_model_id = get_task_model_id(model_id)
+    print(task_model_id)
 
-    print(model_id)
+    model = app.state.MODELS[task_model_id]
 
-    template = app.state.config.SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE
+    if app.state.config.SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE != "":
+        template = app.state.config.SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE
+    else:
+        template = """Given the user's message and interaction history, decide if a web search is necessary. You must be concise and exclusively provide a search query if one is necessary. Refrain from verbose responses or any additional commentary. Prefer suggesting a search if uncertain to provide comprehensive or updated information. If a search isn't needed at all, respond with an empty string. Default to a search query when in doubt. Today's date is {{CURRENT_DATE}}.
+
+User Message:
+{{prompt:end:4000}}
+
+Interaction History:
+{{MESSAGES:END:6}}
+
+Search Query:"""
 
     content = search_query_generation_template(
-        template, form_data["prompt"], {"name": user.name}
+        template, form_data["messages"], {"name": user.name}
     )
 
+    print("content", content)
+
     payload = {
-        "model": model_id,
+        "model": task_model_id,
         "messages": [{"role": "user", "content": content}],
         "stream": False,
-        "max_tokens": 30,
+        **(
+            {"max_tokens": 30}
+            if app.state.MODELS[task_model_id]["owned_by"] == "ollama"
+            else {
+                "max_completion_tokens": 30,
+            }
+        ),
         "metadata": {"task": str(TASKS.QUERY_GENERATION)},
     }
+    log.debug(payload)
 
-    print(payload)
-
+    # Handle pipeline filters
     try:
         payload = filter_pipeline(payload, user)
     except Exception as e:
@@ -1465,7 +1553,6 @@ async def generate_search_query(form_data: dict, user=Depends(get_verified_user)
                 status_code=status.HTTP_400_BAD_REQUEST,
                 content={"detail": str(e)},
             )
-
     if "chat_id" in payload:
         del payload["chat_id"]
 
@@ -1485,16 +1572,16 @@ async def generate_emoji(form_data: dict, user=Depends(get_verified_user)):
 
     # Check if the user has a custom task model
     # If the user has a custom task model, use that model
-    model_id = get_task_model_id(model_id)
+    task_model_id = get_task_model_id(model_id)
+    print(task_model_id)
 
-    print(model_id)
+    model = app.state.MODELS[task_model_id]
 
     template = '''
 Your task is to reflect the speaker's likely facial expression through a fitting emoji. Interpret emotions from the message and reflect their facial expression using fitting, diverse emojis (e.g., 😊, 😢, 😡, 😱).
 
 Message: """{{prompt}}"""
 '''
-
     content = title_generation_template(
         template,
         form_data["prompt"],
@@ -1505,16 +1592,22 @@ Message: """{{prompt}}"""
     )
 
     payload = {
-        "model": model_id,
+        "model": task_model_id,
         "messages": [{"role": "user", "content": content}],
         "stream": False,
-        "max_tokens": 4,
+        **(
+            {"max_tokens": 4}
+            if app.state.MODELS[task_model_id]["owned_by"] == "ollama"
+            else {
+                "max_completion_tokens": 4,
+            }
+        ),
         "chat_id": form_data.get("chat_id", None),
         "metadata": {"task": str(TASKS.EMOJI_GENERATION)},
     }
-
     log.debug(payload)
 
+    # Handle pipeline filters
     try:
         payload = filter_pipeline(payload, user)
     except Exception as e:
@@ -1528,7 +1621,6 @@ Message: """{{prompt}}"""
                 status_code=status.HTTP_400_BAD_REQUEST,
                 content={"detail": str(e)},
             )
-
     if "chat_id" in payload:
         del payload["chat_id"]
 
@@ -1548,8 +1640,10 @@ async def generate_moa_response(form_data: dict, user=Depends(get_verified_user)
 
     # Check if the user has a custom task model
     # If the user has a custom task model, use that model
-    model_id = get_task_model_id(model_id)
-    print(model_id)
+    task_model_id = get_task_model_id(model_id)
+    print(task_model_id)
+
+    model = app.state.MODELS[task_model_id]
 
     template = """You have been provided with a set of responses from various models to the latest user query: "{{prompt}}"
 
@@ -1564,13 +1658,12 @@ Responses from models: {{responses}}"""
     )
 
     payload = {
-        "model": model_id,
+        "model": task_model_id,
         "messages": [{"role": "user", "content": content}],
         "stream": form_data.get("stream", False),
         "chat_id": form_data.get("chat_id", None),
         "metadata": {"task": str(TASKS.MOA_RESPONSE_GENERATION)},
     }
-
     log.debug(payload)
 
     try:
@@ -1586,7 +1679,6 @@ Responses from models: {{responses}}"""
                 status_code=status.HTTP_400_BAD_REQUEST,
                 content={"detail": str(e)},
             )
-
     if "chat_id" in payload:
         del payload["chat_id"]
 

backend/open_webui/migrations/scripts/revision.py

@@ -0,0 +1,19 @@
+from alembic import command
+from alembic.config import Config
+
+from open_webui.env import OPEN_WEBUI_DIR
+
+alembic_cfg = Config(OPEN_WEBUI_DIR / "alembic.ini")
+
+# Set the script location dynamically
+migrations_path = OPEN_WEBUI_DIR / "migrations"
+alembic_cfg.set_main_option("script_location", str(migrations_path))
+
+
+def revision(message: str) -> None:
+    command.revision(alembic_cfg, message=message, autogenerate=False)
+
+
+if __name__ == "__main__":
+    input_message = input("Enter the revision message: ")
+    revision(input_message)

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
 from open_webui.migrations.util import get_existing_tables
 
 # revision identifiers, used by Alembic.
@@ -82,7 +82,7 @@ def upgrade() -> None:
             sa.Column("id", sa.String(), nullable=False),
             sa.Column("user_id", sa.String(), nullable=True),
             sa.Column("filename", sa.Text(), nullable=True),
-            sa.Column("meta", apps.webui.internal.db.JSONField(), nullable=True),
+            sa.Column("meta", JSONField(), nullable=True),
             sa.Column("created_at", sa.BigInteger(), nullable=True),
             sa.PrimaryKeyConstraint("id"),
         )
@@ -95,8 +95,8 @@ def upgrade() -> None:
             sa.Column("name", sa.Text(), nullable=True),
             sa.Column("type", sa.Text(), nullable=True),
             sa.Column("content", sa.Text(), nullable=True),
-            sa.Column("meta", apps.webui.internal.db.JSONField(), nullable=True),
-            sa.Column("valves", apps.webui.internal.db.JSONField(), nullable=True),
+            sa.Column("meta", JSONField(), nullable=True),
+            sa.Column("valves", JSONField(), nullable=True),
             sa.Column("is_active", sa.Boolean(), nullable=True),
             sa.Column("is_global", sa.Boolean(), nullable=True),
             sa.Column("updated_at", sa.BigInteger(), nullable=True),
@@ -122,8 +122,8 @@ def upgrade() -> None:
             sa.Column("user_id", sa.Text(), nullable=True),
             sa.Column("base_model_id", sa.Text(), nullable=True),
             sa.Column("name", sa.Text(), nullable=True),
-            sa.Column("params", apps.webui.internal.db.JSONField(), nullable=True),
-            sa.Column("meta", apps.webui.internal.db.JSONField(), nullable=True),
+            sa.Column("params", JSONField(), nullable=True),
+            sa.Column("meta", JSONField(), nullable=True),
             sa.Column("updated_at", sa.BigInteger(), nullable=True),
             sa.Column("created_at", sa.BigInteger(), nullable=True),
             sa.PrimaryKeyConstraint("id"),
@@ -157,9 +157,9 @@ def upgrade() -> None:
             sa.Column("user_id", sa.String(), nullable=True),
             sa.Column("name", sa.Text(), nullable=True),
             sa.Column("content", sa.Text(), nullable=True),
-            sa.Column("specs", apps.webui.internal.db.JSONField(), nullable=True),
-            sa.Column("meta", apps.webui.internal.db.JSONField(), nullable=True),
-            sa.Column("valves", apps.webui.internal.db.JSONField(), nullable=True),
+            sa.Column("specs", JSONField(), nullable=True),
+            sa.Column("meta", JSONField(), nullable=True),
+            sa.Column("valves", JSONField(), nullable=True),
             sa.Column("updated_at", sa.BigInteger(), nullable=True),
             sa.Column("created_at", sa.BigInteger(), nullable=True),
             sa.PrimaryKeyConstraint("id"),
@@ -177,8 +177,8 @@ def upgrade() -> None:
             sa.Column("updated_at", sa.BigInteger(), nullable=True),
             sa.Column("created_at", sa.BigInteger(), nullable=True),
             sa.Column("api_key", sa.String(), nullable=True),
-            sa.Column("settings", apps.webui.internal.db.JSONField(), nullable=True),
-            sa.Column("info", apps.webui.internal.db.JSONField(), nullable=True),
+            sa.Column("settings", JSONField(), nullable=True),
+            sa.Column("info", JSONField(), nullable=True),
             sa.Column("oauth_sub", sa.Text(), nullable=True),
             sa.PrimaryKeyConstraint("id"),
             sa.UniqueConstraint("api_key"),

backend/open_webui/utils/misc.py

@@ -6,7 +6,14 @@ from datetime import timedelta
 from pathlib import Path
 from typing import Callable, Optional
 
-from open_webui.utils.task import prompt_template
+
+def get_messages_content(messages: list[dict]) -> str:
+    return "\n".join(
+        [
+            f"{message['role'].upper()}: {get_content_from_message(message)}"
+            for message in messages
+        ]
+    )
 
 
 def get_last_user_message_item(messages: list[dict]) -> Optional[dict]:
@@ -30,7 +37,6 @@ def get_last_user_message(messages: list[dict]) -> Optional[str]:
     message = get_last_user_message_item(messages)
     if message is None:
         return None
-
     return get_content_from_message(message)
 
 
@@ -99,103 +105,29 @@ def openai_chat_message_template(model: str):
     }
 
 
-def openai_chat_chunk_message_template(model: str, message: str) -> dict:
+def openai_chat_chunk_message_template(
+    model: str, message: Optional[str] = None
+) -> dict:
     template = openai_chat_message_template(model)
     template["object"] = "chat.completion.chunk"
-    template["choices"][0]["delta"] = {"content": message}
+    if message:
+        template["choices"][0]["delta"] = {"content": message}
+    else:
+        template["choices"][0]["finish_reason"] = "stop"
     return template
 
 
-def openai_chat_completion_message_template(model: str, message: str) -> dict:
+def openai_chat_completion_message_template(
+    model: str, message: Optional[str] = None
+) -> dict:
     template = openai_chat_message_template(model)
     template["object"] = "chat.completion"
-    template["choices"][0]["message"] = {"content": message, "role": "assistant"}
+    if message:
+        template["choices"][0]["message"] = {"content": message, "role": "assistant"}
     template["choices"][0]["finish_reason"] = "stop"
     return template
 
 
-# inplace function: form_data is modified
-def apply_model_system_prompt_to_body(params: dict, form_data: dict, user) -> dict:
-    system = params.get("system", None)
-    if not system:
-        return form_data
-
-    if user:
-        template_params = {
-            "user_name": user.name,
-            "user_location": user.info.get("location") if user.info else None,
-        }
-    else:
-        template_params = {}
-    system = prompt_template(system, **template_params)
-    form_data["messages"] = add_or_update_system_message(
-        system, form_data.get("messages", [])
-    )
-    return form_data
-
-
-# inplace function: form_data is modified
-def apply_model_params_to_body(
-    params: dict, form_data: dict, mappings: dict[str, Callable]
-) -> dict:
-    if not params:
-        return form_data
-
-    for key, cast_func in mappings.items():
-        if (value := params.get(key)) is not None:
-            form_data[key] = cast_func(value)
-
-    return form_data
-
-
-# inplace function: form_data is modified
-def apply_model_params_to_body_openai(params: dict, form_data: dict) -> dict:
-    mappings = {
-        "temperature": float,
-        "top_p": int,
-        "max_tokens": int,
-        "frequency_penalty": int,
-        "seed": lambda x: x,
-        "stop": lambda x: [bytes(s, "utf-8").decode("unicode_escape") for s in x],
-    }
-    return apply_model_params_to_body(params, form_data, mappings)
-
-
-def apply_model_params_to_body_ollama(params: dict, form_data: dict) -> dict:
-    opts = [
-        "temperature",
-        "top_p",
-        "seed",
-        "mirostat",
-        "mirostat_eta",
-        "mirostat_tau",
-        "num_ctx",
-        "num_batch",
-        "num_keep",
-        "repeat_last_n",
-        "tfs_z",
-        "top_k",
-        "min_p",
-        "use_mmap",
-        "use_mlock",
-        "num_thread",
-        "num_gpu",
-    ]
-    mappings = {i: lambda x: x for i in opts}
-    form_data = apply_model_params_to_body(params, form_data, mappings)
-
-    name_differences = {
-        "max_tokens": "num_predict",
-        "frequency_penalty": "repeat_penalty",
-    }
-
-    for key, value in name_differences.items():
-        if (param := params.get(key, None)) is not None:
-            form_data[value] = param
-
-    return form_data
-
-
 def get_gravatar_url(email):
     # Trim leading and trailing whitespace from
     # an email address and force all characters

backend/open_webui/utils/payload.py

@@ -0,0 +1,134 @@
+from open_webui.utils.task import prompt_template
+from open_webui.utils.misc import (
+    add_or_update_system_message,
+)
+
+from typing import Callable, Optional
+
+
+# inplace function: form_data is modified
+def apply_model_system_prompt_to_body(params: dict, form_data: dict, user) -> dict:
+    system = params.get("system", None)
+    if not system:
+        return form_data
+
+    if user:
+        template_params = {
+            "user_name": user.name,
+            "user_location": user.info.get("location") if user.info else None,
+        }
+    else:
+        template_params = {}
+    system = prompt_template(system, **template_params)
+    form_data["messages"] = add_or_update_system_message(
+        system, form_data.get("messages", [])
+    )
+    return form_data
+
+
+# inplace function: form_data is modified
+def apply_model_params_to_body(
+    params: dict, form_data: dict, mappings: dict[str, Callable]
+) -> dict:
+    if not params:
+        return form_data
+
+    for key, cast_func in mappings.items():
+        if (value := params.get(key)) is not None:
+            form_data[key] = cast_func(value)
+
+    return form_data
+
+
+# inplace function: form_data is modified
+def apply_model_params_to_body_openai(params: dict, form_data: dict) -> dict:
+    mappings = {
+        "temperature": float,
+        "top_p": float,
+        "max_tokens": int,
+        "frequency_penalty": float,
+        "seed": lambda x: x,
+        "stop": lambda x: [bytes(s, "utf-8").decode("unicode_escape") for s in x],
+    }
+    return apply_model_params_to_body(params, form_data, mappings)
+
+
+def apply_model_params_to_body_ollama(params: dict, form_data: dict) -> dict:
+    opts = [
+        "temperature",
+        "top_p",
+        "seed",
+        "mirostat",
+        "mirostat_eta",
+        "mirostat_tau",
+        "num_ctx",
+        "num_batch",
+        "num_keep",
+        "repeat_last_n",
+        "tfs_z",
+        "top_k",
+        "min_p",
+        "use_mmap",
+        "use_mlock",
+        "num_thread",
+        "num_gpu",
+    ]
+    mappings = {i: lambda x: x for i in opts}
+    form_data = apply_model_params_to_body(params, form_data, mappings)
+
+    name_differences = {
+        "max_tokens": "num_predict",
+        "frequency_penalty": "repeat_penalty",
+    }
+
+    for key, value in name_differences.items():
+        if (param := params.get(key, None)) is not None:
+            form_data[value] = param
+
+    return form_data
+
+
+def convert_payload_openai_to_ollama(openai_payload: dict) -> dict:
+    """
+    Converts a payload formatted for OpenAI's API to be compatible with Ollama's API endpoint for chat completions.
+
+    Args:
+        openai_payload (dict): The payload originally designed for OpenAI API usage.
+
+    Returns:
+        dict: A modified payload compatible with the Ollama API.
+    """
+    ollama_payload = {}
+
+    # Mapping basic model and message details
+    ollama_payload["model"] = openai_payload.get("model")
+    ollama_payload["messages"] = openai_payload.get("messages")
+    ollama_payload["stream"] = openai_payload.get("stream", False)
+
+    # If there are advanced parameters in the payload, format them in Ollama's options field
+    ollama_options = {}
+
+    # Handle parameters which map directly
+    for param in ["temperature", "top_p", "seed"]:
+        if param in openai_payload:
+            ollama_options[param] = openai_payload[param]
+
+    # Mapping OpenAI's `max_tokens` -> Ollama's `num_predict`
+    if "max_completion_tokens" in openai_payload:
+        ollama_options["num_predict"] = openai_payload["max_completion_tokens"]
+    elif "max_tokens" in openai_payload:
+        ollama_options["num_predict"] = openai_payload["max_tokens"]
+
+    # Handle frequency / presence_penalty, which needs renaming and checking
+    if "frequency_penalty" in openai_payload:
+        ollama_options["repeat_penalty"] = openai_payload["frequency_penalty"]
+
+    if "presence_penalty" in openai_payload and "penalty" not in ollama_options:
+        # We are assuming presence penalty uses a similar concept in Ollama, which needs custom handling if exists.
+        ollama_options["new_topic_penalty"] = openai_payload["presence_penalty"]
+
+    # Add options to payload if any have been set
+    if ollama_options:
+        ollama_payload["options"] = ollama_options
+
+    return ollama_payload

backend/open_webui/utils/response.py

@@ -0,0 +1,32 @@
+import json
+from open_webui.utils.misc import (
+    openai_chat_chunk_message_template,
+    openai_chat_completion_message_template,
+)
+
+
+def convert_response_ollama_to_openai(ollama_response: dict) -> dict:
+    model = ollama_response.get("model", "ollama")
+    message_content = ollama_response.get("message", {}).get("content", "")
+
+    response = openai_chat_completion_message_template(model, message_content)
+    return response
+
+
+async def convert_streaming_response_ollama_to_openai(ollama_streaming_response):
+    async for data in ollama_streaming_response.body_iterator:
+        data = json.loads(data)
+
+        model = data.get("model", "ollama")
+        message_content = data.get("message", {}).get("content", "")
+        done = data.get("done", False)
+
+        data = openai_chat_chunk_message_template(
+            model, message_content if not done else None
+        )
+
+        line = f"data: {json.dumps(data)}\n\n"
+        if done:
+            line += "data: [DONE]\n\n"
+
+        yield line

backend/open_webui/utils/security_headers.py

@@ -0,0 +1,115 @@
+import re
+import os
+
+from fastapi import Request
+from starlette.middleware.base import BaseHTTPMiddleware
+from typing import Dict
+
+
+class SecurityHeadersMiddleware(BaseHTTPMiddleware):
+    async def dispatch(self, request: Request, call_next):
+        response = await call_next(request)
+        response.headers.update(set_security_headers())
+        return response
+
+
+def set_security_headers() -> Dict[str, str]:
+    """
+    Sets security headers based on environment variables.
+
+    This function reads specific environment variables and uses their values
+    to set corresponding security headers. The headers that can be set are:
+    - cache-control
+    - strict-transport-security
+    - referrer-policy
+    - x-content-type-options
+    - x-download-options
+    - x-frame-options
+    - x-permitted-cross-domain-policies
+
+    Each environment variable is associated with a specific setter function
+    that constructs the header. If the environment variable is set, the
+    corresponding header is added to the options dictionary.
+
+    Returns:
+        dict: A dictionary containing the security headers and their values.
+    """
+    options = {}
+    header_setters = {
+        "CACHE_CONTROL": set_cache_control,
+        "HSTS": set_hsts,
+        "REFERRER_POLICY": set_referrer,
+        "XCONTENT_TYPE": set_xcontent_type,
+        "XDOWNLOAD_OPTIONS": set_xdownload_options,
+        "XFRAME_OPTIONS": set_xframe,
+        "XPERMITTED_CROSS_DOMAIN_POLICIES": set_xpermitted_cross_domain_policies,
+    }
+
+    for env_var, setter in header_setters.items():
+        value = os.environ.get(env_var, None)
+        if value:
+            header = setter(value)
+            if header:
+                options.update(header)
+
+    return options
+
+
+# Set HTTP Strict Transport Security(HSTS) response header
+def set_hsts(value: str):
+    pattern = r"^max-age=(\d+)(;includeSubDomains)?(;preload)?$"
+    match = re.match(pattern, value, re.IGNORECASE)
+    if not match:
+        return "max-age=31536000;includeSubDomains"
+    return {"Strict-Transport-Security": value}
+
+
+# Set X-Frame-Options response header
+def set_xframe(value: str):
+    pattern = r"^(DENY|SAMEORIGIN)$"
+    match = re.match(pattern, value, re.IGNORECASE)
+    if not match:
+        value = "DENY"
+    return {"X-Frame-Options": value}
+
+
+# Set Referrer-Policy response header
+def set_referrer(value: str):
+    pattern = r"^(no-referrer|no-referrer-when-downgrade|origin|origin-when-cross-origin|same-origin|strict-origin|strict-origin-when-cross-origin|unsafe-url)$"
+    match = re.match(pattern, value, re.IGNORECASE)
+    if not match:
+        value = "no-referrer"
+    return {"Referrer-Policy": value}
+
+
+# Set Cache-Control response header
+def set_cache_control(value: str):
+    pattern = r"^(public|private|no-cache|no-store|must-revalidate|proxy-revalidate|max-age=\d+|s-maxage=\d+|no-transform|immutable)(,\s*(public|private|no-cache|no-store|must-revalidate|proxy-revalidate|max-age=\d+|s-maxage=\d+|no-transform|immutable))*$"
+    match = re.match(pattern, value, re.IGNORECASE)
+    if not match:
+        value = "no-store, max-age=0"
+
+    return {"Cache-Control": value}
+
+
+# Set X-Download-Options response header
+def set_xdownload_options(value: str):
+    if value != "noopen":
+        value = "noopen"
+    return {"X-Download-Options": value}
+
+
+# Set X-Content-Type-Options response header
+def set_xcontent_type(value: str):
+    if value != "nosniff":
+        value = "nosniff"
+    return {"X-Content-Type-Options": value}
+
+
+# Set X-Permitted-Cross-Domain-Policies response header
+def set_xpermitted_cross_domain_policies(value: str):
+    pattern = r"^(none|master-only|by-content-type|by-ftp-filename)$"
+    match = re.match(pattern, value, re.IGNORECASE)
+    if not match:
+        value = "none"
+    return {"X-Permitted-Cross-Domain-Policies": value}

backend/open_webui/utils/task.py

@@ -4,6 +4,9 @@ from datetime import datetime
 from typing import Optional
 
 
+from open_webui.utils.misc import get_last_user_message, get_messages_content
+
+
 def prompt_template(
     template: str, user_name: Optional[str] = None, user_location: Optional[str] = None
 ) -> str:
@@ -37,9 +40,7 @@ def prompt_template(
     return template
 
 
-def title_generation_template(
-    template: str, prompt: str, user: Optional[dict] = None
-) -> str:
+def replace_prompt_variable(template: str, prompt: str) -> str:
     def replacement_function(match):
         full_match = match.group(0)
         start_length = match.group(1)
@@ -66,7 +67,13 @@ def title_generation_template(
         replacement_function,
         template,
     )
+    return template
 
+
+def title_generation_template(
+    template: str, prompt: str, user: Optional[dict] = None
+) -> str:
+    template = replace_prompt_variable(template, prompt)
     template = prompt_template(
         template,
         **(
@@ -79,36 +86,50 @@ def title_generation_template(
     return template
 
 
-def search_query_generation_template(
-    template: str, prompt: str, user: Optional[dict] = None
-) -> str:
+def replace_messages_variable(template: str, messages: list[str]) -> str:
     def replacement_function(match):
         full_match = match.group(0)
         start_length = match.group(1)
         end_length = match.group(2)
         middle_length = match.group(3)
 
-        if full_match == "{{prompt}}":
-            return prompt
+        # Process messages based on the number of messages required
+        if full_match == "{{MESSAGES}}":
+            return get_messages_content(messages)
         elif start_length is not None:
-            return prompt[: int(start_length)]
+            return get_messages_content(messages[: int(start_length)])
         elif end_length is not None:
-            return prompt[-int(end_length) :]
+            return get_messages_content(messages[-int(end_length) :])
         elif middle_length is not None:
-            middle_length = int(middle_length)
-            if len(prompt) <= middle_length:
-                return prompt
-            start = prompt[: math.ceil(middle_length / 2)]
-            end = prompt[-math.floor(middle_length / 2) :]
-            return f"{start}...{end}"
+            mid = int(middle_length)
+
+            if len(messages) <= mid:
+                return get_messages_content(messages)
+            # Handle middle truncation: split to get start and end portions of the messages list
+            half = mid // 2
+            start_msgs = messages[:half]
+            end_msgs = messages[-half:] if mid % 2 == 0 else messages[-(half + 1) :]
+            formatted_start = get_messages_content(start_msgs)
+            formatted_end = get_messages_content(end_msgs)
+            return f"{formatted_start}\n{formatted_end}"
         return ""
 
     template = re.sub(
-        r"{{prompt}}|{{prompt:start:(\d+)}}|{{prompt:end:(\d+)}}|{{prompt:middletruncate:(\d+)}}",
+        r"{{MESSAGES}}|{{MESSAGES:START:(\d+)}}|{{MESSAGES:END:(\d+)}}|{{MESSAGES:MIDDLETRUNCATE:(\d+)}}",
         replacement_function,
         template,
     )
 
+    return template
+
+
+def search_query_generation_template(
+    template: str, messages: list[dict], user: Optional[dict] = None
+) -> str:
+    prompt = get_last_user_message(messages)
+    template = replace_prompt_variable(template, prompt)
+    template = replace_messages_variable(template, messages)
+
     template = prompt_template(
         template,
         **(

backend/requirements.txt

@@ -40,7 +40,12 @@ langchain-chroma==0.1.2
 
 fake-useragent==1.5.1
 chromadb==0.5.5
+pymilvus==2.4.6
+
 sentence-transformers==3.0.1
+colbert-ai==0.2.21
+einops==0.8.0
+
 pypdf==4.3.1
 docx2txt==0.8
 python-pptx==1.0.0
@@ -64,7 +69,7 @@ rank-bm25==0.2.2
 faster-whisper==1.0.3
 
 PyJWT[crypto]==2.9.0
-authlib==1.3.2
+authlib==1.3.1
 
 black==24.8.0
 langfuse==2.44.0

backend/start_windows.bat

@@ -30,4 +30,4 @@ IF "%WEBUI_SECRET_KEY%%WEBUI_JWT_SECRET_KEY%" == " " (
 
 :: Execute uvicorn
 SET "WEBUI_SECRET_KEY=%WEBUI_SECRET_KEY%"
-uvicorn main:app --host "%HOST%" --port "%PORT%" --forwarded-allow-ips '*'
+uvicorn open_webui.main:app --host "%HOST%" --port "%PORT%" --forwarded-allow-ips '*'

kubernetes/manifest/base/kustomization.yaml

@@ -0,0 +1,8 @@
+resources:
+  - open-webui.yaml
+  - ollama-service.yaml
+  - ollama-statefulset.yaml
+  - webui-deployment.yaml
+  - webui-service.yaml
+  - webui-ingress.yaml
+  - webui-pvc.yaml

kubernetes/manifest/gpu/kustomization.yaml

@@ -0,0 +1,8 @@
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+
+resources:
+  - ../base
+
+patches:
+- path: ollama-statefulset-gpu.yaml

kubernetes/manifest/kustomization.yaml

@@ -1,13 +0,0 @@
-resources:
-- base/open-webui.yaml
-- base/ollama-service.yaml
-- base/ollama-statefulset.yaml
-- base/webui-deployment.yaml
-- base/webui-service.yaml
-- base/webui-ingress.yaml
-- base/webui-pvc.yaml
-
-apiVersion: kustomize.config.k8s.io/v1beta1
-kind: Kustomization
-patches:
-- path: patches/ollama-statefulset-gpu.yaml

package-lock.json

@@ -1,18 +1,19 @@
 {
 	"name": "open-webui",
-	"version": "0.3.17",
+	"version": "0.3.23",
 	"lockfileVersion": 3,
 	"requires": true,
 	"packages": {
 		"": {
 			"name": "open-webui",
-			"version": "0.3.17",
+			"version": "0.3.23",
 			"dependencies": {
 				"@codemirror/lang-javascript": "^6.2.2",
 				"@codemirror/lang-python": "^6.1.6",
 				"@codemirror/theme-one-dark": "^6.1.2",
 				"@pyscript/core": "^0.4.32",
 				"@sveltejs/adapter-node": "^2.0.0",
+				"@xyflow/svelte": "^0.1.19",
 				"async": "^3.2.5",
 				"bits-ui": "^0.19.7",
 				"codemirror": "^6.0.1",
@@ -31,6 +32,7 @@
 				"katex": "^0.16.9",
 				"marked": "^9.1.0",
 				"mermaid": "^10.9.1",
+				"paneforge": "^0.0.6",
 				"pyodide": "^0.26.1",
 				"socket.io-client": "^4.2.0",
 				"sortablejs": "^1.15.2",
@@ -45,7 +47,6 @@
 				"@sveltejs/kit": "^2.5.20",
 				"@sveltejs/vite-plugin-svelte": "^3.1.1",
 				"@tailwindcss/typography": "^0.5.13",
-				"@types/bun": "latest",
 				"@typescript-eslint/eslint-plugin": "^6.17.0",
 				"@typescript-eslint/parser": "^6.17.0",
 				"autoprefixer": "^10.4.16",
@@ -1368,6 +1369,14 @@
 			"resolved": "https://registry.npmjs.org/@socket.io/component-emitter/-/component-emitter-3.1.2.tgz",
 			"integrity": "sha512-9BCxFwvbGg/RsZK9tjXd8s4UcwR0MWeFQ1XEKIQVVvAGJyINdrqKMcTRyLoK8Rse1GjzLV9cwjWV1olXRWEXVA=="
 		},
+		"node_modules/@svelte-put/shortcut": {
+			"version": "3.1.1",
+			"resolved": "https://registry.npmjs.org/@svelte-put/shortcut/-/shortcut-3.1.1.tgz",
+			"integrity": "sha512-2L5EYTZXiaKvbEelVkg5znxqvfZGZai3m97+cAiUBhLZwXnGtviTDpHxOoZBsqz41szlfRMcamW/8o0+fbW3ZQ==",
+			"peerDependencies": {
+				"svelte": "^3.55.0 || ^4.0.0 || ^5.0.0"
+			}
+		},
 		"node_modules/@sveltejs/adapter-auto": {
 			"version": "3.2.2",
 			"resolved": "https://registry.npmjs.org/@sveltejs/adapter-auto/-/adapter-auto-3.2.2.tgz",
@@ -1494,20 +1503,32 @@
 				"tailwindcss": ">=3.0.0 || insiders"
 			}
 		},
-		"node_modules/@types/bun": {
-			"version": "1.0.10",
-			"resolved": "https://registry.npmjs.org/@types/bun/-/bun-1.0.10.tgz",
-			"integrity": "sha512-Jaz6YYAdm1u3NVlgSyEK+qGmrlLQ20sbWeEoXD64b9w6z/YKYNWlfaphu+xF2Kiy5Tpykm5Q9jIquLegwXx4ng==",
-			"dev": true,
-			"dependencies": {
-				"bun-types": "1.0.33"
-			}
-		},
 		"node_modules/@types/cookie": {
 			"version": "0.6.0",
 			"resolved": "https://registry.npmjs.org/@types/cookie/-/cookie-0.6.0.tgz",
 			"integrity": "sha512-4Kh9a6B2bQciAhf7FSuMRRkUWecJgJu9nPnx3yzpsfXX/c50REIqpHY4C82bXP90qrLtXtkDxTZosYO3UpOwlA=="
 		},
+		"node_modules/@types/d3-color": {
+			"version": "3.1.3",
+			"resolved": "https://registry.npmjs.org/@types/d3-color/-/d3-color-3.1.3.tgz",
+			"integrity": "sha512-iO90scth9WAbmgv7ogoq57O9YpKmFBbmoEoCHDB2xMBY0+/KVrqAaCDyCE16dUspeOvIxFFRI+0sEtqDqy2b4A=="
+		},
+		"node_modules/@types/d3-drag": {
+			"version": "3.0.7",
+			"resolved": "https://registry.npmjs.org/@types/d3-drag/-/d3-drag-3.0.7.tgz",
+			"integrity": "sha512-HE3jVKlzU9AaMazNufooRJ5ZpWmLIoc90A37WU2JMmeq28w1FQqCZswHZ3xR+SuxYftzHq6WU6KJHvqxKzTxxQ==",
+			"dependencies": {
+				"@types/d3-selection": "*"
+			}
+		},
+		"node_modules/@types/d3-interpolate": {
+			"version": "3.0.4",
+			"resolved": "https://registry.npmjs.org/@types/d3-interpolate/-/d3-interpolate-3.0.4.tgz",
+			"integrity": "sha512-mgLPETlrpVV1YRJIglr4Ez47g7Yxjl1lj7YKsiMCb27VJH9W8NVM6Bb9d8kkpG/uAQS5AmbA48q2IAolKKo1MA==",
+			"dependencies": {
+				"@types/d3-color": "*"
+			}
+		},
 		"node_modules/@types/d3-scale": {
 			"version": "4.0.8",
 			"resolved": "https://registry.npmjs.org/@types/d3-scale/-/d3-scale-4.0.8.tgz",
@@ -1521,11 +1542,33 @@
 			"resolved": "https://registry.npmjs.org/@types/d3-scale-chromatic/-/d3-scale-chromatic-3.0.3.tgz",
 			"integrity": "sha512-laXM4+1o5ImZv3RpFAsTRn3TEkzqkytiOY0Dz0sq5cnd1dtNlk6sHLon4OvqaiJb28T0S/TdsBI3Sjsy+keJrw=="
 		},
+		"node_modules/@types/d3-selection": {
+			"version": "3.0.10",
+			"resolved": "https://registry.npmjs.org/@types/d3-selection/-/d3-selection-3.0.10.tgz",
+			"integrity": "sha512-cuHoUgS/V3hLdjJOLTT691+G2QoqAjCVLmr4kJXR4ha56w1Zdu8UUQ5TxLRqudgNjwXeQxKMq4j+lyf9sWuslg=="
+		},
 		"node_modules/@types/d3-time": {
 			"version": "3.0.3",
 			"resolved": "https://registry.npmjs.org/@types/d3-time/-/d3-time-3.0.3.tgz",
 			"integrity": "sha512-2p6olUZ4w3s+07q3Tm2dbiMZy5pCDfYwtLXXHUnVzXgQlZ/OyPtUz6OL382BkOuGlLXqfT+wqv8Fw2v8/0geBw=="
 		},
+		"node_modules/@types/d3-transition": {
+			"version": "3.0.8",
+			"resolved": "https://registry.npmjs.org/@types/d3-transition/-/d3-transition-3.0.8.tgz",
+			"integrity": "sha512-ew63aJfQ/ms7QQ4X7pk5NxQ9fZH/z+i24ZfJ6tJSfqxJMrYLiK01EAs2/Rtw/JreGUsS3pLPNV644qXFGnoZNQ==",
+			"dependencies": {
+				"@types/d3-selection": "*"
+			}
+		},
+		"node_modules/@types/d3-zoom": {
+			"version": "3.0.8",
+			"resolved": "https://registry.npmjs.org/@types/d3-zoom/-/d3-zoom-3.0.8.tgz",
+			"integrity": "sha512-iqMC4/YlFCSlO8+2Ii1GGGliCAY4XdeG748w5vQUbevlbDu0zSjH/+jojorQVBK/se0j6DUFNPBGSqD3YWYnDw==",
+			"dependencies": {
+				"@types/d3-interpolate": "*",
+				"@types/d3-selection": "*"
+			}
+		},
 		"node_modules/@types/debug": {
 			"version": "4.1.12",
 			"resolved": "https://registry.npmjs.org/@types/debug/-/debug-4.1.12.tgz",
@@ -1568,7 +1611,7 @@
 			"version": "20.11.30",
 			"resolved": "https://registry.npmjs.org/@types/node/-/node-20.11.30.tgz",
 			"integrity": "sha512-dHM6ZxwlmuZaRmUPfv1p+KrdD1Dci04FbdEm/9wEMouFqxYoFl5aMkt0VMAUtYRQDyYvD41WJLukhq/ha3YuTw==",
-			"devOptional": true,
+			"optional": true,
 			"dependencies": {
 				"undici-types": "~5.26.4"
 			}
@@ -1613,15 +1656,6 @@
 			"resolved": "https://registry.npmjs.org/@types/unist/-/unist-2.0.10.tgz",
 			"integrity": "sha512-IfYcSBWE3hLpBg8+X2SEa8LVkJdJEkT2Ese2aaLs3ptGdVtABxndrMaxuFlQ1qdFf9Q5rDvDpxI3WwgvKFAsQA=="
 		},
-		"node_modules/@types/ws": {
-			"version": "8.5.10",
-			"resolved": "https://registry.npmjs.org/@types/ws/-/ws-8.5.10.tgz",
-			"integrity": "sha512-vmQSUcfalpIq0R9q7uTo2lXs6eGIpt9wtnLdMv9LVpIjCA/+ufZRozlVoVelIYixx1ugCBKDhn89vnsEGOCx9A==",
-			"dev": true,
-			"dependencies": {
-				"@types/node": "*"
-			}
-		},
 		"node_modules/@types/yauzl": {
 			"version": "2.10.3",
 			"resolved": "https://registry.npmjs.org/@types/yauzl/-/yauzl-2.10.3.tgz",
@@ -1942,6 +1976,33 @@
 			"resolved": "https://registry.npmjs.org/@webreflection/fetch/-/fetch-0.1.5.tgz",
 			"integrity": "sha512-zCcqCJoNLvdeF41asAK71XPlwSPieeRDsE09albBunJEksuYPYNillKNQjf8p5BqSoTKTuKrW3lUm3MNodUC4g=="
 		},
+		"node_modules/@xyflow/svelte": {
+			"version": "0.1.19",
+			"resolved": "https://registry.npmjs.org/@xyflow/svelte/-/svelte-0.1.19.tgz",
+			"integrity": "sha512-yW5w5aI+Yqkob4kLQpVDo/ZmX+E9Pw7459kqwLfv4YG4N1NYXrsDRh9cyph/rapbuDnPi6zqK5E8LKrgaCQC0w==",
+			"dependencies": {
+				"@svelte-put/shortcut": "^3.1.0",
+				"@xyflow/system": "0.0.42",
+				"classcat": "^5.0.4"
+			},
+			"peerDependencies": {
+				"svelte": "^3.0.0 || ^4.0.0"
+			}
+		},
+		"node_modules/@xyflow/system": {
+			"version": "0.0.42",
+			"resolved": "https://registry.npmjs.org/@xyflow/system/-/system-0.0.42.tgz",
+			"integrity": "sha512-kWYj+Y0GOct0jKYTdyRMNOLPxGNbb2TYvPg2gTmJnZ31DOOMkL5uRBLX825DR2gOACDu+i5FHLxPJUPf/eGOJw==",
+			"dependencies": {
+				"@types/d3-drag": "^3.0.7",
+				"@types/d3-selection": "^3.0.10",
+				"@types/d3-transition": "^3.0.8",
+				"@types/d3-zoom": "^3.0.8",
+				"d3-drag": "^3.0.0",
+				"d3-selection": "^3.0.0",
+				"d3-zoom": "^3.0.0"
+			}
+		},
 		"node_modules/acorn": {
 			"version": "8.11.3",
 			"resolved": "https://registry.npmjs.org/acorn/-/acorn-8.11.3.tgz",
@@ -2533,16 +2594,6 @@
 				"url": "https://github.com/sponsors/sindresorhus"
 			}
 		},
-		"node_modules/bun-types": {
-			"version": "1.0.33",
-			"resolved": "https://registry.npmjs.org/bun-types/-/bun-types-1.0.33.tgz",
-			"integrity": "sha512-L5tBIf9g6rBBkvshqysi5NoLQ9NnhSPU1pfJ9FzqoSfofYdyac3WLUnOIuQ+M5za/sooVUOP2ko+E6Tco0OLIA==",
-			"dev": true,
-			"dependencies": {
-				"@types/node": "~20.11.3",
-				"@types/ws": "~8.5.10"
-			}
-		},
 		"node_modules/cac": {
 			"version": "6.7.14",
 			"resolved": "https://registry.npmjs.org/cac/-/cac-6.7.14.tgz",
@@ -2777,6 +2828,11 @@
 				"node": ">=8"
 			}
 		},
+		"node_modules/classcat": {
+			"version": "5.0.5",
+			"resolved": "https://registry.npmjs.org/classcat/-/classcat-5.0.5.tgz",
+			"integrity": "sha512-JhZUT7JFcQy/EzW605k/ktHtncoo9vnyW/2GspNYwFlN1C/WmjuV/xtS04e9SOkL2sTdw0VAZ2UGCcQ9lR6p6w=="
+		},
 		"node_modules/clean-stack": {
 			"version": "2.2.0",
 			"resolved": "https://registry.npmjs.org/clean-stack/-/clean-stack-2.2.0.tgz",
@@ -6931,6 +6987,17 @@
 				"url": "https://github.com/sponsors/sindresorhus"
 			}
 		},
+		"node_modules/paneforge": {
+			"version": "0.0.6",
+			"resolved": "https://registry.npmjs.org/paneforge/-/paneforge-0.0.6.tgz",
+			"integrity": "sha512-jYeN/wdREihja5c6nK3S5jritDQ+EbCqC5NrDo97qCZzZ9GkmEcN5C0ZCjF4nmhBwkDKr6tLIgz4QUKWxLXjAw==",
+			"dependencies": {
+				"nanoid": "^5.0.4"
+			},
+			"peerDependencies": {
+				"svelte": "^4.0.0 || ^5.0.0-next.1"
+			}
+		},
 		"node_modules/parent-module": {
 			"version": "1.0.1",
 			"resolved": "https://registry.npmjs.org/parent-module/-/parent-module-1.0.1.tgz",
@@ -7094,9 +7161,9 @@
 			}
 		},
 		"node_modules/picocolors": {
-			"version": "1.0.1",
-			"resolved": "https://registry.npmjs.org/picocolors/-/picocolors-1.0.1.tgz",
-			"integrity": "sha512-anP1Z8qwhkbmu7MFP5iTt+wQKXgwzf7zTyGlcdzabySa9vd0Xt392U0rVmz9poOaBj0uHJKyyo9/upk0HrEQew=="
+			"version": "1.1.0",
+			"resolved": "https://registry.npmjs.org/picocolors/-/picocolors-1.1.0.tgz",
+			"integrity": "sha512-TQ92mBOW0l3LeMeyLV6mzy/kWr8lkd/hp3mTg7wYK7zJhuBStmGMBG0BdeDZS/dZx1IukaX6Bk11zcln25o1Aw=="
 		},
 		"node_modules/picomatch": {
 			"version": "2.3.1",
@@ -7162,9 +7229,9 @@
 			}
 		},
 		"node_modules/postcss": {
-			"version": "8.4.41",
-			"resolved": "https://registry.npmjs.org/postcss/-/postcss-8.4.41.tgz",
-			"integrity": "sha512-TesUflQ0WKZqAvg52PWL6kHgLKP6xB6heTOdoYM0Wt2UHyxNa4K25EZZMgKns3BH1RLVbZCREPpLY0rhnNoHVQ==",
+			"version": "8.4.47",
+			"resolved": "https://registry.npmjs.org/postcss/-/postcss-8.4.47.tgz",
+			"integrity": "sha512-56rxCq7G/XfB4EkXq9Egn5GCqugWvDFjafDOThIdMBsI15iqPqR5r15TfSr1YPYeEI19YeaXMCbY6u88Y76GLQ==",
 			"funding": [
 				{
 					"type": "opencollective",
@@ -7181,8 +7248,8 @@
 			],
 			"dependencies": {
 				"nanoid": "^3.3.7",
-				"picocolors": "^1.0.1",
-				"source-map-js": "^1.2.0"
+				"picocolors": "^1.1.0",
+				"source-map-js": "^1.2.1"
 			},
 			"engines": {
 				"node": "^10 || ^12 || >=14"
@@ -8195,9 +8262,9 @@
 			"integrity": "sha512-FJF5jgdfvoKn1MAKSdGs33bIqLi3LmsgVTliuX6iITj834F+JRQZN90Z93yql8h0K2t0RwDPBmxwlbZfDcxNZA=="
 		},
 		"node_modules/source-map-js": {
-			"version": "1.2.0",
-			"resolved": "https://registry.npmjs.org/source-map-js/-/source-map-js-1.2.0.tgz",
-			"integrity": "sha512-itJW8lvSA0TXEphiRoawsCksnlf8SyvmFzIhltqAHluXd88pkCd+cXJVHTDwdCr0IzwptSm035IHQktUu1QUMg==",
+			"version": "1.2.1",
+			"resolved": "https://registry.npmjs.org/source-map-js/-/source-map-js-1.2.1.tgz",
+			"integrity": "sha512-UXWMKhLOwVKb728IUtQPXxfYU+usdybtUrK/8uGE8CQMvrhOpwvzDBwj0QhSL7MQc7vIsISBG8VQ8+IDQxpfQA==",
 			"engines": {
 				"node": ">=0.10.0"
 			}
@@ -9112,7 +9179,7 @@
 			"version": "5.26.5",
 			"resolved": "https://registry.npmjs.org/undici-types/-/undici-types-5.26.5.tgz",
 			"integrity": "sha512-JlCMO+ehdEIKqlFxk6IfVoAUVmgz7cU7zD/h9XZ0qzeosSHmUJVOzSQvvYSYWXkFXC+IfLKSIffhv0sVZup6pA==",
-			"devOptional": true
+			"optional": true
 		},
 		"node_modules/unist-util-stringify-position": {
 			"version": "3.0.3",
@@ -9329,13 +9396,13 @@
 			}
 		},
 		"node_modules/vite": {
-			"version": "5.4.0",
-			"resolved": "https://registry.npmjs.org/vite/-/vite-5.4.0.tgz",
-			"integrity": "sha512-5xokfMX0PIiwCMCMb9ZJcMyh5wbBun0zUzKib+L65vAZ8GY9ePZMXxFrHbr/Kyll2+LSCY7xtERPpxkBDKngwg==",
+			"version": "5.4.6",
+			"resolved": "https://registry.npmjs.org/vite/-/vite-5.4.6.tgz",
+			"integrity": "sha512-IeL5f8OO5nylsgzd9tq4qD2QqI0k2CQLGrWD0rCN0EQJZpBK5vJAx0I+GDkMOXxQX/OfFHMuLIx6ddAxGX/k+Q==",
 			"dependencies": {
 				"esbuild": "^0.21.3",
-				"postcss": "^8.4.40",
-				"rollup": "^4.13.0"
+				"postcss": "^8.4.43",
+				"rollup": "^4.20.0"
 			},
 			"bin": {
 				"vite": "bin/vite.js"

package.json

@@ -1,6 +1,6 @@
 {
 	"name": "open-webui",
-	"version": "0.3.17",
+	"version": "0.3.23",
 	"private": true,
 	"scripts": {
 		"dev": "npm run pyodide:fetch && vite dev --host",
@@ -25,7 +25,6 @@
 		"@sveltejs/kit": "^2.5.20",
 		"@sveltejs/vite-plugin-svelte": "^3.1.1",
 		"@tailwindcss/typography": "^0.5.13",
-		"@types/bun": "latest",
 		"@typescript-eslint/eslint-plugin": "^6.17.0",
 		"@typescript-eslint/parser": "^6.17.0",
 		"autoprefixer": "^10.4.16",
@@ -54,6 +53,7 @@
 		"@codemirror/theme-one-dark": "^6.1.2",
 		"@pyscript/core": "^0.4.32",
 		"@sveltejs/adapter-node": "^2.0.0",
+		"@xyflow/svelte": "^0.1.19",
 		"async": "^3.2.5",
 		"bits-ui": "^0.19.7",
 		"codemirror": "^6.0.1",
@@ -72,6 +72,7 @@
 		"katex": "^0.16.9",
 		"marked": "^9.1.0",
 		"mermaid": "^10.9.1",
+		"paneforge": "^0.0.6",
 		"pyodide": "^0.26.1",
 		"socket.io-client": "^4.2.0",
 		"sortablejs": "^1.15.2",

pyproject.toml

@@ -47,7 +47,12 @@ dependencies = [
 
     "fake-useragent==1.5.1",
     "chromadb==0.5.5",
+    "pymilvus==2.4.6",
+
     "sentence-transformers==3.0.1",
+    "colbert-ai==0.2.21",
+    "einops==0.8.0",
+    
     "pypdf==4.3.1",
     "docx2txt==0.8",
     "python-pptx==1.0.0",
@@ -71,7 +76,7 @@ dependencies = [
     "faster-whisper==1.0.3",
 
     "PyJWT[crypto]==2.9.0",
-    "authlib==1.3.2",
+    "authlib==1.3.1",
 
     "black==24.8.0",
     "langfuse==2.44.0",

src/app.css

@@ -50,21 +50,6 @@ iframe {
 	@apply rounded-lg;
 }
 
-ol > li {
-	counter-increment: list-number;
-	display: block;
-	margin-bottom: 0;
-	margin-top: 0;
-	min-height: 28px;
-}
-
-.prose ol > li::before {
-	content: counters(list-number, '.') '.';
-	padding-right: 0.5rem;
-	color: var(--tw-prose-counters);
-	font-weight: 400;
-}
-
 li p {
 	display: inline;
 }
@@ -171,3 +156,20 @@ input[type='number'] {
 	font-weight: 600;
 	@apply rounded-md dark:bg-gray-800 bg-gray-100 mx-0.5;
 }
+
+.svelte-flow {
+	background-color: transparent !important;
+}
+
+.svelte-flow__edge > path {
+	stroke-width: 0.5;
+}
+
+.svelte-flow__edge.animated > path {
+	stroke-width: 2;
+	@apply stroke-gray-600 dark:stroke-gray-500;
+}
+
+.bg-gray-950-90 {
+	background-color: rgba(var(--color-gray-950, #0d0d0d), 0.9);
+}

src/app.html

@@ -4,7 +4,11 @@
 		<meta charset="utf-8" />
 		<link rel="icon" href="%sveltekit.assets%/favicon.png" />
 		<link rel="manifest" href="%sveltekit.assets%/manifest.json" crossorigin="use-credentials" />
-		<meta name="viewport" content="width=device-width, initial-scale=1, maximum-scale=1" />
+		<meta
+			name="viewport"
+			content="width=device-width, initial-scale=1, maximum-scale=1, viewport-fit=cover"
+		/>
+		<meta name="theme-color" content="#171717" />
 		<meta name="robots" content="noindex,nofollow" />
 		<meta name="description" content="Open WebUI" />
 		<link
@@ -23,33 +27,33 @@
 		<script>
 			// On page load or when changing themes, best to add inline in `head` to avoid FOUC
 			(() => {
+				const metaThemeColorTag = document.querySelector('meta[name="theme-color"]');
+				const prefersDarkTheme = window.matchMedia('(prefers-color-scheme: dark)').matches;
+
 				if (!localStorage?.theme) {
 					localStorage.theme = 'system';
 				}
 
-				if (localStorage?.theme && localStorage?.theme.includes('oled')) {
+				if (localStorage.theme === 'system') {
+					document.documentElement.classList.add(prefersDarkTheme ? 'dark' : 'light');
+					metaThemeColorTag.setAttribute('content', prefersDarkTheme ? '#171717' : '#ffffff');
+				} else if (localStorage.theme === 'oled-dark') {
 					document.documentElement.style.setProperty('--color-gray-800', '#101010');
 					document.documentElement.style.setProperty('--color-gray-850', '#050505');
 					document.documentElement.style.setProperty('--color-gray-900', '#000000');
 					document.documentElement.style.setProperty('--color-gray-950', '#000000');
 					document.documentElement.classList.add('dark');
-				} else if (
-					localStorage.theme === 'light' ||
-					(!('theme' in localStorage) && window.matchMedia('(prefers-color-scheme: light)').matches)
-				) {
+					metaThemeColorTag.setAttribute('content', '#000000');
+				} else if (localStorage.theme === 'light') {
 					document.documentElement.classList.add('light');
-				} else if (localStorage.theme && localStorage.theme !== 'system') {
-					localStorage.theme.split(' ').forEach((e) => {
-						document.documentElement.classList.add(e);
-					});
-				} else if (localStorage.theme && localStorage.theme === 'system') {
-					systemTheme = window.matchMedia('(prefers-color-scheme: dark)').matches;
-					document.documentElement.classList.add(systemTheme ? 'dark' : 'light');
-				} else if (localStorage.theme && localStorage.theme === 'her') {
+					metaThemeColorTag.setAttribute('content', '#ffffff');
+				} else if (localStorage.theme === 'her') {
 					document.documentElement.classList.add('dark');
 					document.documentElement.classList.add('her');
+					metaThemeColorTag.setAttribute('content', '#983724');
 				} else {
 					document.documentElement.classList.add('dark');
+					metaThemeColorTag.setAttribute('content', '#171717');
 				}
 
 				window.matchMedia('(prefers-color-scheme: dark)').addListener((e) => {
@@ -57,9 +61,11 @@
 						if (e.matches) {
 							document.documentElement.classList.add('dark');
 							document.documentElement.classList.remove('light');
+							metaThemeColorTag.setAttribute('content', '#171717');
 						} else {
 							document.documentElement.classList.add('light');
 							document.documentElement.classList.remove('dark');
+							metaThemeColorTag.setAttribute('content', '#ffffff');
 						}
 					}
 				});

src/lib/components/admin/Settings/Audio.svelte

@@ -31,6 +31,8 @@
 	let TTS_MODEL = '';
 	let TTS_VOICE = '';
 	let TTS_SPLIT_ON: TTS_RESPONSE_SPLIT = TTS_RESPONSE_SPLIT.PUNCTUATION;
+	let TTS_AZURE_SPEECH_REGION = '';
+	let TTS_AZURE_SPEECH_OUTPUT_FORMAT = '';
 
 	let STT_OPENAI_API_BASE_URL = '';
 	let STT_OPENAI_API_KEY = '';
@@ -87,7 +89,9 @@
 				ENGINE: TTS_ENGINE,
 				MODEL: TTS_MODEL,
 				VOICE: TTS_VOICE,
-				SPLIT_ON: TTS_SPLIT_ON
+				SPLIT_ON: TTS_SPLIT_ON,
+				AZURE_SPEECH_REGION: TTS_AZURE_SPEECH_REGION,
+				AZURE_SPEECH_OUTPUT_FORMAT: TTS_AZURE_SPEECH_OUTPUT_FORMAT
 			},
 			stt: {
 				OPENAI_API_BASE_URL: STT_OPENAI_API_BASE_URL,
@@ -120,6 +124,9 @@
 
 			TTS_SPLIT_ON = res.tts.SPLIT_ON || TTS_RESPONSE_SPLIT.PUNCTUATION;
 
+			TTS_AZURE_SPEECH_OUTPUT_FORMAT = res.tts.AZURE_SPEECH_OUTPUT_FORMAT;
+			TTS_AZURE_SPEECH_REGION = res.tts.AZURE_SPEECH_REGION;
+
 			STT_OPENAI_API_BASE_URL = res.stt.OPENAI_API_BASE_URL;
 			STT_OPENAI_API_KEY = res.stt.OPENAI_API_KEY;
 
@@ -224,6 +231,7 @@
 							<option value="">{$i18n.t('Web API')}</option>
 							<option value="openai">{$i18n.t('OpenAI')}</option>
 							<option value="elevenlabs">{$i18n.t('ElevenLabs')}</option>
+							<option value="azure">{$i18n.t('Azure AI Speech')}</option>
 						</select>
 					</div>
 				</div>
@@ -252,6 +260,23 @@
 							/>
 						</div>
 					</div>
+				{:else if TTS_ENGINE === 'azure'}
+					<div>
+						<div class="mt-1 flex gap-2 mb-1">
+							<input
+								class="flex-1 w-full rounded-lg py-2 pl-4 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none"
+								placeholder={$i18n.t('API Key')}
+								bind:value={TTS_API_KEY}
+								required
+							/>
+							<input
+								class="flex-1 w-full rounded-lg py-2 pl-4 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none"
+								placeholder={$i18n.t('Azure Region')}
+								bind:value={TTS_AZURE_SPEECH_REGION}
+								required
+							/>
+						</div>
+					</div>
 				{/if}
 
 				<hr class=" dark:border-gray-850 my-2" />
@@ -359,6 +384,49 @@
 							</div>
 						</div>
 					</div>
+				{:else if TTS_ENGINE === 'azure'}
+					<div class=" flex gap-2">
+						<div class="w-full">
+							<div class=" mb-1.5 text-sm font-medium">{$i18n.t('TTS Voice')}</div>
+							<div class="flex w-full">
+								<div class="flex-1">
+									<input
+										list="voice-list"
+										class="w-full rounded-lg py-2 px-4 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none"
+										bind:value={TTS_VOICE}
+										placeholder="Select a voice"
+									/>
+
+									<datalist id="voice-list">
+										{#each voices as voice}
+											<option value={voice.id}>{voice.name}</option>
+										{/each}
+									</datalist>
+								</div>
+							</div>
+						</div>
+						<div class="w-full">
+							<div class=" mb-1.5 text-sm font-medium">
+								{$i18n.t('Output format')}
+								<a
+									href="https://learn.microsoft.com/en-us/azure/ai-services/speech-service/rest-text-to-speech?tabs=streaming#audio-outputs"
+									target="_blank"
+								>
+									<small>{$i18n.t('Available list')}</small>
+								</a>
+							</div>
+							<div class="flex w-full">
+								<div class="flex-1">
+									<input
+										list="tts-model-list"
+										class="w-full rounded-lg py-2 px-4 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none"
+										bind:value={TTS_AZURE_SPEECH_OUTPUT_FORMAT}
+										placeholder="Select a output format"
+									/>
+								</div>
+							</div>
+						</div>
+					</div>
 				{/if}
 
 				<hr class="dark:border-gray-850 my-2" />

src/lib/components/admin/Settings/Connections.svelte

@@ -150,18 +150,20 @@
 				})()
 			]);
 
-			OPENAI_API_BASE_URLS.forEach(async (url, idx) => {
-				const res = await getOpenAIModels(localStorage.token, idx);
-				if (res.pipelines) {
-					pipelineUrls[url] = true;
-				}
-			});
-
 			const ollamaConfig = await getOllamaConfig(localStorage.token);
 			const openaiConfig = await getOpenAIConfig(localStorage.token);
 
 			ENABLE_OPENAI_API = openaiConfig.ENABLE_OPENAI_API;
 			ENABLE_OLLAMA_API = ollamaConfig.ENABLE_OLLAMA_API;
+
+			if (ENABLE_OPENAI_API) {
+				OPENAI_API_BASE_URLS.forEach(async (url, idx) => {
+					const res = await getOpenAIModels(localStorage.token, idx);
+					if (res.pipelines) {
+						pipelineUrls[url] = true;
+					}
+				});
+			}
 		}
 	});
 </script>

src/lib/components/admin/Settings/Documents.svelte

@@ -732,11 +732,17 @@
 
 			<div>
 				<div class=" mb-2.5 text-sm font-medium">{$i18n.t('RAG Template')}</div>
-				<textarea
-					bind:value={querySettings.template}
-					class="w-full rounded-lg px-4 py-3 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none resize-none"
-					rows="4"
-				/>
+				<Tooltip
+					content={$i18n.t('Leave empty to use the default prompt, or enter a custom prompt')}
+					placement="top-start"
+				>
+					<textarea
+						bind:value={querySettings.template}
+						placeholder={$i18n.t('Leave empty to use the default prompt, or enter a custom prompt')}
+						class="w-full rounded-lg px-4 py-3 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none resize-none"
+						rows="4"
+					/>
+				</Tooltip>
 			</div>
 		</div>
 

src/lib/components/admin/Settings/Images.svelte

@@ -27,6 +27,43 @@
 
 	let models = null;
 
+	let samplers = [
+		'DPM++ 2M',
+		'DPM++ SDE',
+		'DPM++ 2M SDE',
+		'DPM++ 2M SDE Heun',
+		'DPM++ 2S a',
+		'DPM++ 3M SDE',
+		'Euler a',
+		'Euler',
+		'LMS',
+		'Heun',
+		'DPM2',
+		'DPM2 a',
+		'DPM fast',
+		'DPM adaptive',
+		'Restart',
+		'DDIM',
+		'DDIM CFG++',
+		'PLMS',
+		'UniPC'
+	];
+
+	let schedulers = [
+		'Automatic',
+		'Uniform',
+		'Karras',
+		'Exponential',
+		'Polyexponential',
+		'SGM Uniform',
+		'KL Optimal',
+		'Align Your Steps',
+		'Simple',
+		'Normal',
+		'DDIM',
+		'Beta'
+	];
+
 	let requiredWorkflowNodes = [
 		{
 			type: 'prompt',
@@ -326,6 +363,66 @@
 							</a>
 						</div>
 					</div>
+
+					<!---Sampler-->
+					<div>
+						<div class=" mb-2.5 text-sm font-medium">{$i18n.t('Set Sampler')}</div>
+						<div class="flex w-full">
+							<div class="flex-1 mr-2">
+								<Tooltip content={$i18n.t('Enter Sampler (e.g. Euler a)')} placement="top-start">
+									<input
+										list="sampler-list"
+										class="w-full rounded-lg py-2 px-4 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none"
+										placeholder={$i18n.t('Enter Sampler (e.g. Euler a)')}
+										bind:value={config.automatic1111.AUTOMATIC1111_SAMPLER}
+									/>
+
+									<datalist id="sampler-list">
+										{#each samplers ?? [] as sampler}
+											<option value={sampler}>{sampler}</option>
+										{/each}
+									</datalist>
+								</Tooltip>
+							</div>
+						</div>
+					</div>
+					<!---Scheduler-->
+					<div>
+						<div class=" mb-2.5 text-sm font-medium">{$i18n.t('Set Scheduler')}</div>
+						<div class="flex w-full">
+							<div class="flex-1 mr-2">
+								<Tooltip content={$i18n.t('Enter Scheduler (e.g. Karras)')} placement="top-start">
+									<input
+										list="scheduler-list"
+										class="w-full rounded-lg py-2 px-4 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none"
+										placeholder={$i18n.t('Enter Scheduler (e.g. Karras)')}
+										bind:value={config.automatic1111.AUTOMATIC1111_SCHEDULER}
+									/>
+
+									<datalist id="scheduler-list">
+										{#each schedulers ?? [] as scheduler}
+											<option value={scheduler}>{scheduler}</option>
+										{/each}
+									</datalist>
+								</Tooltip>
+							</div>
+						</div>
+					</div>
+					<!---CFG scale-->
+					<div>
+						<div class=" mb-2.5 text-sm font-medium">{$i18n.t('Set CFG Scale')}</div>
+						<div class="flex w-full">
+							<div class="flex-1 mr-2">
+								<Tooltip content={$i18n.t('Enter CFG Scale (e.g. 7.0)')} placement="top-start">
+									<input
+										class="w-full rounded-lg py-2 px-4 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none"
+										placeholder={$i18n.t('Enter CFG Scale (e.g. 7.0)')}
+										bind:value={config.automatic1111.AUTOMATIC1111_CFG_SCALE}
+									/>
+								</Tooltip>
+							</div>
+						</div>
+					</div>
 				{:else if config?.engine === 'comfyui'}
 					<div class="">
 						<div class=" mb-2 text-sm font-medium">{$i18n.t('ComfyUI Base URL')}</div>

src/lib/components/admin/Settings/Interface.svelte

@@ -23,8 +23,8 @@
 		TASK_MODEL: '',
 		TASK_MODEL_EXTERNAL: '',
 		TITLE_GENERATION_PROMPT_TEMPLATE: '',
-		SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE: '',
-		SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD: 0
+		ENABLE_SEARCH_QUERY: true,
+		SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE: ''
 	};
 
 	let promptSuggestions = [];
@@ -43,7 +43,6 @@
 		taskConfig = await getTaskConfig(localStorage.token);
 
 		promptSuggestions = $config?.default_prompt_suggestions;
-
 		banners = await getBanners(localStorage.token);
 	});
 
@@ -119,33 +118,50 @@
 			</div>
 
 			<div class="mt-3">
-				<div class=" mb-2.5 text-sm font-medium">{$i18n.t('Title Generation Prompt')}</div>
-				<textarea
-					bind:value={taskConfig.TITLE_GENERATION_PROMPT_TEMPLATE}
-					class="w-full rounded-lg py-3 px-4 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none resize-none"
-					rows="6"
-				/>
+				<div class=" mb-2.5 text-xs font-medium">{$i18n.t('Title Generation Prompt')}</div>
+
+				<Tooltip
+					content={$i18n.t('Leave empty to use the default prompt, or enter a custom prompt')}
+					placement="top-start"
+				>
+					<textarea
+						bind:value={taskConfig.TITLE_GENERATION_PROMPT_TEMPLATE}
+						class="w-full rounded-lg py-3 px-4 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none resize-none"
+						rows="3"
+						placeholder={$i18n.t('Leave empty to use the default prompt, or enter a custom prompt')}
+					/>
+				</Tooltip>
 			</div>
 
-			<div class="mt-3">
-				<div class=" mb-2.5 text-sm font-medium">{$i18n.t('Search Query Generation Prompt')}</div>
-				<textarea
-					bind:value={taskConfig.SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE}
-					class="w-full rounded-lg py-3 px-4 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none resize-none"
-					rows="6"
-				/>
-			</div>
+			<hr class=" dark:border-gray-850 my-3" />
 
-			<div class="mt-3">
-				<div class=" mb-2.5 text-sm font-medium">
-					{$i18n.t('Search Query Generation Prompt Length Threshold')}
+			<div class="my-3 flex w-full items-center justify-between">
+				<div class=" self-center text-xs font-medium">
+					{$i18n.t('Enable Web Search Query Generation')}
 				</div>
-				<input
-					bind:value={taskConfig.SEARCH_QUERY_PROMPT_LENGTH_THRESHOLD}
-					class="w-full rounded-lg py-2 px-4 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none resize-none"
-					type="number"
-				/>
+
+				<Switch bind:state={taskConfig.ENABLE_SEARCH_QUERY} />
 			</div>
+
+			{#if taskConfig.ENABLE_SEARCH_QUERY}
+				<div class="">
+					<div class=" mb-2.5 text-xs font-medium">{$i18n.t('Search Query Generation Prompt')}</div>
+
+					<Tooltip
+						content={$i18n.t('Leave empty to use the default prompt, or enter a custom prompt')}
+						placement="top-start"
+					>
+						<textarea
+							bind:value={taskConfig.SEARCH_QUERY_GENERATION_PROMPT_TEMPLATE}
+							class="w-full rounded-lg py-3 px-4 text-sm bg-gray-50 dark:text-gray-300 dark:bg-gray-850 outline-none resize-none"
+							rows="3"
+							placeholder={$i18n.t(
+								'Leave empty to use the default prompt, or enter a custom prompt'
+							)}
+						/>
+					</Tooltip>
+				</div>
+			{/if}
 		</div>
 
 		<hr class=" dark:border-gray-850 my-3" />
@@ -291,9 +307,10 @@
 									/>
 								</div>
 
-								<input
-									class="px-3 py-1.5 text-xs w-full bg-transparent outline-none border-r border-gray-100 dark:border-gray-800"
+								<textarea
+									class="px-3 py-1.5 text-xs w-full bg-transparent outline-none border-r border-gray-100 dark:border-gray-800 resize-none"
 									placeholder={$i18n.t('Prompt (e.g. Tell me a fun fact about the Roman Empire)')}
+									rows="3"
 									bind:value={prompt.content}
 								/>
 							</div>

src/lib/components/chat/Chat.svelte

@@ -2,6 +2,7 @@
 	import { v4 as uuidv4 } from 'uuid';
 	import { toast } from 'svelte-sonner';
 	import mermaid from 'mermaid';
+	import { PaneGroup, Pane, PaneResizer } from 'paneforge';
 
 	import { getContext, onDestroy, onMount, tick } from 'svelte';
 	import { goto } from '$app/navigation';
@@ -23,9 +24,12 @@
 		banners,
 		user,
 		socket,
+		showControls,
 		showCallOverlay,
 		currentChatPage,
-		temporaryChatEnabled
+		temporaryChatEnabled,
+		mobile,
+		showOverview
 	} from '$lib/stores';
 	import {
 		convertMessagesToHistory,
@@ -63,14 +67,15 @@
 	import Navbar from '$lib/components/layout/Navbar.svelte';
 	import ChatControls from './ChatControls.svelte';
 	import EventConfirmDialog from '../common/ConfirmDialog.svelte';
+	import EllipsisVertical from '../icons/EllipsisVertical.svelte';
 
 	const i18n: Writable<i18nType> = getContext('i18n');
 
 	export let chatIdProp = '';
 	let loaded = false;
 	const eventTarget = new EventTarget();
+	let controlPane;
 
-	let showControls = false;
 	let stopResponseFlag = false;
 	let autoScroll = true;
 	let processing = '';
@@ -115,13 +120,14 @@
 
 	$: if (history.currentId !== null) {
 		let _messages = [];
-
 		let currentMessage = history.messages[history.currentId];
-		while (currentMessage !== null) {
+		while (currentMessage) {
 			_messages.unshift({ ...currentMessage });
 			currentMessage =
 				currentMessage.parentId !== null ? history.messages[currentMessage.parentId] : null;
 		}
+
+		// This is most likely causing the performance issue
 		messages = _messages;
 	} else {
 		messages = [];
@@ -143,6 +149,28 @@
 		})();
 	}
 
+	const showMessage = async (message) => {
+		let _messageId = JSON.parse(JSON.stringify(message.id));
+
+		let messageChildrenIds = history.messages[_messageId].childrenIds;
+
+		while (messageChildrenIds.length !== 0) {
+			_messageId = messageChildrenIds.at(-1);
+			messageChildrenIds = history.messages[_messageId].childrenIds;
+		}
+
+		history.currentId = _messageId;
+
+		await tick();
+		await tick();
+		await tick();
+
+		const messageElement = document.getElementById(`message-${message.id}`);
+		if (messageElement) {
+			messageElement.scrollIntoView({ behavior: 'smooth' });
+		}
+	};
+
 	const chatEventHandler = async (event, cb) => {
 		if (event.chat_id === $chatId) {
 			await tick();
@@ -256,6 +284,29 @@
 				await goto('/');
 			}
 		}
+
+		showControls.subscribe(async (value) => {
+			if (controlPane && !$mobile) {
+				try {
+					if (value) {
+						controlPane.resize(
+							parseInt(localStorage.getItem('chat-controls-size') || '35')
+								? parseInt(localStorage.getItem('chat-controls-size') || '35')
+								: 35
+						);
+					} else {
+						controlPane.resize(0);
+					}
+				} catch (e) {
+					// ignore
+				}
+			}
+
+			if (!value) {
+				showCallOverlay.set(false);
+				showOverview.set(false);
+			}
+		});
 	});
 
 	onDestroy(() => {
@@ -732,6 +783,8 @@
 					responseMessage.userContext = userContext;
 
 					const chatEventEmitter = await getChatEventEmitter(model.id, _chatId);
+
+					scrollToBottom();
 					if (webSearchEnabled) {
 						await getWebSearchResults(model.id, parentId, responseMessageId);
 					}
@@ -858,8 +911,9 @@
 
 		await tick();
 
+		const stream = $settings?.streamResponse ?? true;
 		const [res, controller] = await generateChatCompletion(localStorage.token, {
-			stream: true,
+			stream: stream,
 			model: model.id,
 			messages: messagesBody,
 			options: {
@@ -884,142 +938,162 @@
 		});
 
 		if (res && res.ok) {
-			console.log('controller', controller);
+			if (!stream) {
+				const response = await res.json();
+				console.log(response);
 
-			const reader = res.body
-				.pipeThrough(new TextDecoderStream())
-				.pipeThrough(splitStream('\n'))
-				.getReader();
+				responseMessage.content = response.message.content;
+				responseMessage.info = {
+					eval_count: response.eval_count,
+					eval_duration: response.eval_duration,
+					load_duration: response.load_duration,
+					prompt_eval_count: response.prompt_eval_count,
+					prompt_eval_duration: response.prompt_eval_duration,
+					total_duration: response.total_duration
+				};
+				responseMessage.done = true;
+			} else {
+				console.log('controller', controller);
 
-			while (true) {
-				const { value, done } = await reader.read();
-				if (done || stopResponseFlag || _chatId !== $chatId) {
-					responseMessage.done = true;
-					messages = messages;
+				const reader = res.body
+					.pipeThrough(new TextDecoderStream())
+					.pipeThrough(splitStream('\n'))
+					.getReader();
 
-					if (stopResponseFlag) {
-						controller.abort('User: Stop Response');
-					} else {
-						const messages = createMessagesList(responseMessageId);
-						await chatCompletedHandler(_chatId, model.id, responseMessageId, messages);
+				while (true) {
+					const { value, done } = await reader.read();
+					if (done || stopResponseFlag || _chatId !== $chatId) {
+						responseMessage.done = true;
+						messages = messages;
+
+						if (stopResponseFlag) {
+							controller.abort('User: Stop Response');
+						}
+
+						_response = responseMessage.content;
+						break;
 					}
 
-					_response = responseMessage.content;
-					break;
-				}
+					try {
+						let lines = value.split('\n');
 
-				try {
-					let lines = value.split('\n');
+						for (const line of lines) {
+							if (line !== '') {
+								console.log(line);
+								let data = JSON.parse(line);
 
-					for (const line of lines) {
-						if (line !== '') {
-							console.log(line);
-							let data = JSON.parse(line);
-
-							if ('citations' in data) {
-								responseMessage.citations = data.citations;
-								// Only remove status if it was initially set
-								if (model?.info?.meta?.knowledge ?? false) {
-									responseMessage.statusHistory = responseMessage.statusHistory.filter(
-										(status) => status.action !== 'knowledge_search'
-									);
-								}
-								continue;
-							}
-
-							if ('detail' in data) {
-								throw data;
-							}
-
-							if (data.done == false) {
-								if (responseMessage.content == '' && data.message.content == '\n') {
-									continue;
-								} else {
-									responseMessage.content += data.message.content;
-
-									if (navigator.vibrate && ($settings?.hapticFeedback ?? false)) {
-										navigator.vibrate(5);
-									}
-
-									const messageContentParts = getMessageContentParts(
-										responseMessage.content,
-										$config?.audio?.tts?.split_on ?? 'punctuation'
-									);
-									messageContentParts.pop();
-
-									// dispatch only last sentence and make sure it hasn't been dispatched before
-									if (
-										messageContentParts.length > 0 &&
-										messageContentParts[messageContentParts.length - 1] !==
-											responseMessage.lastSentence
-									) {
-										responseMessage.lastSentence =
-											messageContentParts[messageContentParts.length - 1];
-										eventTarget.dispatchEvent(
-											new CustomEvent('chat', {
-												detail: {
-													id: responseMessageId,
-													content: messageContentParts[messageContentParts.length - 1]
-												}
-											})
+								if ('citations' in data) {
+									responseMessage.citations = data.citations;
+									// Only remove status if it was initially set
+									if (model?.info?.meta?.knowledge ?? false) {
+										responseMessage.statusHistory = responseMessage.statusHistory.filter(
+											(status) => status.action !== 'knowledge_search'
 										);
 									}
-
-									messages = messages;
+									continue;
 								}
-							} else {
-								responseMessage.done = true;
 
-								if (responseMessage.content == '') {
-									responseMessage.error = {
-										code: 400,
-										content: `Oops! No text generated from Ollama, Please try again.`
+								if ('detail' in data) {
+									throw data;
+								}
+
+								if (data.done == false) {
+									if (responseMessage.content == '' && data.message.content == '\n') {
+										continue;
+									} else {
+										responseMessage.content += data.message.content;
+
+										if (navigator.vibrate && ($settings?.hapticFeedback ?? false)) {
+											navigator.vibrate(5);
+										}
+
+										const messageContentParts = getMessageContentParts(
+											responseMessage.content,
+											$config?.audio?.tts?.split_on ?? 'punctuation'
+										);
+										messageContentParts.pop();
+
+										// dispatch only last sentence and make sure it hasn't been dispatched before
+										if (
+											messageContentParts.length > 0 &&
+											messageContentParts[messageContentParts.length - 1] !==
+												responseMessage.lastSentence
+										) {
+											responseMessage.lastSentence =
+												messageContentParts[messageContentParts.length - 1];
+											eventTarget.dispatchEvent(
+												new CustomEvent('chat', {
+													detail: {
+														id: responseMessageId,
+														content: messageContentParts[messageContentParts.length - 1]
+													}
+												})
+											);
+										}
+
+										messages = messages;
+									}
+								} else {
+									responseMessage.done = true;
+
+									if (responseMessage.content == '') {
+										responseMessage.error = {
+											code: 400,
+											content: `Oops! No text generated from Ollama, Please try again.`
+										};
+									}
+
+									responseMessage.context = data.context ?? null;
+									responseMessage.info = {
+										total_duration: data.total_duration,
+										load_duration: data.load_duration,
+										sample_count: data.sample_count,
+										sample_duration: data.sample_duration,
+										prompt_eval_count: data.prompt_eval_count,
+										prompt_eval_duration: data.prompt_eval_duration,
+										eval_count: data.eval_count,
+										eval_duration: data.eval_duration
 									};
-								}
+									messages = messages;
 
-								responseMessage.context = data.context ?? null;
-								responseMessage.info = {
-									total_duration: data.total_duration,
-									load_duration: data.load_duration,
-									sample_count: data.sample_count,
-									sample_duration: data.sample_duration,
-									prompt_eval_count: data.prompt_eval_count,
-									prompt_eval_duration: data.prompt_eval_duration,
-									eval_count: data.eval_count,
-									eval_duration: data.eval_duration
-								};
-								messages = messages;
+									if ($settings.notificationEnabled && !document.hasFocus()) {
+										const notification = new Notification(`${model.id}`, {
+											body: responseMessage.content,
+											icon: `${WEBUI_BASE_URL}/static/favicon.png`
+										});
+									}
 
-								if ($settings.notificationEnabled && !document.hasFocus()) {
-									const notification = new Notification(`${model.id}`, {
-										body: responseMessage.content,
-										icon: `${WEBUI_BASE_URL}/static/favicon.png`
-									});
-								}
+									if ($settings?.responseAutoCopy ?? false) {
+										copyToClipboard(responseMessage.content);
+									}
 
-								if ($settings?.responseAutoCopy ?? false) {
-									copyToClipboard(responseMessage.content);
-								}
-
-								if ($settings.responseAutoPlayback && !$showCallOverlay) {
-									await tick();
-									document.getElementById(`speak-button-${responseMessage.id}`)?.click();
+									if ($settings.responseAutoPlayback && !$showCallOverlay) {
+										await tick();
+										document.getElementById(`speak-button-${responseMessage.id}`)?.click();
+									}
 								}
 							}
 						}
+					} catch (error) {
+						console.log(error);
+						if ('detail' in error) {
+							toast.error(error.detail);
+						}
+						break;
 					}
-				} catch (error) {
-					console.log(error);
-					if ('detail' in error) {
-						toast.error(error.detail);
-					}
-					break;
-				}
 
-				if (autoScroll) {
-					scrollToBottom();
+					if (autoScroll) {
+						scrollToBottom();
+					}
 				}
 			}
+
+			await chatCompletedHandler(
+				_chatId,
+				model.id,
+				responseMessageId,
+				createMessagesList(responseMessageId)
+			);
 		} else {
 			if (res !== null) {
 				const error = await res.json();
@@ -1131,17 +1205,19 @@
 		await tick();
 
 		try {
+			const stream = $settings?.streamResponse ?? true;
 			const [res, controller] = await generateOpenAIChatCompletion(
 				localStorage.token,
 				{
-					stream: true,
+					stream: stream,
 					model: model.id,
-					stream_options:
-						(model.info?.meta?.capabilities?.usage ?? false)
-							? {
+					...(stream && (model.info?.meta?.capabilities?.usage ?? false)
+						? {
+								stream_options: {
 									include_usage: true
 								}
-							: undefined,
+							}
+						: {}),
 					messages: [
 						params?.system || $settings.system || (responseMessage?.userContext ?? null)
 							? {
@@ -1219,85 +1295,95 @@
 			scrollToBottom();
 
 			if (res && res.ok && res.body) {
-				const textStream = await createOpenAITextStream(res.body, $settings.splitLargeChunks);
+				if (!stream) {
+					const response = await res.json();
+					console.log(response);
 
-				for await (const update of textStream) {
-					const { value, done, citations, error, usage } = update;
-					if (error) {
-						await handleOpenAIError(error, null, model, responseMessage);
-						break;
-					}
-					if (done || stopResponseFlag || _chatId !== $chatId) {
-						responseMessage.done = true;
-						messages = messages;
+					responseMessage.content = response.choices[0].message.content;
+					responseMessage.info = { ...response.usage, openai: true };
+					responseMessage.done = true;
+				} else {
+					const textStream = await createOpenAITextStream(res.body, $settings.splitLargeChunks);
 
-						if (stopResponseFlag) {
-							controller.abort('User: Stop Response');
+					for await (const update of textStream) {
+						const { value, done, citations, error, usage } = update;
+						if (error) {
+							await handleOpenAIError(error, null, model, responseMessage);
+							break;
+						}
+						if (done || stopResponseFlag || _chatId !== $chatId) {
+							responseMessage.done = true;
+							messages = messages;
+
+							if (stopResponseFlag) {
+								controller.abort('User: Stop Response');
+							}
+							_response = responseMessage.content;
+							break;
+						}
+
+						if (usage) {
+							responseMessage.info = { ...usage, openai: true };
+						}
+
+						if (citations) {
+							responseMessage.citations = citations;
+							// Only remove status if it was initially set
+							if (model?.info?.meta?.knowledge ?? false) {
+								responseMessage.statusHistory = responseMessage.statusHistory.filter(
+									(status) => status.action !== 'knowledge_search'
+								);
+							}
+							continue;
+						}
+
+						if (responseMessage.content == '' && value == '\n') {
+							continue;
 						} else {
-							const messages = createMessagesList(responseMessageId);
+							responseMessage.content += value;
 
-							await chatCompletedHandler(_chatId, model.id, responseMessageId, messages);
-						}
+							if (navigator.vibrate && ($settings?.hapticFeedback ?? false)) {
+								navigator.vibrate(5);
+							}
 
-						_response = responseMessage.content;
-
-						break;
-					}
-
-					if (usage) {
-						responseMessage.info = { ...usage, openai: true };
-					}
-
-					if (citations) {
-						responseMessage.citations = citations;
-						// Only remove status if it was initially set
-						if (model?.info?.meta?.knowledge ?? false) {
-							responseMessage.statusHistory = responseMessage.statusHistory.filter(
-								(status) => status.action !== 'knowledge_search'
+							const messageContentParts = getMessageContentParts(
+								responseMessage.content,
+								$config?.audio?.tts?.split_on ?? 'punctuation'
 							);
-						}
-						continue;
-					}
+							messageContentParts.pop();
 
-					if (responseMessage.content == '' && value == '\n') {
-						continue;
-					} else {
-						responseMessage.content += value;
+							// dispatch only last sentence and make sure it hasn't been dispatched before
+							if (
+								messageContentParts.length > 0 &&
+								messageContentParts[messageContentParts.length - 1] !== responseMessage.lastSentence
+							) {
+								responseMessage.lastSentence = messageContentParts[messageContentParts.length - 1];
+								eventTarget.dispatchEvent(
+									new CustomEvent('chat', {
+										detail: {
+											id: responseMessageId,
+											content: messageContentParts[messageContentParts.length - 1]
+										}
+									})
+								);
+							}
 
-						if (navigator.vibrate && ($settings?.hapticFeedback ?? false)) {
-							navigator.vibrate(5);
+							messages = messages;
 						}
 
-						const messageContentParts = getMessageContentParts(
-							responseMessage.content,
-							$config?.audio?.tts?.split_on ?? 'punctuation'
-						);
-						messageContentParts.pop();
-
-						// dispatch only last sentence and make sure it hasn't been dispatched before
-						if (
-							messageContentParts.length > 0 &&
-							messageContentParts[messageContentParts.length - 1] !== responseMessage.lastSentence
-						) {
-							responseMessage.lastSentence = messageContentParts[messageContentParts.length - 1];
-							eventTarget.dispatchEvent(
-								new CustomEvent('chat', {
-									detail: {
-										id: responseMessageId,
-										content: messageContentParts[messageContentParts.length - 1]
-									}
-								})
-							);
+						if (autoScroll) {
+							scrollToBottom();
 						}
-
-						messages = messages;
-					}
-
-					if (autoScroll) {
-						scrollToBottom();
 					}
 				}
 
+				await chatCompletedHandler(
+					_chatId,
+					model.id,
+					responseMessageId,
+					createMessagesList(responseMessageId)
+				);
+
 				if ($settings.notificationEnabled && !document.hasFocus()) {
 					const notification = new Notification(`${model.id}`, {
 						body: responseMessage.content,
@@ -1512,23 +1598,25 @@
 		messages = messages;
 
 		const prompt = userMessage.content;
-		let searchQuery = await generateSearchQuery(localStorage.token, model, messages, prompt).catch(
-			(error) => {
-				console.log(error);
-				return prompt;
-			}
-		);
+		let searchQuery = await generateSearchQuery(
+			localStorage.token,
+			model,
+			messages.filter((message) => message?.content?.trim()),
+			prompt
+		).catch((error) => {
+			console.log(error);
+			return prompt;
+		});
 
-		if (!searchQuery) {
-			toast.warning($i18n.t('No search query generated'));
+		if (!searchQuery || searchQuery == '') {
 			responseMessage.statusHistory.push({
 				done: true,
 				error: true,
 				action: 'web_search',
-				description: 'No search query generated'
+				description: $i18n.t('No search query generated')
 			});
-
 			messages = messages;
+			return;
 		}
 
 		responseMessage.statusHistory.push({
@@ -1699,117 +1787,117 @@
 			{title}
 			bind:selectedModels
 			bind:showModelSelector
-			bind:showControls
 			shareEnabled={messages.length > 0}
 			{chat}
 			{initNewChat}
 		/>
 
-		{#if $banners.length > 0 && messages.length === 0 && !$chatId && selectedModels.length <= 1}
-			<div
-				class="absolute top-[4.25rem] w-full {$showSidebar
-					? 'md:max-w-[calc(100%-260px)]'
-					: ''} {showControls ? 'lg:pr-[24rem]' : ''} z-20"
-			>
-				<div class=" flex flex-col gap-1 w-full">
-					{#each $banners.filter( (b) => (b.dismissible ? !JSON.parse(localStorage.getItem('dismissedBannerIds') ?? '[]').includes(b.id) : true) ) as banner}
-						<Banner
-							{banner}
-							on:dismiss={(e) => {
-								const bannerId = e.detail;
+		<PaneGroup direction="horizontal" class="w-full h-full">
+			<Pane defaultSize={50} class="h-full flex w-full relative">
+				{#if $banners.length > 0 && messages.length === 0 && !$chatId && selectedModels.length <= 1}
+					<div class="absolute top-3 left-0 right-0 w-full z-20">
+						<div class=" flex flex-col gap-1 w-full">
+							{#each $banners.filter( (b) => (b.dismissible ? !JSON.parse(localStorage.getItem('dismissedBannerIds') ?? '[]').includes(b.id) : true) ) as banner}
+								<Banner
+									{banner}
+									on:dismiss={(e) => {
+										const bannerId = e.detail;
 
-								localStorage.setItem(
-									'dismissedBannerIds',
-									JSON.stringify(
-										[
-											bannerId,
-											...JSON.parse(localStorage.getItem('dismissedBannerIds') ?? '[]')
-										].filter((id) => $banners.find((b) => b.id === id))
-									)
-								);
+										localStorage.setItem(
+											'dismissedBannerIds',
+											JSON.stringify(
+												[
+													bannerId,
+													...JSON.parse(localStorage.getItem('dismissedBannerIds') ?? '[]')
+												].filter((id) => $banners.find((b) => b.id === id))
+											)
+										);
+									}}
+								/>
+							{/each}
+						</div>
+					</div>
+				{/if}
+
+				<div class="flex flex-col flex-auto z-10 w-full">
+					<div
+						class=" pb-2.5 flex flex-col justify-between w-full flex-auto overflow-auto h-0 max-w-full z-10 scrollbar-hidden"
+						id="messages-container"
+						bind:this={messagesContainerElement}
+						on:scroll={(e) => {
+							autoScroll =
+								messagesContainerElement.scrollHeight - messagesContainerElement.scrollTop <=
+								messagesContainerElement.clientHeight + 5;
+						}}
+					>
+						<div class=" h-full w-full flex flex-col {chatIdProp ? 'py-4' : 'pt-2 pb-4'}">
+							<Messages
+								chatId={$chatId}
+								{selectedModels}
+								{processing}
+								bind:history
+								bind:messages
+								bind:autoScroll
+								bind:prompt
+								bottomPadding={files.length > 0}
+								{sendPrompt}
+								{continueGeneration}
+								{regenerateResponse}
+								{mergeResponses}
+								{chatActionHandler}
+								{showMessage}
+							/>
+						</div>
+					</div>
+
+					<div class="">
+						<MessageInput
+							bind:files
+							bind:prompt
+							bind:autoScroll
+							bind:selectedToolIds
+							bind:webSearchEnabled
+							bind:atSelectedModel
+							availableToolIds={selectedModelIds.reduce((a, e, i, arr) => {
+								const model = $models.find((m) => m.id === e);
+								if (model?.info?.meta?.toolIds ?? false) {
+									return [...new Set([...a, ...model.info.meta.toolIds])];
+								}
+								return a;
+							}, [])}
+							transparentBackground={$settings?.backgroundImageUrl ?? false}
+							{selectedModels}
+							{messages}
+							{submitPrompt}
+							{stopResponse}
+							on:call={async () => {
+								await showControls.set(true);
 							}}
 						/>
-					{/each}
+					</div>
 				</div>
-			</div>
-		{/if}
+			</Pane>
 
-		<div class="flex flex-col flex-auto z-10">
-			<div
-				class=" pb-2.5 flex flex-col justify-between w-full flex-auto overflow-auto h-0 max-w-full z-10 scrollbar-hidden {showControls
-					? 'lg:pr-[24rem]'
-					: ''}"
-				id="messages-container"
-				bind:this={messagesContainerElement}
-				on:scroll={(e) => {
-					autoScroll =
-						messagesContainerElement.scrollHeight - messagesContainerElement.scrollTop <=
-						messagesContainerElement.clientHeight + 5;
-				}}
-			>
-				<div class=" h-full w-full flex flex-col {chatIdProp ? 'py-4' : 'pt-2 pb-4'}">
-					<Messages
-						chatId={$chatId}
-						{selectedModels}
-						{processing}
-						bind:history
-						bind:messages
-						bind:autoScroll
-						bind:prompt
-						bottomPadding={files.length > 0}
-						{sendPrompt}
-						{continueGeneration}
-						{regenerateResponse}
-						{mergeResponses}
-						{chatActionHandler}
-					/>
-				</div>
-			</div>
-
-			<div class={showControls ? 'lg:pr-[24rem]' : ''}>
-				<MessageInput
-					bind:files
-					bind:prompt
-					bind:autoScroll
-					bind:selectedToolIds
-					bind:webSearchEnabled
-					bind:atSelectedModel
-					availableToolIds={selectedModelIds.reduce((a, e, i, arr) => {
-						const model = $models.find((m) => m.id === e);
-						if (model?.info?.meta?.toolIds ?? false) {
-							return [...new Set([...a, ...model.info.meta.toolIds])];
-						}
-						return a;
-					}, [])}
-					transparentBackground={$settings?.backgroundImageUrl ?? false}
-					{selectedModels}
-					{messages}
-					{submitPrompt}
-					{stopResponse}
-					on:call={() => {
-						showControls = true;
-					}}
-				/>
-			</div>
-		</div>
+			<ChatControls
+				models={selectedModelIds.reduce((a, e, i, arr) => {
+					const model = $models.find((m) => m.id === e);
+					if (model) {
+						return [...a, model];
+					}
+					return a;
+				}, [])}
+				bind:history
+				bind:chatFiles
+				bind:params
+				bind:files
+				bind:pane={controlPane}
+				{submitPrompt}
+				{stopResponse}
+				{showMessage}
+				modelId={selectedModelIds?.at(0) ?? null}
+				chatId={$chatId}
+				{eventTarget}
+			/>
+		</PaneGroup>
 	</div>
 {/if}
-
-<ChatControls
-	models={selectedModelIds.reduce((a, e, i, arr) => {
-		const model = $models.find((m) => m.id === e);
-		if (model) {
-			return [...a, model];
-		}
-		return a;
-	}, [])}
-	bind:show={showControls}
-	bind:chatFiles
-	bind:params
-	bind:files
-	{submitPrompt}
-	{stopResponse}
-	modelId={selectedModelIds?.at(0) ?? null}
-	chatId={$chatId}
-	{eventTarget}
-/>

src/lib/components/chat/ChatControls.svelte

@@ -1,13 +1,20 @@
 <script lang="ts">
+	import { SvelteFlowProvider } from '@xyflow/svelte';
 	import { slide } from 'svelte/transition';
+
+	import { onDestroy, onMount } from 'svelte';
+	import { mobile, showControls, showCallOverlay, showOverview } from '$lib/stores';
+
 	import Modal from '../common/Modal.svelte';
 	import Controls from './Controls/Controls.svelte';
-	import { onMount } from 'svelte';
-	import { mobile, showCallOverlay } from '$lib/stores';
 	import CallOverlay from './MessageInput/CallOverlay.svelte';
+	import Drawer from '../common/Drawer.svelte';
+	import Overview from './Overview.svelte';
+	import { Pane, PaneResizer } from 'paneforge';
+	import EllipsisVertical from '../icons/EllipsisVertical.svelte';
+	import { get } from 'svelte/store';
 
-	export let show = false;
-
+	export let history;
 	export let models = [];
 
 	export let chatId = null;
@@ -17,10 +24,13 @@
 	export let eventTarget: EventTarget;
 	export let submitPrompt: Function;
 	export let stopResponse: Function;
+	export let showMessage: Function;
 	export let files;
 	export let modelId;
 
+	export let pane;
 	let largeScreen = false;
+
 	onMount(() => {
 		// listen to resize 1024px
 		const mediaQuery = window.matchMedia('(min-width: 1024px)');
@@ -30,6 +40,7 @@
 				largeScreen = true;
 			} else {
 				largeScreen = false;
+				pane = null;
 			}
 		};
 
@@ -41,28 +52,58 @@
 			mediaQuery.removeEventListener('change', handleMediaQuery);
 		};
 	});
+
+	onDestroy(() => {
+		showControls.set(false);
+	});
+
+	$: if (!chatId) {
+		showOverview.set(false);
+	}
 </script>
 
-{#if largeScreen}
-	{#if show}
-		<div class=" absolute bottom-0 right-0 z-20 h-full pointer-events-none">
-			<div class="pr-4 pt-14 pb-8 w-[24rem] h-full" in:slide={{ duration: 200, axis: 'x' }}>
+<SvelteFlowProvider>
+	{#if !largeScreen}
+		{#if $showControls}
+			<Drawer
+				show={$showControls}
+				on:close={() => {
+					showControls.set(false);
+				}}
+			>
 				<div
-					class="w-full h-full px-5 py-4 bg-white dark:shadow-lg dark:bg-gray-850 border border-gray-50 dark:border-gray-800 rounded-xl z-50 pointer-events-auto overflow-y-auto scrollbar-hidden"
+					class=" {$showCallOverlay || $showOverview ? ' h-screen  w-screen' : 'px-6 py-4'} h-full"
 				>
 					{#if $showCallOverlay}
-						<CallOverlay
-							bind:files
-							{submitPrompt}
-							{stopResponse}
-							{modelId}
-							{chatId}
-							{eventTarget}
+						<div
+							class=" h-full max-h-[100dvh] bg-white text-gray-700 dark:bg-black dark:text-gray-300 flex justify-center"
+						>
+							<CallOverlay
+								bind:files
+								{submitPrompt}
+								{stopResponse}
+								{modelId}
+								{chatId}
+								{eventTarget}
+								on:close={() => {
+									showControls.set(false);
+								}}
+							/>
+						</div>
+					{:else if $showOverview}
+						<Overview
+							{history}
+							on:nodeclick={(e) => {
+								showMessage(e.detail.node.data.message);
+							}}
+							on:close={() => {
+								showControls.set(false);
+							}}
 						/>
 					{:else}
 						<Controls
 							on:close={() => {
-								show = false;
+								showControls.set(false);
 							}}
 							{models}
 							bind:chatFiles
@@ -70,38 +111,77 @@
 						/>
 					{/if}
 				</div>
+			</Drawer>
+		{/if}
+	{:else}
+		<!-- if $showControls -->
+		<PaneResizer class="relative flex w-2 items-center justify-center bg-background">
+			<div class="z-10 flex h-7 w-5 items-center justify-center rounded-sm">
+				<EllipsisVertical />
 			</div>
-		</div>
-	{/if}
-{:else if $showCallOverlay}
-	<div class=" absolute w-full h-screen max-h-[100dvh] flex z-[999] overflow-hidden">
-		<div
-			class="absolute w-full h-screen max-h-[100dvh] bg-white text-gray-700 dark:bg-black dark:text-gray-300 flex justify-center"
+		</PaneResizer>
+		<Pane
+			bind:pane
+			defaultSize={$showControls
+				? parseInt(localStorage.getItem('chat-controls-size') || '35')
+					? parseInt(localStorage.getItem('chat-controls-size') || '35')
+					: 35
+				: 0}
+			onResize={(size) => {
+				if (size === 0) {
+					showControls.set(false);
+				} else {
+					if (!$showControls) {
+						showControls.set(true);
+					}
+					localStorage.setItem('chat-controls-size', size);
+				}
+			}}
 		>
-			<CallOverlay
-				bind:files
-				{submitPrompt}
-				{stopResponse}
-				{modelId}
-				{chatId}
-				{eventTarget}
-				on:close={() => {
-					show = false;
-				}}
-			/>
-		</div>
-	</div>
-{:else}
-	<Modal bind:show>
-		<div class="  px-6 py-4 h-full">
-			<Controls
-				on:close={() => {
-					show = false;
-				}}
-				{models}
-				bind:chatFiles
-				bind:params
-			/>
-		</div>
-	</Modal>
-{/if}
+			{#if $showControls}
+				<div class="pr-4 pb-8 flex max-h-full min-h-full">
+					<div
+						class="w-full {$showOverview && !$showCallOverlay
+							? ' '
+							: 'px-5 py-4 bg-white dark:shadow-lg dark:bg-gray-850  border border-gray-50 dark:border-gray-800'}  rounded-lg z-50 pointer-events-auto overflow-y-auto scrollbar-hidden"
+					>
+						{#if $showCallOverlay}
+							<div class="w-full h-full flex justify-center">
+								<CallOverlay
+									bind:files
+									{submitPrompt}
+									{stopResponse}
+									{modelId}
+									{chatId}
+									{eventTarget}
+									on:close={() => {
+										showControls.set(false);
+									}}
+								/>
+							</div>
+						{:else if $showOverview}
+							<Overview
+								{history}
+								on:nodeclick={(e) => {
+									showMessage(e.detail.node.data.message);
+								}}
+								on:close={() => {
+									showControls.set(false);
+								}}
+							/>
+						{:else}
+							<Controls
+								on:close={() => {
+									showControls.set(false);
+								}}
+								{models}
+								bind:chatFiles
+								bind:params
+							/>
+						{/if}
+					</div>
+				</div>
+			{/if}
+		</Pane>
+	{/if}
+</SvelteFlowProvider>

src/lib/components/chat/MessageInput.svelte

@@ -93,20 +93,6 @@
 	const uploadFileHandler = async (file) => {
 		console.log(file);
 
-		// Check if the file is an audio file and transcribe/convert it to text file
-		if (['audio/mpeg', 'audio/wav'].includes(file['type'])) {
-			const res = await transcribeAudio(localStorage.token, file).catch((error) => {
-				toast.error(error);
-				return null;
-			});
-
-			if (res) {
-				console.log(res);
-				const blob = new Blob([res.text], { type: 'text/plain' });
-				file = blobToFile(blob, `${file.name}.txt`);
-			}
-		}
-
 		const fileItem = {
 			type: 'file',
 			file: '',
@@ -120,6 +106,23 @@
 		};
 		files = [...files, fileItem];
 
+		// Check if the file is an audio file and transcribe/convert it to text file
+		if (['audio/mpeg', 'audio/wav', 'audio/ogg'].includes(file['type'])) {
+			const res = await transcribeAudio(localStorage.token, file).catch((error) => {
+				toast.error(error);
+				return null;
+			});
+
+			if (res) {
+				console.log(res);
+				const blob = new Blob([res.text], { type: 'text/plain' });
+				file = blobToFile(blob, `${file.name}.txt`);
+
+				fileItem.name = file.name;
+				fileItem.size = file.size;
+			}
+		}
+
 		try {
 			const uploadedFile = await uploadFile(localStorage.token, file);
 
@@ -349,7 +352,7 @@
 	</div>
 
 	<div class="{transparentBackground ? 'bg-transparent' : 'bg-white dark:bg-gray-900'} ">
-		<div class="max-w-6xl px-2.5 md:px-6 mx-auto inset-x-0">
+		<div class="max-w-6xl px-2.5 md:px-6 mx-auto inset-x-0 pb-safe-bottom">
 			<div class=" pb-2">
 				<input
 					bind:this={filesInputElement}

src/lib/components/chat/MessageInput/CallOverlay.svelte

@@ -1,15 +1,13 @@
 <script lang="ts">
 	import { config, models, settings, showCallOverlay } from '$lib/stores';
 	import { onMount, tick, getContext, onDestroy, createEventDispatcher } from 'svelte';
+	import { DropdownMenu } from 'bits-ui';
+	import Dropdown from '$lib/components/common/Dropdown.svelte';
+	import { flyAndScale } from '$lib/utils/transitions';
 
 	const dispatch = createEventDispatcher();
 
-	import {
-		blobToFile,
-		calculateSHA256,
-		extractSentencesForAudio,
-		findWordIndices
-	} from '$lib/utils';
+	import { blobToFile } from '$lib/utils';
 	import { generateEmoji } from '$lib/apis';
 	import { synthesizeOpenAISpeech, transcribeAudio } from '$lib/apis/audio';
 
@@ -222,7 +220,9 @@
 	};
 
 	const startRecording = async () => {
-		audioStream = await navigator.mediaDevices.getUserMedia({ audio: true });
+		if (!audioStream) {
+			audioStream = await navigator.mediaDevices.getUserMedia({ audio: true });
+		}
 		mediaRecorder = new MediaRecorder(audioStream);
 
 		mediaRecorder.onstart = () => {
@@ -238,7 +238,7 @@
 		};
 
 		mediaRecorder.onstop = (e) => {
-			console.log('Recording stopped', e);
+			console.log('Recording stopped', audioStream, e);
 			stopRecordingCallback();
 		};
 
@@ -246,10 +246,11 @@
 	};
 
 	const stopAudioStream = async () => {
-		if (audioStream) {
-			const tracks = audioStream.getTracks();
-			tracks.forEach((track) => track.stop());
-		}
+		if (!audioStream) return;
+
+		audioStream.getAudioTracks().forEach(function (track) {
+			track.stop();
+		});
 
 		audioStream = null;
 	};
@@ -360,6 +361,7 @@
 								?.at(0) ?? undefined;
 
 						currentUtterance = new SpeechSynthesisUtterance(content);
+						currentUtterance.rate = $settings.audio?.tts?.speedRate ?? 1;
 
 						if (voice) {
 							currentUtterance.voice = voice;
@@ -381,11 +383,12 @@
 	const playAudio = (audio) => {
 		if ($showCallOverlay) {
 			return new Promise((resolve) => {
-				const audioElement = document.getElementById('audioElement');
+				const audioElement = document.getElementById('audioElement') as HTMLAudioElement;
 
 				if (audioElement) {
 					audioElement.src = audio.src;
 					audioElement.muted = true;
+					audioElement.playbackRate = $settings.audio?.tts?.speedRate ?? 1;
 
 					audioElement
 						.play()
@@ -525,6 +528,60 @@
 		console.log(`Audio monitoring and playing stopped for message ID ${id}`);
 	};
 
+	const chatStartHandler = async (e) => {
+		const { id } = e.detail;
+
+		chatStreaming = true;
+
+		if (currentMessageId !== id) {
+			console.log(`Received chat start event for message ID ${id}`);
+
+			currentMessageId = id;
+			if (audioAbortController) {
+				audioAbortController.abort();
+			}
+			audioAbortController = new AbortController();
+
+			assistantSpeaking = true;
+			// Start monitoring and playing audio for the message ID
+			monitorAndPlayAudio(id, audioAbortController.signal);
+		}
+	};
+
+	const chatEventHandler = async (e) => {
+		const { id, content } = e.detail;
+		// "id" here is message id
+		// if "id" is not the same as "currentMessageId" then do not process
+		// "content" here is a sentence from the assistant,
+		// there will be many sentences for the same "id"
+
+		if (currentMessageId === id) {
+			console.log(`Received chat event for message ID ${id}: ${content}`);
+
+			try {
+				if (messages[id] === undefined) {
+					messages[id] = [content];
+				} else {
+					messages[id].push(content);
+				}
+
+				console.log(content);
+
+				fetchAudio(content);
+			} catch (error) {
+				console.error('Failed to fetch or play audio:', error);
+			}
+		}
+	};
+
+	const chatFinishHandler = async (e) => {
+		const { id, content } = e.detail;
+		// "content" here is the entire message from the assistant
+		finishedMessages[id] = true;
+
+		chatStreaming = false;
+	};
+
 	onMount(async () => {
 		const setWakeLock = async () => {
 			try {
@@ -558,65 +615,15 @@
 
 		startRecording();
 
-		const chatStartHandler = async (e) => {
-			const { id } = e.detail;
-
-			chatStreaming = true;
-
-			if (currentMessageId !== id) {
-				console.log(`Received chat start event for message ID ${id}`);
-
-				currentMessageId = id;
-				if (audioAbortController) {
-					audioAbortController.abort();
-				}
-				audioAbortController = new AbortController();
-
-				assistantSpeaking = true;
-				// Start monitoring and playing audio for the message ID
-				monitorAndPlayAudio(id, audioAbortController.signal);
-			}
-		};
-
-		const chatEventHandler = async (e) => {
-			const { id, content } = e.detail;
-			// "id" here is message id
-			// if "id" is not the same as "currentMessageId" then do not process
-			// "content" here is a sentence from the assistant,
-			// there will be many sentences for the same "id"
-
-			if (currentMessageId === id) {
-				console.log(`Received chat event for message ID ${id}: ${content}`);
-
-				try {
-					if (messages[id] === undefined) {
-						messages[id] = [content];
-					} else {
-						messages[id].push(content);
-					}
-
-					console.log(content);
-
-					fetchAudio(content);
-				} catch (error) {
-					console.error('Failed to fetch or play audio:', error);
-				}
-			}
-		};
-
-		const chatFinishHandler = async (e) => {
-			const { id, content } = e.detail;
-			// "content" here is the entire message from the assistant
-			finishedMessages[id] = true;
-
-			chatStreaming = false;
-		};
-
 		eventTarget.addEventListener('chat:start', chatStartHandler);
 		eventTarget.addEventListener('chat', chatEventHandler);
 		eventTarget.addEventListener('chat:finish', chatFinishHandler);
 
 		return async () => {
+			await stopAllAudio();
+
+			stopAudioStream();
+
 			eventTarget.removeEventListener('chat:start', chatStartHandler);
 			eventTarget.removeEventListener('chat', chatEventHandler);
 			eventTarget.removeEventListener('chat:finish', chatFinishHandler);
@@ -633,6 +640,17 @@
 
 	onDestroy(async () => {
 		await stopAllAudio();
+		stopAudioStream();
+
+		eventTarget.removeEventListener('chat:start', chatStartHandler);
+		eventTarget.removeEventListener('chat', chatEventHandler);
+		eventTarget.removeEventListener('chat:finish', chatFinishHandler);
+
+		audioAbortController.abort();
+		await tick();
+
+		await stopAllAudio();
+
 		await stopRecordingCallback(false);
 		await stopCamera();
 	});
@@ -922,8 +940,12 @@
 				<button
 					class=" p-3 rounded-full bg-gray-50 dark:bg-gray-900"
 					on:click={async () => {
-						stopAudioStream();
-						stopVideoStream();
+						await stopAudioStream();
+						await stopVideoStream();
+
+						console.log(audioStream);
+						console.log(cameraStream);
+
 						showCallOverlay.set(false);
 						dispatch('close');
 					}}

src/lib/components/chat/MessageInput/Suggestions.svelte

@@ -9,7 +9,7 @@
 
 	let prompts = [];
 
-	$: prompts = suggestionPrompts
+	$: prompts = (suggestionPrompts ?? [])
 		.reduce((acc, current) => [...acc, ...[current]], [])
 		.sort(() => Math.random() - 0.5);
 	// suggestionPrompts.length <= 4

src/lib/components/chat/MessageInput/VoiceRecording.svelte

@@ -44,6 +44,7 @@
 		return `${minutes}:${formattedSeconds}`;
 	};
 
+	let stream;
 	let speechRecognition;
 
 	let mediaRecorder;
@@ -159,7 +160,7 @@
 	const startRecording = async () => {
 		startDurationCounter();
 
-		const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
+		stream = await navigator.mediaDevices.getUserMedia({ audio: true });
 		mediaRecorder = new MediaRecorder(stream);
 		mediaRecorder.onstart = () => {
 			console.log('Recording started');
@@ -251,6 +252,13 @@
 		}
 		stopDurationCounter();
 		audioChunks = [];
+
+		if (stream) {
+			const tracks = stream.getTracks();
+			tracks.forEach((track) => track.stop());
+		}
+
+		stream = null;
 	};
 
 	const confirmRecording = async () => {

src/lib/components/chat/Messages.svelte

@@ -21,6 +21,7 @@
 	export let regenerateResponse: Function;
 	export let mergeResponses: Function;
 	export let chatActionHandler: Function;
+	export let showMessage: Function = () => {};
 
 	export let user = $_user;
 	export let prompt;
@@ -108,6 +109,33 @@
 		await updateChatMessages();
 	};
 
+	const saveNewResponseMessage = async (message, content) => {
+		const responseMessageId = uuidv4();
+		const parentId = message.parentId;
+
+		const responseMessage = {
+			...message,
+			id: responseMessageId,
+			parentId: parentId,
+			childrenIds: [],
+			content: content,
+			timestamp: Math.floor(Date.now() / 1000) // Unix epoch
+		};
+
+		history.messages[responseMessageId] = responseMessage;
+		history.currentId = responseMessageId;
+
+		// Append messageId to childrenIds of parent message
+		if (parentId !== null) {
+			history.messages[parentId].childrenIds = [
+				...history.messages[parentId].childrenIds,
+				responseMessageId
+			];
+		}
+
+		await updateChatMessages();
+	};
+
 	const rateMessage = async (messageId, rating) => {
 		history.messages[messageId].annotation = {
 			...history.messages[messageId].annotation,
@@ -217,49 +245,39 @@
 
 	const deleteMessageHandler = async (messageId) => {
 		const messageToDelete = history.messages[messageId];
-
 		const parentMessageId = messageToDelete.parentId;
 		const childMessageIds = messageToDelete.childrenIds ?? [];
 
-		const hasDescendantMessages = childMessageIds.some(
-			(childId) => history.messages[childId]?.childrenIds?.length > 0
+		// Collect all grandchildren
+		const grandchildrenIds = childMessageIds.flatMap(
+			(childId) => history.messages[childId]?.childrenIds ?? []
 		);
 
-		history.currentId = parentMessageId;
-		await tick();
+		// Update parent's children
+		if (parentMessageId && history.messages[parentMessageId]) {
+			history.messages[parentMessageId].childrenIds = [
+				...history.messages[parentMessageId].childrenIds.filter((id) => id !== messageId),
+				...grandchildrenIds
+			];
+		}
 
-		// Remove the message itself from the parent message's children array
-		history.messages[parentMessageId].childrenIds = history.messages[
-			parentMessageId
-		].childrenIds.filter((id) => id !== messageId);
-
-		await tick();
-
-		childMessageIds.forEach((childId) => {
-			const childMessage = history.messages[childId];
-
-			if (childMessage && childMessage.childrenIds) {
-				if (childMessage.childrenIds.length === 0 && !hasDescendantMessages) {
-					// If there are no other responses/prompts
-					history.messages[parentMessageId].childrenIds = [];
-				} else {
-					childMessage.childrenIds.forEach((grandChildId) => {
-						if (history.messages[grandChildId]) {
-							history.messages[grandChildId].parentId = parentMessageId;
-							history.messages[parentMessageId].childrenIds.push(grandChildId);
-						}
-					});
-				}
+		// Update grandchildren's parent
+		grandchildrenIds.forEach((grandchildId) => {
+			if (history.messages[grandchildId]) {
+				history.messages[grandchildId].parentId = parentMessageId;
 			}
+		});
 
-			// Remove child message id from the parent message's children array
-			history.messages[parentMessageId].childrenIds = history.messages[
-				parentMessageId
-			].childrenIds.filter((id) => id !== childId);
+		// Delete the message and its children
+		[messageId, ...childMessageIds].forEach((id) => {
+			delete history.messages[id];
 		});
 
 		await tick();
 
+		showMessage({ id: parentMessageId });
+
+		// Update the chat
 		await updateChatById(localStorage.token, chatId, {
 			messages: messages,
 			history: history
@@ -309,7 +327,7 @@
 	{:else}
 		<div class="w-full pt-2">
 			{#key chatId}
-				{#each messages as message, messageIdx}
+				{#each messages as message, messageIdx (message.id)}
 					<div class=" w-full {messageIdx === messages.length - 1 ? ' pb-12' : ''}">
 						<div
 							class="flex flex-col justify-between px-5 mb-3 {($settings?.widescreenMode ?? null)
@@ -334,7 +352,7 @@
 									copyToClipboard={copyToClipboardWithToast}
 								/>
 							{:else if (history.messages[message.parentId]?.models?.length ?? 1) === 1}
-								{#key message.id && history.currentId}
+								{#key message.id}
 									<ResponseMessage
 										{message}
 										siblings={history.messages[message.parentId]?.childrenIds ?? []}
@@ -342,6 +360,7 @@
 										{readOnly}
 										{updateChatMessages}
 										{confirmEditResponseMessage}
+										{saveNewResponseMessage}
 										{showPreviousMessage}
 										{showNextMessage}
 										{rateMessage}

src/lib/components/chat/Messages/Citations.svelte

@@ -3,14 +3,9 @@
 
 	export let citations = [];
 
-	let showCitationModal = false;
-	let selectedCitation = null;
-</script>
+	let _citations = [];
 
-<CitationsModal bind:show={showCitationModal} citation={selectedCitation} />
-
-<div class="mt-1 mb-2 w-full flex gap-1 items-center flex-wrap">
-	{#each citations.reduce((acc, citation) => {
+	$: _citations = citations.reduce((acc, citation) => {
 		citation.document.forEach((document, index) => {
 			const metadata = citation.metadata?.[index];
 			const id = metadata?.source ?? 'N/A';
@@ -31,26 +26,42 @@
 				existingSource.document.push(document);
 				existingSource.metadata.push(metadata);
 			} else {
-				acc.push( { id: id, source: source, document: [document], metadata: metadata ? [metadata] : [] } );
+				acc.push({
+					id: id,
+					source: source,
+					document: [document],
+					metadata: metadata ? [metadata] : []
+				});
 			}
 		});
 		return acc;
-	}, []) as citation, idx}
-		<div class="flex gap-1 text-xs font-semibold">
-			<button
-				class="flex dark:text-gray-300 py-1 px-1 bg-gray-50 hover:bg-gray-100 dark:bg-gray-850 dark:hover:bg-gray-800 transition rounded-xl"
-				on:click={() => {
-					showCitationModal = true;
-					selectedCitation = citation;
-				}}
-			>
-				<div class="bg-white dark:bg-gray-700 rounded-full size-4">
-					{idx + 1}
-				</div>
-				<div class="flex-1 mx-2 line-clamp-1">
-					{citation.source.name}
-				</div>
-			</button>
-		</div>
-	{/each}
-</div>
+	}, []);
+
+	let showCitationModal = false;
+	let selectedCitation = null;
+</script>
+
+<CitationsModal bind:show={showCitationModal} citation={selectedCitation} />
+
+{#if _citations.length > 0}
+	<div class="mt-1 mb-2 w-full flex gap-1 items-center flex-wrap">
+		{#each _citations as citation, idx}
+			<div class="flex gap-1 text-xs font-semibold">
+				<button
+					class="flex dark:text-gray-300 py-1 px-1 bg-gray-50 hover:bg-gray-100 dark:bg-gray-850 dark:hover:bg-gray-800 transition rounded-xl max-w-96"
+					on:click={() => {
+						showCitationModal = true;
+						selectedCitation = citation;
+					}}
+				>
+					<div class="bg-white dark:bg-gray-700 rounded-full size-4">
+						{idx + 1}
+					</div>
+					<div class="flex-1 mx-2 line-clamp-1">
+						{citation.source.name}
+					</div>
+				</button>
+			</div>
+		{/each}
+	</div>
+{/if}

src/lib/components/chat/Messages/CitationsModal.svelte

@@ -66,6 +66,7 @@
 								>
 									{document?.metadata?.name ?? document.source.name}
 								</a>
+								{document?.metadata?.page ? `(page ${document.metadata.page + 1})` : ''}
 							</div>
 						{:else}
 							<div class="text-sm dark:text-gray-400">

src/lib/components/chat/Messages/Markdown/MarkdownInlineTokens.svelte

@@ -30,7 +30,13 @@
 			{token.text}
 		{/if}
 	{:else if token.type === 'link'}
-		<a href={token.href} target="_blank" rel="nofollow" title={token.title}>{token.text}</a>
+		{#if token.tokens}
+			<a href={token.href} target="_blank" rel="nofollow" title={token.title}>
+				<svelte:self id={`${id}-a`} tokens={token.tokens} />
+			</a>
+		{:else}
+			<a href={token.href} target="_blank" rel="nofollow" title={token.title}>{token.text}</a>
+		{/if}
 	{:else if token.type === 'image'}
 		<Image src={token.href} alt={token.text} />
 	{:else if token.type === 'strong'}

src/lib/components/chat/Messages/Markdown/MarkdownTokens.svelte

@@ -34,34 +34,36 @@
 			code={revertSanitizedResponseContent(token?.text ?? '')}
 		/>
 	{:else if token.type === 'table'}
-		<table>
-			<thead>
-				<tr>
-					{#each token.header as header, headerIdx}
-						<th style={token.align[headerIdx] ? '' : `text-align: ${token.align[headerIdx]}`}>
-							<MarkdownInlineTokens
-								id={`${id}-${tokenIdx}-header-${headerIdx}`}
-								tokens={header.tokens}
-							/>
-						</th>
-					{/each}
-				</tr>
-			</thead>
-			<tbody>
-				{#each token.rows as row, rowIdx}
+		<div class="scrollbar-hidden relative whitespace-nowrap overflow-x-auto max-w-full">
+			<table class="w-full">
+				<thead>
 					<tr>
-						{#each row ?? [] as cell, cellIdx}
-							<td style={token.align[cellIdx] ? '' : `text-align: ${token.align[cellIdx]}`}>
+						{#each token.header as header, headerIdx}
+							<th style={token.align[headerIdx] ? '' : `text-align: ${token.align[headerIdx]}`}>
 								<MarkdownInlineTokens
-									id={`${id}-${tokenIdx}-row-${rowIdx}-${cellIdx}`}
-									tokens={cell.tokens}
+									id={`${id}-${tokenIdx}-header-${headerIdx}`}
+									tokens={header.tokens}
 								/>
-							</td>
+							</th>
 						{/each}
 					</tr>
-				{/each}
-			</tbody>
-		</table>
+				</thead>
+				<tbody>
+					{#each token.rows as row, rowIdx}
+						<tr>
+							{#each row ?? [] as cell, cellIdx}
+								<td style={token.align[cellIdx] ? '' : `text-align: ${token.align[cellIdx]}`}>
+									<MarkdownInlineTokens
+										id={`${id}-${tokenIdx}-row-${rowIdx}-${cellIdx}`}
+										tokens={cell.tokens}
+									/>
+								</td>
+							{/each}
+						</tr>
+					{/each}
+				</tbody>
+			</table>
+		</div>
 	{:else if token.type === 'blockquote'}
 		<blockquote>
 			<svelte:self id={`${id}-${tokenIdx}`} tokens={token.tokens} />
@@ -142,7 +144,7 @@
 			/>
 		{/if}
 	{:else if token.type === 'space'}
-		<div class="my-0.5" />
+		<div class="my-2" />
 	{:else}
 		{console.log('Unknown token', token)}
 	{/if}

src/lib/components/chat/Messages/Placeholder.svelte

@@ -34,9 +34,9 @@
 </script>
 
 {#key mounted}
-	<div class="m-auto w-full max-w-6xl px-8 lg:px-20 pb-10">
+	<div class="m-auto w-full max-w-6xl px-8 lg:px-20">
 		<div class="flex justify-start">
-			<div class="flex -space-x-4 mb-1" in:fade={{ duration: 200 }}>
+			<div class="flex -space-x-4 mb-0.5" in:fade={{ duration: 200 }}>
 				{#each models as model, modelIdx}
 					<button
 						on:click={() => {
@@ -78,7 +78,7 @@
 		{/if}
 
 		<div
-			class=" mt-2 mb-4 text-3xl text-gray-800 dark:text-gray-100 font-semibold text-left flex items-center gap-4 font-primary"
+			class=" mt-2 mb-4 text-3xl text-gray-800 dark:text-gray-100 font-medium text-left flex items-center gap-4 font-primary"
 		>
 			<div>
 				<div class=" capitalize line-clamp-1" in:fade={{ duration: 200 }}>

src/lib/components/chat/Messages/ProfileImage.svelte

@@ -1,23 +1,11 @@
 <script lang="ts">
 	import { settings } from '$lib/stores';
-	import { WEBUI_BASE_URL } from '$lib/constants';
+	import ProfileImageBase from './ProfileImageBase.svelte';
 
 	export let className = 'size-8';
-
-	export let src = '/user.png';
+	export let src = '';
 </script>
 
 <div class={`flex-shrink-0 ${($settings?.chatDirection ?? 'LTR') === 'LTR' ? 'mr-3' : 'ml-3'}`}>
-	<img
-		crossorigin="anonymous"
-		src={src.startsWith(WEBUI_BASE_URL) ||
-		src.startsWith('https://www.gravatar.com/avatar/') ||
-		src.startsWith('data:') ||
-		src.startsWith('/')
-			? src
-			: `/user.png`}
-		class=" {className} object-cover rounded-full -translate-y-[1px]"
-		alt="profile"
-		draggable="false"
-	/>
+	<ProfileImageBase {src} {className} />
 </div>

src/lib/components/chat/Messages/ProfileImageBase.svelte

@@ -0,0 +1,21 @@
+<script lang="ts">
+	import { WEBUI_BASE_URL } from '$lib/constants';
+
+	export let className = 'size-8';
+	export let src = `${WEBUI_BASE_URL}/static/favicon.png`;
+</script>
+
+<img
+	crossorigin="anonymous"
+	src={src === ''
+		? `${WEBUI_BASE_URL}/static/favicon.png`
+		: src.startsWith(WEBUI_BASE_URL) ||
+			  src.startsWith('https://www.gravatar.com/avatar/') ||
+			  src.startsWith('data:') ||
+			  src.startsWith('/')
+			? src
+			: `/user.png`}
+	class=" {className} object-cover rounded-full -translate-y-[1px]"
+	alt="profile"
+	draggable="false"
+/>

src/lib/components/chat/Messages/ResponseMessage.svelte

@@ -85,6 +85,8 @@
 
 	export let updateChatMessages: Function;
 	export let confirmEditResponseMessage: Function;
+	export let saveNewResponseMessage: Function = () => {};
+
 	export let showPreviousMessage: Function;
 	export let showNextMessage: Function;
 	export let rateMessage: Function;
@@ -202,6 +204,8 @@
 					const blob = await res.blob();
 					const blobUrl = URL.createObjectURL(blob);
 					const audio = new Audio(blobUrl);
+					audio.playbackRate = $settings.audio?.tts?.speedRate ?? 1;
+
 					audioParts[idx] = audio;
 					loadingSpeech = false;
 					lastPlayedAudioPromise = lastPlayedAudioPromise.then(() => playAudio(idx));
@@ -224,6 +228,7 @@
 					console.log(voice);
 
 					const speak = new SpeechSynthesisUtterance(message.content);
+					speak.rate = $settings.audio?.tts?.speedRate ?? 1;
 
 					console.log(speak);
 
@@ -267,6 +272,15 @@
 		await tick();
 	};
 
+	const saveNewMessageHandler = async () => {
+		saveNewResponseMessage(message, editedContent);
+
+		edit = false;
+		editedContent = '';
+
+		await tick();
+	};
+
 	const cancelEditMessage = async () => {
 		edit = false;
 		editedContent = '';
@@ -333,7 +347,7 @@
 						{#each message.files as file}
 							<div>
 								{#if file.type === 'image'}
-									<Image src={file.url} />
+									<Image src={file.url} alt={message.content} />
 								{/if}
 							</div>
 						{/each}
@@ -356,7 +370,11 @@
 								{#if status?.action === 'web_search' && status?.urls}
 									<WebSearchResults {status}>
 										<div class="flex flex-col justify-center -space-y-0.5">
-											<div class="shimmer text-base line-clamp-1 text-wrap">
+											<div
+												class="{status?.done === false
+													? 'shimmer'
+													: ''} text-base line-clamp-1 text-wrap"
+											>
 												{status?.description}
 											</div>
 										</div>
@@ -364,7 +382,9 @@
 								{:else}
 									<div class="flex flex-col justify-center -space-y-0.5">
 										<div
-											class="shimmer text-gray-500 dark:text-gray-500 text-base line-clamp-1 text-wrap"
+											class="{status?.done === false
+												? 'shimmer'
+												: ''} text-gray-500 dark:text-gray-500 text-base line-clamp-1 text-wrap"
 										>
 											{status?.description}
 										</div>
@@ -393,31 +413,45 @@
 										const isEnterPressed = e.key === 'Enter';
 
 										if (isCmdOrCtrlPressed && isEnterPressed) {
-											document.getElementById('save-edit-message-button')?.click();
+											document.getElementById('confirm-edit-message-button')?.click();
 										}
 									}}
 								/>
 
-								<div class=" mt-2 mb-1 flex justify-end space-x-1.5 text-sm font-medium">
-									<button
-										id="close-edit-message-button"
-										class="px-4 py-2 bg-white hover:bg-gray-100 text-gray-800 transition rounded-3xl"
-										on:click={() => {
-											cancelEditMessage();
-										}}
-									>
-										{$i18n.t('Cancel')}
-									</button>
+								<div class=" mt-2 mb-1 flex justify-between text-sm font-medium">
+									<div>
+										<button
+											id="save-new-message-button"
+											class=" px-4 py-2 bg-gray-50 hover:bg-gray-100 dark:bg-gray-800 dark:hover:bg-gray-700 border dark:border-gray-700 text-gray-700 dark:text-gray-200 transition rounded-3xl"
+											on:click={() => {
+												saveNewMessageHandler();
+											}}
+										>
+											{$i18n.t('Save As Copy')}
+										</button>
+									</div>
 
-									<button
-										id="save-edit-message-button"
-										class=" px-4 py-2 bg-gray-900 hover:bg-gray-850 text-gray-100 transition rounded-3xl"
-										on:click={() => {
-											editMessageConfirmHandler();
-										}}
-									>
-										{$i18n.t('Save')}
-									</button>
+									<div class="flex space-x-1.5">
+										<button
+											id="close-edit-message-button"
+											class="px-4 py-2 bg-white dark:bg-gray-900 hover:bg-gray-100 text-gray-800 dark:text-gray-100 transition rounded-3xl"
+											on:click={() => {
+												cancelEditMessage();
+											}}
+										>
+											{$i18n.t('Cancel')}
+										</button>
+
+										<button
+											id="confirm-edit-message-button"
+											class=" px-4 py-2 bg-gray-900 dark:bg-white hover:bg-gray-850 text-gray-100 dark:text-gray-800 transition rounded-3xl"
+											on:click={() => {
+												editMessageConfirmHandler();
+											}}
+										>
+											{$i18n.t('Save')}
+										</button>
+									</div>
 								</div>
 							</div>
 						{:else}

src/lib/components/chat/Messages/Skeleton.svelte

@@ -1,4 +1,4 @@
-<div class="w-full mt-2 mb-4">
+<div class="w-full mt-2 mb-2">
 	<div class="animate-pulse flex w-full">
 		<div class="space-y-2 w-full">
 			<div class="h-2 bg-gray-200 dark:bg-gray-600 rounded mr-14" />

src/lib/components/chat/Messages/UserMessage.svelte

@@ -62,7 +62,7 @@
 	};
 </script>
 
-<div class=" flex w-full user-message" dir={$settings.chatDirection}>
+<div class=" flex w-full user-message" dir={$settings.chatDirection} id="message-{message.id}">
 	{#if !($settings?.chatBubble ?? true)}
 		<ProfileImage
 			src={message.user

src/lib/components/chat/ModelSelector.svelte

@@ -55,7 +55,9 @@
 			</div>
 
 			{#if selectedModelIdx === 0}
-				<div class="  self-center mr-2 disabled:text-gray-600 disabled:hover:text-gray-600">
+				<div
+					class="  self-center mx-1 disabled:text-gray-600 disabled:hover:text-gray-600 -translate-y-[0.5px]"
+				>
 					<Tooltip content={$i18n.t('Add Model')}>
 						<button
 							class=" "
@@ -79,7 +81,9 @@
 					</Tooltip>
 				</div>
 			{:else}
-				<div class="  self-center disabled:text-gray-600 disabled:hover:text-gray-600 mr-2">
+				<div
+					class="  self-center mx-1 disabled:text-gray-600 disabled:hover:text-gray-600 -translate-y-[0.5px]"
+				>
 					<Tooltip content={$i18n.t('Remove Model')}>
 						<button
 							{disabled}
@@ -95,7 +99,7 @@
 								viewBox="0 0 24 24"
 								stroke-width="2"
 								stroke="currentColor"
-								class="size-3.5"
+								class="size-3"
 							>
 								<path stroke-linecap="round" stroke-linejoin="round" d="M19.5 12h-15" />
 							</svg>
@@ -107,8 +111,8 @@
 	{/each}
 </div>
 
-{#if showSetDefault && !$mobile}
-	<div class="text-left mt-0.5 ml-1 text-[0.7rem] text-gray-500 font-primary">
+{#if showSetDefault}
+	<div class=" absolute text-left mt-[1px] ml-1 text-[0.7rem] text-gray-500 font-primary">
 		<button on:click={saveDefaultModel}> {$i18n.t('Set as default')}</button>
 	</div>
 {/if}

src/lib/components/chat/ModelSelector/Selector.svelte

@@ -231,7 +231,7 @@
 >
 	<DropdownMenu.Trigger class="relative w-full font-primary" aria-label={placeholder}>
 		<div
-			class="flex w-full text-left px-0.5 outline-none bg-transparent truncate text-lg font-semibold placeholder-gray-400 focus:outline-none"
+			class="flex w-full text-left px-0.5 outline-none bg-transparent truncate text-lg font-medium placeholder-gray-400 focus:outline-none"
 		>
 			{#if selectedModel}
 				{selectedModel.label}
@@ -302,7 +302,7 @@
 					>
 						<div class="flex flex-col">
 							{#if $mobile && (item?.model?.info?.meta?.tags ?? []).length > 0}
-								<div class="flex gap-0.5 self-start h-full mb-0.5 -translate-x-1">
+								<div class="flex gap-0.5 self-start h-full mb-1.5 -translate-x-1">
 									{#each item.model?.info?.meta.tags as tag}
 										<div
 											class=" text-xs font-bold px-1 rounded uppercase line-clamp-1 bg-gray-500/20 text-gray-700 dark:text-gray-200"
@@ -418,7 +418,7 @@
 						</div>
 
 						{#if value === item.value}
-							<div class="ml-auto pl-2">
+							<div class="ml-auto pl-2 pr-2 md:pr-0">
 								<Check />
 							</div>
 						{/if}

src/lib/components/chat/Overview.svelte

@@ -0,0 +1,185 @@
+<script lang="ts">
+	import { getContext, createEventDispatcher, onDestroy } from 'svelte';
+	import { useSvelteFlow, useNodesInitialized, useStore } from '@xyflow/svelte';
+
+	const dispatch = createEventDispatcher();
+	const i18n = getContext('i18n');
+
+	import { onMount, tick } from 'svelte';
+
+	import { writable } from 'svelte/store';
+	import { models, showOverview, theme, user } from '$lib/stores';
+
+	import '@xyflow/svelte/dist/style.css';
+
+	import CustomNode from './Overview/Node.svelte';
+	import Flow from './Overview/Flow.svelte';
+	import XMark from '../icons/XMark.svelte';
+
+	const { width, height } = useStore();
+
+	const { fitView, getViewport } = useSvelteFlow();
+	const nodesInitialized = useNodesInitialized();
+
+	export let history;
+
+	let selectedMessageId = null;
+
+	const nodes = writable([]);
+	const edges = writable([]);
+
+	const nodeTypes = {
+		custom: CustomNode
+	};
+
+	$: if (history) {
+		drawFlow();
+	}
+
+	$: if (history && history.currentId) {
+		focusNode();
+		selectedMessageId = null;
+	}
+
+	const focusNode = async () => {
+		if (selectedMessageId === null) {
+			await fitView({ nodes: [{ id: history.currentId }] });
+		}
+	};
+
+	const drawFlow = async () => {
+		const nodeList = [];
+		const edgeList = [];
+		const levelOffset = 150; // Vertical spacing between layers
+		const siblingOffset = 250; // Horizontal spacing between nodes at the same layer
+
+		// Map to keep track of node positions at each level
+		let positionMap = new Map();
+
+		// Helper function to truncate labels
+		function createLabel(content) {
+			const maxLength = 100;
+			return content.length > maxLength ? content.substr(0, maxLength) + '...' : content;
+		}
+
+		// Create nodes and map children to ensure alignment in width
+		let layerWidths = {}; // Track widths of each layer
+
+		Object.keys(history.messages).forEach((id) => {
+			const message = history.messages[id];
+			const level = message.parentId ? (positionMap.get(message.parentId)?.level ?? -1) + 1 : 0;
+			if (!layerWidths[level]) layerWidths[level] = 0;
+
+			positionMap.set(id, {
+				id: message.id,
+				level,
+				position: layerWidths[level]++
+			});
+		});
+
+		// Adjust positions based on siblings count to centralize vertical spacing
+		Object.keys(history.messages).forEach((id) => {
+			const pos = positionMap.get(id);
+			const xOffset = pos.position * siblingOffset;
+			const y = pos.level * levelOffset;
+			const x = xOffset;
+
+			nodeList.push({
+				id: pos.id,
+				type: 'custom',
+				data: {
+					user: $user,
+					message: history.messages[id],
+					model: $models.find((model) => model.id === history.messages[id].model)
+				},
+				position: { x, y }
+			});
+
+			// Create edges
+			const parentId = history.messages[id].parentId;
+			if (parentId) {
+				edgeList.push({
+					id: parentId + '-' + pos.id,
+					source: parentId,
+					target: pos.id,
+					selectable: false,
+					class: ' dark:fill-gray-300 fill-gray-300',
+					type: 'smoothstep',
+					animated: history.currentId === id || recurseCheckChild(id, history.currentId)
+				});
+			}
+		});
+
+		await edges.set([...edgeList]);
+		await nodes.set([...nodeList]);
+	};
+
+	const recurseCheckChild = (nodeId, currentId) => {
+		const node = history.messages[nodeId];
+		return (
+			node.childrenIds &&
+			node.childrenIds.some((id) => id === currentId || recurseCheckChild(id, currentId))
+		);
+	};
+
+	onMount(() => {
+		drawFlow();
+
+		nodesInitialized.subscribe(async (initialized) => {
+			if (initialized) {
+				await tick();
+				const res = await fitView({ nodes: [{ id: history.currentId }] });
+			}
+		});
+
+		width.subscribe((value) => {
+			if (value) {
+				// fitView();
+				fitView({ nodes: [{ id: history.currentId }] });
+			}
+		});
+
+		height.subscribe((value) => {
+			if (value) {
+				// fitView();
+				fitView({ nodes: [{ id: history.currentId }] });
+			}
+		});
+	});
+
+	onDestroy(() => {
+		console.log('Overview destroyed');
+
+		nodes.set([]);
+		edges.set([]);
+	});
+</script>
+
+<div class="w-full h-full relative">
+	<div class=" absolute z-50 w-full flex justify-between dark:text-gray-100 px-5 py-4">
+		<div class=" text-lg font-medium self-center font-primary">{$i18n.t('Chat Overview')}</div>
+		<button
+			class="self-center"
+			on:click={() => {
+				dispatch('close');
+				showOverview.set(false);
+			}}
+		>
+			<XMark className="size-4" />
+		</button>
+	</div>
+
+	{#if $nodes.length > 0}
+		<Flow
+			{nodes}
+			{nodeTypes}
+			{edges}
+			on:nodeclick={(e) => {
+				console.log(e.detail.node.data);
+				dispatch('nodeclick', e.detail);
+				selectedMessageId = e.detail.node.data.message.id;
+				fitView({ nodes: [{ id: selectedMessageId }] });
+			}}
+		/>
+	{/if}
+</div>

src/lib/components/chat/Overview/Flow.svelte

@@ -0,0 +1,36 @@
+<script>
+	import { createEventDispatcher } from 'svelte';
+
+	const dispatch = createEventDispatcher();
+
+	import { theme } from '$lib/stores';
+	import { Background, Controls, SvelteFlow, BackgroundVariant } from '@xyflow/svelte';
+
+	export let nodes;
+	export let nodeTypes;
+	export let edges;
+</script>
+
+<SvelteFlow
+	{nodes}
+	{nodeTypes}
+	{edges}
+	fitView
+	minZoom={0.001}
+	colorMode={$theme.includes('dark')
+		? 'dark'
+		: $theme === 'system'
+			? window.matchMedia('(prefers-color-scheme: dark)').matches
+				? 'dark'
+				: 'light'
+			: 'light'}
+	nodesConnectable={false}
+	nodesDraggable={false}
+	on:nodeclick={(e) => dispatch('nodeclick', e.detail)}
+	oninit={() => {
+		console.log('Flow initialized');
+	}}
+>
+	<Controls showLock={false} />
+	<Background variant={BackgroundVariant.Dots} />
+</SvelteFlow>

src/lib/components/chat/Overview/Node.svelte

@@ -0,0 +1,62 @@
+<script lang="ts">
+	import { WEBUI_BASE_URL } from '$lib/constants';
+	import { Handle, Position, type NodeProps } from '@xyflow/svelte';
+
+	import ProfileImageBase from '../Messages/ProfileImageBase.svelte';
+	import Tooltip from '$lib/components/common/Tooltip.svelte';
+
+	type $$Props = NodeProps;
+	export let data: $$Props['data'];
+</script>
+
+<div
+	class="px-4 py-3 shadow-md rounded-xl dark:bg-black bg-white border dark:border-gray-900 w-60 h-20"
+>
+	<Tooltip
+		content={data?.message?.error ? data.message.error.content : data.message.content}
+		class="w-full"
+	>
+		{#if data.message.role === 'user'}
+			<div class="flex w-full">
+				<ProfileImageBase
+					src={data.user?.profile_image_url ?? '/user.png'}
+					className={'size-5 -translate-y-[1px]'}
+				/>
+				<div class="ml-2">
+					<div class="text-xs text-black dark:text-white font-medium">
+						{data?.user?.name ?? 'User'}
+					</div>
+
+					{#if data?.message?.error}
+						<div class="text-red-500 line-clamp-2 text-xs mt-0.5">{data.message.error.content}</div>
+					{:else}
+						<div class="text-gray-500 line-clamp-2 text-xs mt-0.5">{data.message.content}</div>
+					{/if}
+				</div>
+			</div>
+		{:else}
+			<div class="flex w-full">
+				<ProfileImageBase
+					src={data?.model?.info?.meta?.profile_image_url ?? ''}
+					className={'size-5 -translate-y-[1px]'}
+				/>
+
+				<div class="ml-2">
+					<div class="text-xs text-black dark:text-white font-medium">
+						{data?.model?.name ?? data?.message?.model ?? 'Assistant'}
+					</div>
+
+					{#if data?.message?.error}
+						<div class="text-red-500 line-clamp-2 text-xs mt-0.5">
+							{data.message.error.content}
+						</div>
+					{:else}
+						<div class="text-gray-500 line-clamp-2 text-xs mt-0.5">{data.message.content}</div>
+					{/if}
+				</div>
+			</div>
+		{/if}
+	</Tooltip>
+	<Handle type="target" position={Position.Top} class="w-2 rounded-full dark:bg-gray-900" />
+	<Handle type="source" position={Position.Bottom} class="w-2 rounded-full dark:bg-gray-900" />
+</div>

src/lib/components/chat/Settings/Advanced/AdvancedParams.svelte

@@ -41,7 +41,7 @@
 	}
 </script>
 
-<div class=" space-y-1 text-xs">
+<div class=" space-y-1 text-xs pb-safe-bottom">
 	<div class=" py-0.5 w-full justify-between">
 		<div class="flex w-full justify-between">
 			<div class=" self-center text-xs font-medium">{$i18n.t('Seed')}</div>

src/lib/components/chat/Settings/Audio.svelte

@@ -23,6 +23,10 @@
 	let voices = [];
 	let voice = '';
 
+	// Audio speed control
+	let speechRate = 1;
+	const speedOptions = [2, 1.75, 1.5, 1.25, 1, 0.75, 0.5];
+
 	const getVoices = async () => {
 		if ($config.audio.tts.engine === '') {
 			const getVoicesLoop = setInterval(async () => {
@@ -56,6 +60,7 @@
 	};
 
 	onMount(async () => {
+		speechRate = $settings.audio?.tts?.speedRate ?? 1;
 		conversationMode = $settings.conversationMode ?? false;
 		speechAutoSend = $settings.speechAutoSend ?? false;
 		responseAutoPlayback = $settings.responseAutoPlayback ?? false;
@@ -83,6 +88,7 @@
 					engine: STTEngine !== '' ? STTEngine : undefined
 				},
 				tts: {
+					speedRate: speechRate,
 					voice: voice !== '' ? voice : undefined,
 					defaultVoice: $config?.audio?.tts?.voice ?? '',
 					nonLocalVoices: $config.audio.tts.engine === '' ? nonLocalVoices : undefined
@@ -153,6 +159,21 @@
 					{/if}
 				</button>
 			</div>
+
+			<div class=" py-0.5 flex w-full justify-between">
+				<div class=" self-center text-xs font-medium">{$i18n.t('Speed Rate')}</div>
+
+				<div class="flex items-center relative">
+					<select
+						class="dark:bg-gray-900 w-fit pr-8 rounded px-2 p-1 text-xs bg-transparent outline-none text-right"
+						bind:value={speechRate}
+					>
+						{#each speedOptions as option}
+							<option value={option} selected={speechRate === option}>{option}x</option>
+						{/each}
+					</select>
+				</div>
+			</div>
 		</div>
 
 		<hr class=" dark:border-gray-850" />

src/lib/components/chat/Settings/General.svelte

@@ -114,6 +114,29 @@
 			document.documentElement.classList.add(e);
 		});
 
+		const metaThemeColor = document.querySelector('meta[name="theme-color"]');
+		if (metaThemeColor) {
+			if (_theme.includes('system')) {
+				const systemTheme = window.matchMedia('(prefers-color-scheme: dark)').matches
+					? 'dark'
+					: 'light';
+				console.log('Setting system meta theme color: ' + systemTheme);
+				metaThemeColor.setAttribute('content', systemTheme === 'light' ? '#ffffff' : '#171717');
+			} else {
+				console.log('Setting meta theme color: ' + _theme);
+				metaThemeColor.setAttribute(
+					'content',
+					_theme === 'dark'
+						? '#171717'
+						: _theme === 'oled-dark'
+							? '#000000'
+							: _theme === 'her'
+								? '#983724'
+								: '#ffffff'
+				);
+			}
+		}
+
 		console.log(_theme);
 	};
 

src/lib/components/chat/Settings/Interface.svelte

@@ -36,11 +36,18 @@
 	let voiceInterruption = false;
 	let hapticFeedback = false;
 
+	let streamResponse = true;
+
 	const toggleSplitLargeChunks = async () => {
 		splitLargeChunks = !splitLargeChunks;
 		saveSettings({ splitLargeChunks: splitLargeChunks });
 	};
 
+	const toggleStreamResponse = async () => {
+		streamResponse = !streamResponse;
+		saveSettings({ streamResponse: streamResponse });
+	};
+
 	const togglesScrollOnBranchChange = async () => {
 		scrollOnBranchChange = !scrollOnBranchChange;
 		saveSettings({ scrollOnBranchChange: scrollOnBranchChange });
@@ -158,6 +165,7 @@
 		userLocation = $settings.userLocation ?? false;
 
 		hapticFeedback = $settings.hapticFeedback ?? false;
+		streamResponse = $settings?.streamResponse ?? true;
 
 		defaultModelId = $settings?.models?.at(0) ?? '';
 		if ($config?.default_models) {
@@ -311,6 +319,28 @@
 				</div>
 			</div>
 
+			<div>
+				<div class=" py-0.5 flex w-full justify-between">
+					<div class=" self-center text-xs">
+						{$i18n.t('Stream Chat Response')}
+					</div>
+
+					<button
+						class="p-1 px-3 text-xs flex rounded transition"
+						on:click={() => {
+							toggleStreamResponse();
+						}}
+						type="button"
+					>
+						{#if streamResponse === true}
+							<span class="ml-2 self-center">{$i18n.t('On')}</span>
+						{:else}
+							<span class="ml-2 self-center">{$i18n.t('Off')}</span>
+						{/if}
+					</button>
+				</div>
+			</div>
+
 			<div>
 				<div class=" py-0.5 flex w-full justify-between">
 					<div class=" self-center text-xs">

src/lib/components/common/ConfirmDialog.svelte

@@ -49,20 +49,20 @@
 	<!-- svelte-ignore a11y-no-static-element-interactions -->
 	<div
 		bind:this={modalElement}
-		class=" fixed top-0 right-0 left-0 bottom-0 bg-black/60 w-full min-h-screen h-screen flex justify-center z-[9999] overflow-hidden overscroll-contain"
+		class=" fixed top-0 right-0 left-0 bottom-0 bg-black/60 w-full h-screen max-h-[100dvh] flex justify-center z-[9999] overflow-hidden overscroll-contain"
 		in:fade={{ duration: 10 }}
 		on:mousedown={() => {
 			show = false;
 		}}
 	>
 		<div
-			class=" m-auto rounded-2xl max-w-full w-[32rem] mx-2 bg-gray-50 dark:bg-gray-950 shadow-3xl border border-gray-850"
+			class=" m-auto rounded-2xl max-w-full w-[32rem] mx-2 bg-gray-50 dark:bg-gray-950 max-h-[100dvh] shadow-3xl border border-gray-850"
 			in:flyAndScale
 			on:mousedown={(e) => {
 				e.stopPropagation();
 			}}
 		>
-			<div class="px-[1.75rem] py-6">
+			<div class="px-[1.75rem] py-6 flex flex-col">
 				<div class=" text-lg font-semibold dark:text-gray-200 mb-2.5">
 					{#if title !== ''}
 						{title}
@@ -72,7 +72,7 @@
 				</div>
 
 				<slot>
-					<div class=" text-sm text-gray-500">
+					<div class=" text-sm text-gray-500 flex-1">
 						{#if message !== ''}
 							{message}
 						{:else}

src/lib/components/common/Drawer.svelte

@@ -0,0 +1,103 @@
+<script lang="ts">
+	import { onDestroy, onMount, createEventDispatcher } from 'svelte';
+	import { flyAndScale } from '$lib/utils/transitions';
+	import { fade, fly, slide } from 'svelte/transition';
+
+	const dispatch = createEventDispatcher();
+
+	export let show = false;
+	export let size = 'md';
+
+	let modalElement = null;
+	let mounted = false;
+
+	const sizeToWidth = (size) => {
+		if (size === 'xs') {
+			return 'w-[16rem]';
+		} else if (size === 'sm') {
+			return 'w-[30rem]';
+		} else if (size === 'md') {
+			return 'w-[48rem]';
+		} else {
+			return 'w-[56rem]';
+		}
+	};
+
+	const handleKeyDown = (event: KeyboardEvent) => {
+		if (event.key === 'Escape' && isTopModal()) {
+			console.log('Escape');
+			show = false;
+		}
+	};
+
+	const isTopModal = () => {
+		const modals = document.getElementsByClassName('modal');
+		return modals.length && modals[modals.length - 1] === modalElement;
+	};
+
+	onMount(() => {
+		mounted = true;
+	});
+
+	$: if (show && modalElement) {
+		document.body.appendChild(modalElement);
+		window.addEventListener('keydown', handleKeyDown);
+		document.body.style.overflow = 'hidden';
+	} else if (modalElement) {
+		dispatch('close');
+		window.removeEventListener('keydown', handleKeyDown);
+
+		if (document.body.contains(modalElement)) {
+			document.body.removeChild(modalElement);
+			document.body.style.overflow = 'unset';
+		}
+	}
+
+	onDestroy(() => {
+		show = false;
+		if (modalElement) {
+			if (document.body.contains(modalElement)) {
+				document.body.removeChild(modalElement);
+				document.body.style.overflow = 'unset';
+			}
+		}
+	});
+</script>
+
+<!-- svelte-ignore a11y-click-events-have-key-events -->
+<!-- svelte-ignore a11y-no-static-element-interactions -->
+
+<div
+	bind:this={modalElement}
+	class="modal fixed right-0 left-0 bottom-0 bg-black/60 w-full h-screen max-h-[100dvh] flex justify-center z-[9999] overflow-hidden overscroll-contain"
+	in:fly={{ y: 100, duration: 100 }}
+	on:mousedown={() => {
+		show = false;
+	}}
+>
+	<div
+		class=" mt-auto max-w-full w-full bg-gray-50 dark:bg-gray-900 max-h-[100dvh] overflow-y-auto scrollbar-hidden"
+		on:mousedown={(e) => {
+			e.stopPropagation();
+		}}
+	>
+		<slot />
+	</div>
+</div>
+
+<style>
+	.modal-content {
+		animation: scaleUp 0.1s ease-out forwards;
+	}
+
+	@keyframes scaleUp {
+		from {
+			transform: scale(0.985);
+			opacity: 0;
+		}
+		to {
+			transform: scale(1);
+			opacity: 1;
+		}
+	}
+</style>

src/lib/components/common/ImagePreview.svelte

@@ -9,14 +9,14 @@
 
 	let previewElement = null;
 
-	const downloadImage = (url, filename) => {
+	const downloadImage = (url, filename, prefixName = '') => {
 		fetch(url)
 			.then((response) => response.blob())
 			.then((blob) => {
 				const objectUrl = window.URL.createObjectURL(blob);
 				const link = document.createElement('a');
 				link.href = objectUrl;
-				link.download = filename;
+				link.download = `${prefixName}${filename}`;
 				document.body.appendChild(link);
 				link.click();
 				document.body.removeChild(link);
@@ -87,7 +87,7 @@
 				<button
 					class=" p-5"
 					on:click={() => {
-						downloadImage(src, src.substring(src.lastIndexOf('/') + 1));
+						downloadImage(src, src.substring(src.lastIndexOf('/') + 1), alt);
 					}}
 				>
 					<svg

src/lib/components/common/Modal.svelte

@@ -61,7 +61,7 @@
 	<!-- svelte-ignore a11y-no-static-element-interactions -->
 	<div
 		bind:this={modalElement}
-		class="modal fixed top-0 right-0 left-0 bottom-0 bg-black/60 w-full min-h-screen h-screen flex justify-center z-[9999] overflow-hidden overscroll-contain"
+		class="modal fixed top-0 right-0 left-0 bottom-0 bg-black/60 w-full h-screen max-h-[100dvh] flex justify-center z-[9999] overflow-hidden overscroll-contain"
 		in:fade={{ duration: 10 }}
 		on:mousedown={() => {
 			show = false;

src/lib/components/icons/ArrowUpCircle.svelte

@@ -0,0 +1,19 @@
+<script lang="ts">
+	export let className = 'size-4';
+	export let strokeWidth = '1.5';
+</script>
+
+<svg
+	xmlns="http://www.w3.org/2000/svg"
+	fill="none"
+	viewBox="0 0 24 24"
+	stroke-width={strokeWidth}
+	stroke="currentColor"
+	class={className}
+>
+	<path
+		stroke-linecap="round"
+		stroke-linejoin="round"
+		d="m15 11.25-3-3m0 0-3 3m3-3v7.5M21 12a9 9 0 1 1-18 0 9 9 0 0 1 18 0Z"
+	/>
+</svg>

src/lib/components/icons/Clipboard.svelte

@@ -0,0 +1,19 @@
+<script lang="ts">
+	export let className = 'size-4';
+	export let strokeWidth = '2';
+</script>
+
+<svg
+	xmlns="http://www.w3.org/2000/svg"
+	fill="none"
+	viewBox="0 0 24 24"
+	stroke-width={strokeWidth}
+	stroke="currentColor"
+	class={className}
+>
+	<path
+		stroke-linecap="round"
+		stroke-linejoin="round"
+		d="M15.666 3.888A2.25 2.25 0 0 0 13.5 2.25h-3c-1.03 0-1.9.693-2.166 1.638m7.332 0c.055.194.084.4.084.612v0a.75.75 0 0 1-.75.75H9a.75.75 0 0 1-.75-.75v0c0-.212.03-.418.084-.612m7.332 0c.646.049 1.288.11 1.927.184 1.1.128 1.907 1.077 1.907 2.185V19.5a2.25 2.25 0 0 1-2.25 2.25H6.75A2.25 2.25 0 0 1 4.5 19.5V6.257c0-1.108.806-2.057 1.907-2.185a48.208 48.208 0 0 1 1.927-.184"
+	/>
+</svg>

src/lib/components/icons/Map.svelte

@@ -0,0 +1,19 @@
+<script lang="ts">
+	export let className = 'size-4';
+	export let strokeWidth = '2';
+</script>
+
+<svg
+	xmlns="http://www.w3.org/2000/svg"
+	fill="none"
+	viewBox="0 0 24 24"
+	stroke-width={strokeWidth}
+	stroke="currentColor"
+	class={className}
+>
+	<path
+		stroke-linecap="round"
+		stroke-linejoin="round"
+		d="M9 6.75V15m6-6v8.25m.503 3.498 4.875-2.437c.381-.19.622-.58.622-1.006V4.82c0-.836-.88-1.38-1.628-1.006l-3.869 1.934c-.317.159-.69.159-1.006 0L9.503 3.252a1.125 1.125 0 0 0-1.006 0L3.622 5.689C3.24 5.88 3 6.27 3 6.695V19.18c0 .836.88 1.38 1.628 1.006l3.869-1.934c.317-.159.69-.159 1.006 0l4.994 2.497c.317.158.69.158 1.006 0Z"
+	/>
+</svg>

src/lib/components/layout/Navbar.svelte

@@ -8,7 +8,7 @@
 		mobile,
 		settings,
 		showArchivedChats,
-		showSettings,
+		showControls,
 		showSidebar,
 		user
 	} from '$lib/stores';
@@ -22,6 +22,7 @@
 	import UserMenu from './Sidebar/UserMenu.svelte';
 	import MenuLines from '../icons/MenuLines.svelte';
 	import AdjustmentsHorizontal from '../icons/AdjustmentsHorizontal.svelte';
+	import Map from '../icons/Map.svelte';
 
 	const i18n = getContext('i18n');
 
@@ -31,9 +32,7 @@
 
 	export let chat;
 	export let selectedModels;
-
 	export let showModelSelector = true;
-	export let showControls = false;
 
 	let showShareChatModal = false;
 	let showDownloadChatModal = false;
@@ -83,7 +82,7 @@
 						}}
 					>
 						<button
-							class="hidden md:flex cursor-pointer px-2 py-2 rounded-xl hover:bg-gray-50 dark:hover:bg-gray-850 transition"
+							class="flex cursor-pointer px-2 py-2 rounded-xl hover:bg-gray-50 dark:hover:bg-gray-850 transition"
 							id="chat-context-menu-button"
 						>
 							<div class=" m-auto self-center">
@@ -106,19 +105,21 @@
 					</Menu>
 				{/if}
 
-				<Tooltip content={$i18n.t('Controls')}>
-					<button
-						class=" flex cursor-pointer px-2 py-2 rounded-xl hover:bg-gray-50 dark:hover:bg-gray-850 transition"
-						on:click={() => {
-							showControls = !showControls;
-						}}
-						aria-label="Controls"
-					>
-						<div class=" m-auto self-center">
-							<AdjustmentsHorizontal className=" size-5" strokeWidth="0.5" />
-						</div>
-					</button>
-				</Tooltip>
+				{#if !$mobile}
+					<Tooltip content={$i18n.t('Controls')}>
+						<button
+							class=" flex cursor-pointer px-2 py-2 rounded-xl hover:bg-gray-50 dark:hover:bg-gray-850 transition"
+							on:click={async () => {
+								await showControls.set(!$showControls);
+							}}
+							aria-label="Controls"
+						>
+							<div class=" m-auto self-center">
+								<AdjustmentsHorizontal className=" size-5" strokeWidth="0.5" />
+							</div>
+						</button>
+					</Tooltip>
+				{/if}
 
 				<Tooltip content={$i18n.t('New Chat')}>
 					<button

src/lib/components/layout/Navbar/Menu.svelte

@@ -5,13 +5,19 @@
 	import fileSaver from 'file-saver';
 	const { saveAs } = fileSaver;
 
-	import { showSettings } from '$lib/stores';
+	import { downloadChatAsPDF } from '$lib/apis/utils';
+	import { copyToClipboard } from '$lib/utils';
+
+	import { showOverview, showControls, mobile } from '$lib/stores';
 	import { flyAndScale } from '$lib/utils/transitions';
 
 	import Dropdown from '$lib/components/common/Dropdown.svelte';
 	import Tags from '$lib/components/chat/Tags.svelte';
-
-	import { downloadChatAsPDF } from '$lib/apis/utils';
+	import Map from '$lib/components/icons/Map.svelte';
+	import { get } from 'svelte/store';
+	import Clipboard from '$lib/components/icons/Clipboard.svelte';
+	import { toast } from 'svelte-sonner';
+	import AdjustmentsHorizontal from '$lib/components/icons/AdjustmentsHorizontal.svelte';
 
 	const i18n = getContext('i18n');
 
@@ -24,19 +30,23 @@
 	export let chat;
 	export let onClose: Function = () => {};
 
-	const downloadTxt = async () => {
+	const getChatAsText = async () => {
 		const _chat = chat.chat;
-		console.log('download', chat);
-
 		const chatText = _chat.messages.reduce((a, message, i, arr) => {
 			return `${a}### ${message.role.toUpperCase()}\n${message.content}\n\n`;
 		}, '');
 
+		return chatText.trim();
+	};
+
+	const downloadTxt = async () => {
+		const chatText = await getChatAsText();
+
 		let blob = new Blob([chatText], {
 			type: 'text/plain'
 		});
 
-		saveAs(blob, `chat-${_chat.title}.txt`);
+		saveAs(blob, `chat-${chat.chat.title}.txt`);
 	};
 
 	const downloadPdf = async () => {
@@ -117,6 +127,48 @@
 				<div class="flex items-center">{$i18n.t('Settings')}</div>
 			</DropdownMenu.Item> -->
 
+			{#if $mobile}
+				<DropdownMenu.Item
+					class="flex gap-2 items-center px-3 py-2 text-sm  cursor-pointer hover:bg-gray-50 dark:hover:bg-gray-800 rounded-md"
+					id="chat-controls-button"
+					on:click={async () => {
+						await showControls.set(true);
+					}}
+				>
+					<AdjustmentsHorizontal className=" size-4" strokeWidth="0.5" />
+					<div class="flex items-center">{$i18n.t('Controls')}</div>
+				</DropdownMenu.Item>
+			{/if}
+
+			<DropdownMenu.Item
+				class="flex gap-2 items-center px-3 py-2 text-sm  cursor-pointer hover:bg-gray-50 dark:hover:bg-gray-800 rounded-md"
+				id="chat-overview-button"
+				on:click={async () => {
+					await showControls.set(true);
+					await showOverview.set(true);
+				}}
+			>
+				<Map className=" size-4" strokeWidth="1.5" />
+				<div class="flex items-center">{$i18n.t('Overview')}</div>
+			</DropdownMenu.Item>
+
+			<DropdownMenu.Item
+				class="flex gap-2 items-center px-3 py-2 text-sm  cursor-pointer hover:bg-gray-50 dark:hover:bg-gray-800 rounded-md"
+				id="chat-copy-button"
+				on:click={async () => {
+					const res = await copyToClipboard(await getChatAsText()).catch((e) => {
+						console.error(e);
+					});
+
+					if (res) {
+						toast.success($i18n.t('Copied to clipboard'));
+					}
+				}}
+			>
+				<Clipboard className=" size-4" strokeWidth="1.5" />
+				<div class="flex items-center">{$i18n.t('Copy')}</div>
+			</DropdownMenu.Item>
+
 			<DropdownMenu.Item
 				class="flex gap-2 items-center px-3 py-2 text-sm  cursor-pointer hover:bg-gray-50 dark:hover:bg-gray-800 rounded-md"
 				id="chat-share-button"
@@ -138,12 +190,7 @@
 				</svg>
 				<div class="flex items-center">{$i18n.t('Share')}</div>
 			</DropdownMenu.Item>
-			<!-- <DropdownMenu.Item
-					class="flex gap-2 items-center px-3 py-2 text-sm  font-medium cursor-pointer"
-					on:click={() => {
-						downloadHandler();
-					}}
-				/> -->
+
 			<DropdownMenu.Sub>
 				<DropdownMenu.SubTrigger
 					class="flex gap-2 items-center px-3 py-2 text-sm  cursor-pointer hover:bg-gray-50 dark:hover:bg-gray-800 rounded-md"

src/lib/components/layout/Sidebar.svelte

@@ -53,6 +53,8 @@
 	let showDeleteConfirm = false;
 	let showDropdown = false;
 
+	let selectedTagName = null;
+
 	let filteredChatList = [];
 
 	// Pagination variables
@@ -240,7 +242,7 @@
 		deleteChatHandler(deleteChat.id);
 	}}
 >
-	<div class=" text-sm text-gray-500">
+	<div class=" text-sm text-gray-500 flex-1 line-clamp-3">
 		{$i18n.t('This will delete')} <span class="  font-semibold">{deleteChat.title}</span>.
 	</div>
 </DeleteConfirmDialog>
@@ -261,7 +263,7 @@
 	id="sidebar"
 	class="h-screen max-h-[100dvh] min-h-screen select-none {$showSidebar
 		? 'md:relative w-[260px]'
-		: '-translate-x-[260px] w-[0px]'} bg-gray-50 text-gray-900 dark:bg-gray-950 dark:text-gray-200 text-sm transition fixed z-50 top-0 left-0 rounded-r-2xl
+		: '-translate-x-[260px] w-[0px]'} bg-gray-50 text-gray-900 dark:bg-gray-950 dark:text-gray-200 text-sm transition fixed z-50 top-0 left-0
         "
 	data-state={$showSidebar}
 >
@@ -273,7 +275,7 @@
 		<div class="px-2.5 flex justify-between space-x-1 text-gray-600 dark:text-gray-400">
 			<a
 				id="sidebar-new-chat-button"
-				class="flex flex-1 justify-between rounded-xl px-2 py-2 hover:bg-gray-100 dark:hover:bg-gray-900 transition"
+				class="flex flex-1 justify-between rounded-xl px-2 h-full hover:bg-gray-100 dark:hover:bg-gray-900 transition"
 				href="/"
 				draggable="false"
 				on:click={async () => {
@@ -380,11 +382,13 @@
 			</div>
 		{/if}
 
-		<div class="relative flex flex-col flex-1 overflow-y-auto">
+		<div
+			class="relative flex flex-col flex-1 overflow-y-auto {$temporaryChatEnabled
+				? 'opacity-20'
+				: ''}"
+		>
 			{#if $temporaryChatEnabled}
-				<div
-					class="absolute z-40 w-full h-full bg-gray-50/90 dark:bg-black/90 flex justify-center"
-				></div>
+				<div class="absolute z-40 w-full h-full flex justify-center"></div>
 			{/if}
 
 			<div class="px-2 mt-0.5 mb-2 flex justify-center space-x-2">
@@ -419,10 +423,13 @@
 			</div>
 
 			{#if $tags.filter((t) => t.name !== 'pinned').length > 0}
-				<div class="px-2.5 mb-2 flex gap-1 flex-wrap">
+				<div class="px-3.5 mb-1 flex gap-0.5 flex-wrap">
 					<button
-						class="px-2.5 text-xs font-medium bg-gray-50 dark:bg-gray-900 dark:hover:bg-gray-800 transition rounded-full"
+						class="px-2.5 py-[1px] text-xs transition {selectedTagName === null
+							? 'bg-gray-100 dark:bg-gray-900'
+							: ' '} rounded-md font-medium"
 						on:click={async () => {
+							selectedTagName = null;
 							await enablePagination();
 						}}
 					>
@@ -430,8 +437,11 @@
 					</button>
 					{#each $tags.filter((t) => t.name !== 'pinned') as tag}
 						<button
-							class="px-2.5 text-xs font-medium bg-gray-50 dark:bg-gray-900 dark:hover:bg-gray-800 transition rounded-full"
+							class="px-2.5 py-[1px] text-xs transition {selectedTagName === tag.name
+								? 'bg-gray-100 dark:bg-gray-900'
+								: ''}  rounded-md font-medium"
 							on:click={async () => {
+								selectedTagName = tag.name;
 								scrollPaginationEnabled.set(false);
 								let chatIds = await getChatListByTagName(localStorage.token, tag.name);
 								if (chatIds.length === 0) {
@@ -551,7 +561,7 @@
 			</div>
 		</div>
 
-		<div class="px-2.5">
+		<div class="px-2.5 pb-safe-bottom">
 			<!-- <hr class=" border-gray-900 mb-1 w-full" /> -->
 
 			<div class="flex flex-col font-primary">

src/lib/components/workspace/Documents.svelte

@@ -55,7 +55,7 @@
 	const uploadDoc = async (file, tags?: object) => {
 		console.log(file);
 		// Check if the file is an audio file and transcribe/convert it to text file
-		if (['audio/mpeg', 'audio/wav'].includes(file['type'])) {
+		if (['audio/mpeg', 'audio/wav', 'audio/ogg'].includes(file['type'])) {
 			const transcribeRes = await transcribeAudio(localStorage.token, file).catch((error) => {
 				toast.error(error);
 				return null;
@@ -215,7 +215,11 @@
 
 <div class="mb-3">
 	<div class="flex justify-between items-center">
-		<div class=" text-lg font-semibold self-center">{$i18n.t('Documents')}</div>
+		<div class="flex md:self-center text-lg font-medium px-0.5">
+			{$i18n.t('Documents')}
+			<div class="flex self-center w-[1px] h-6 mx-2.5 bg-gray-200 dark:bg-gray-700" />
+			<span class="text-lg font-medium text-gray-500 dark:text-gray-300">{$documents.length}</span>
+		</div>
 	</div>
 </div>
 

src/lib/components/workspace/Models.svelte

@@ -94,6 +94,58 @@
 		window.addEventListener('message', messageHandler, false);
 	};
 
+	const moveToTopHandler = async (model) => {
+		// find models with position 0 and set them to 1
+		const topModels = _models.filter((m) => m.info?.meta?.position === 0);
+		for (const m of topModels) {
+			let info = m.info;
+			if (!info) {
+				info = {
+					id: m.id,
+					name: m.name,
+					meta: {
+						position: 1
+					},
+					params: {}
+				};
+			}
+
+			info.meta = {
+				...info.meta,
+				position: 1
+			};
+
+			await updateModelById(localStorage.token, info.id, info);
+		}
+
+		let info = model.info;
+
+		if (!info) {
+			info = {
+				id: model.id,
+				name: model.name,
+				meta: {
+					position: 0
+				},
+				params: {}
+			};
+		}
+
+		info.meta = {
+			...info.meta,
+			position: 0
+		};
+
+		const res = await updateModelById(localStorage.token, info.id, info);
+
+		if (res) {
+			toast.success($i18n.t(`Model {{name}} is now at the top`, { name: info.id }));
+		}
+
+		await models.set(await getModels(localStorage.token));
+		_models = $models;
+	};
+
 	const hideModelHandler = async (model) => {
 		let info = model.info;
 
@@ -322,7 +374,7 @@
 			>
 				<div class=" self-start w-8 pt-0.5">
 					<div
-						class=" rounded-full bg-stone-700 {(model?.info?.meta?.hidden ?? false)
+						class=" rounded-full object-cover {(model?.info?.meta?.hidden ?? false)
 							? 'brightness-90 dark:brightness-50'
 							: ''} "
 					>
@@ -440,6 +492,9 @@
 						exportHandler={() => {
 							exportModelHandler(model);
 						}}
+						moveToTopHandler={() => {
+							moveToTopHandler(model);
+						}}
 						hideHandler={() => {
 							hideModelHandler(model);
 						}}

src/lib/components/workspace/Models/Capabilities.svelte

@@ -0,0 +1,44 @@
+<script lang="ts">
+	import { getContext } from 'svelte';
+	import Checkbox from '$lib/components/common/Checkbox.svelte';
+	import Tooltip from '$lib/components/common/Tooltip.svelte';
+	import { marked } from 'marked';
+
+	const i18n = getContext('i18n');
+
+	const helpText = {
+		vision: $i18n.t('Model accepts image inputs'),
+		usage: $i18n.t(
+			'Sends `stream_options: { include_usage: true }` in the request.\nSupported providers will return token usage information in the response when set.'
+		)
+	};
+
+	export let capabilities: {
+		vision?: boolean;
+		usage?: boolean;
+	} = {};
+</script>
+
+<div>
+	<div class="flex w-full justify-between mb-1">
+		<div class=" self-center text-sm font-semibold">{$i18n.t('Capabilities')}</div>
+	</div>
+	<div class="flex flex-col">
+		{#each Object.keys(capabilities) as capability}
+			<div class=" flex items-center gap-2">
+				<Checkbox
+					state={capabilities[capability] ? 'checked' : 'unchecked'}
+					on:change={(e) => {
+						capabilities[capability] = e.detail === 'checked';
+					}}
+				/>
+
+				<div class=" py-0.5 text-sm capitalize">
+					<Tooltip content={marked.parse(helpText[capability])}>
+						{$i18n.t(capability)}
+					</Tooltip>
+				</div>
+			</div>
+		{/each}
+	</div>
+</div>

src/lib/components/workspace/Models/ModelMenu.svelte

@@ -12,6 +12,7 @@
 	import ArchiveBox from '$lib/components/icons/ArchiveBox.svelte';
 	import DocumentDuplicate from '$lib/components/icons/DocumentDuplicate.svelte';
 	import ArrowDownTray from '$lib/components/icons/ArrowDownTray.svelte';
+	import ArrowUpCircle from '$lib/components/icons/ArrowUpCircle.svelte';
 
 	const i18n = getContext('i18n');
 
@@ -21,6 +22,7 @@
 	export let cloneHandler: Function;
 	export let exportHandler: Function;
 
+	export let moveToTopHandler: Function;
 	export let hideHandler: Function;
 	export let deleteHandler: Function;
 	export let onClose: Function;
@@ -80,6 +82,17 @@
 				<div class="flex items-center">{$i18n.t('Export')}</div>
 			</DropdownMenu.Item>
 
+			<DropdownMenu.Item
+				class="flex gap-2 items-center px-3 py-2 text-sm  font-medium cursor-pointer hover:bg-gray-50 dark:hover:bg-gray-800 rounded-md"
+				on:click={() => {
+					moveToTopHandler();
+				}}
+			>
+				<ArrowUpCircle />
+
+				<div class="flex items-center">{$i18n.t('Move to Top')}</div>
+			</DropdownMenu.Item>
+
 			<DropdownMenu.Item
 				class="flex  gap-2  items-center px-3 py-2 text-sm  font-medium cursor-pointer hover:bg-gray-50 dark:hover:bg-gray-800 rounded-md"
 				on:click={() => {

src/lib/i18n/locales/ar-BH/translation.json

@@ -73,7 +73,10 @@
 	"AUTOMATIC1111 Api Auth String": "",
 	"AUTOMATIC1111 Base URL": "AUTOMATIC1111 الرابط الرئيسي",
 	"AUTOMATIC1111 Base URL is required.": "AUTOMATIC1111 الرابط مطلوب",
+	"Available list": "",
 	"available!": "متاح",
+	"Azure AI Speech": "",
+	"Azure Region": "",
 	"Back": "خلف",
 	"Bad Response": "استجابة خطاء",
 	"Banners": "لافتات",
@@ -94,6 +97,7 @@
 	"Chat Bubble UI": "UI الدردشة",
 	"Chat Controls": "",
 	"Chat direction": "اتجاه المحادثة",
+	"Chat Overview": "",
 	"Chats": "المحادثات",
 	"Check Again": "تحقق مرة اخرى",
 	"Check for updates": "تحقق من التحديثات",
@@ -230,6 +234,7 @@
 	"Enable Message Rating": "",
 	"Enable New Sign Ups": "تفعيل عمليات التسجيل الجديدة",
 	"Enable Web Search": "تمكين بحث الويب",
+	"Enable Web Search Query Generation": "",
 	"Enabled": "",
 	"Engine": "",
 	"Ensure your CSV file includes 4 columns in this order: Name, Email, Password, Role.": "تأكد من أن ملف CSV الخاص بك يتضمن 4 أعمدة بهذا الترتيب: Name, Email, Password, Role.",
@@ -237,6 +242,7 @@
 	"Enter a detail about yourself for your LLMs to recall": "ادخل معلومات عنك تريد أن يتذكرها الموديل",
 	"Enter api auth string (e.g. username:password)": "",
 	"Enter Brave Search API Key": "أدخل مفتاح واجهة برمجة تطبيقات البحث الشجاع",
+	"Enter CFG Scale (e.g. 7.0)": "",
 	"Enter Chunk Overlap": "أدخل الChunk Overlap",
 	"Enter Chunk Size": "أدخل Chunk الحجم",
 	"Enter Github Raw URL": "أدخل عنوان URL ل Github Raw",
@@ -247,6 +253,8 @@
 	"Enter Model ID": "",
 	"Enter model tag (e.g. {{modelTag}})": "(e.g. {{modelTag}}) أدخل الموديل تاق",
 	"Enter Number of Steps (e.g. 50)": "(e.g. 50) أدخل عدد الخطوات",
+	"Enter Sampler (e.g. Euler a)": "",
+	"Enter Scheduler (e.g. Karras)": "",
 	"Enter Score": "أدخل النتيجة",
 	"Enter SearchApi API Key": "",
 	"Enter SearchApi Engine": "",
@@ -367,6 +375,7 @@
 	"Last Active": "آخر نشاط",
 	"Last Modified": "",
 	"Leave empty for unlimited": "",
+	"Leave empty to use the default prompt, or enter a custom prompt": "",
 	"Light": "فاتح",
 	"Listening...": "",
 	"LLMs can make mistakes. Verify important information.": "يمكن أن تصدر بعض الأخطاء. لذلك يجب التحقق من المعلومات المهمة",
@@ -406,6 +415,8 @@
 	"Model {{modelId}} not found": "لم يتم العثور على النموذج {{modelId}}.",
 	"Model {{modelName}} is not vision capable": "نموذج {{modelName}} غير قادر على الرؤية",
 	"Model {{name}} is now {{status}}": "نموذج {{name}} هو الآن {{status}}",
+	"Model {{name}} is now at the top": "",
+	"Model accepts image inputs": "",
 	"Model created successfully!": "",
 	"Model filesystem path detected. Model shortname is required for update, cannot continue.": "تم اكتشاف مسار نظام الملفات النموذجي. الاسم المختصر للنموذج مطلوب للتحديث، ولا يمكن الاستمرار.",
 	"Model ID": "رقم الموديل",
@@ -417,6 +428,7 @@
 	"Modelfile Content": "محتوى الملف النموذجي",
 	"Models": "الموديلات",
 	"More": "المزيد",
+	"Move to Top": "",
 	"Name": "الأسم",
 	"Name Tag": "أسم التاق",
 	"Name your model": "قم بتسمية النموذج الخاص بك",
@@ -462,6 +474,8 @@
 	"OpenAI URL/Key required.": "URL/مفتاح OpenAI.مطلوب عنوان ",
 	"or": "أو",
 	"Other": "آخر",
+	"Output format": "",
+	"Overview": "",
 	"Password": "الباسورد",
 	"PDF document (.pdf)": "PDF ملف (.pdf)",
 	"PDF Extract Images (OCR)": "PDF أستخرج الصور (OCR)",
@@ -524,6 +538,7 @@
 	"Save": "حفظ",
 	"Save & Create": "حفظ وإنشاء",
 	"Save & Update": "حفظ وتحديث",
+	"Save As Copy": "",
 	"Save Tag": "",
 	"Saving chat logs directly to your browser's storage is no longer supported. Please take a moment to download and delete your chat logs by clicking the button below. Don't worry, you can easily re-import your chat logs to the backend through": "لم يعد حفظ سجلات الدردشة مباشرة في مساحة تخزين متصفحك مدعومًا. يرجى تخصيص بعض الوقت لتنزيل وحذف سجلات الدردشة الخاصة بك عن طريق النقر على الزر أدناه. لا تقلق، يمكنك بسهولة إعادة استيراد سجلات الدردشة الخاصة بك إلى الواجهة الخلفية من خلاله",
 	"Scan": "مسح",
@@ -538,7 +553,6 @@
 	"Search Models": "نماذج البحث",
 	"Search Prompts": "أبحث حث",
 	"Search Query Generation Prompt": "",
-	"Search Query Generation Prompt Length Threshold": "",
 	"Search Result Count": "عدد نتائج البحث",
 	"Search Tools": "",
 	"SearchApi API Key": "",
@@ -571,16 +585,20 @@
 	"Send": "تم",
 	"Send a Message": "يُرجى إدخال طلبك هنا",
 	"Send message": "يُرجى إدخال طلبك هنا.",
+	"Sends `stream_options: { include_usage: true }` in the request.\nSupported providers will return token usage information in the response when set.": "",
 	"September": "سبتمبر",
 	"Serper API Key": "مفتاح واجهة برمجة تطبيقات سيربر",
 	"Serply API Key": "",
 	"Serpstack API Key": "مفتاح واجهة برمجة تطبيقات Serpstack",
 	"Server connection verified": "تم التحقق من اتصال الخادم",
 	"Set as default": "الافتراضي",
+	"Set CFG Scale": "",
 	"Set Default Model": "تفعيد الموديل الافتراضي",
 	"Set embedding model (e.g. {{model}})": "ضبط نموذج المتجهات (على سبيل المثال: {{model}})",
 	"Set Image Size": "حجم الصورة",
 	"Set reranking model (e.g. {{model}})": "ضبط نموذج إعادة الترتيب (على سبيل المثال: {{model}})",
+	"Set Sampler": "",
+	"Set Scheduler": "",
 	"Set Steps": "ضبط الخطوات",
 	"Set Task Model": "تعيين نموذج المهمة",
 	"Set Voice": "ضبط الصوت",
@@ -603,7 +621,9 @@
 	"Source": "المصدر",
 	"Speech recognition error: {{error}}": "{{error}} خطأ في التعرف على الكلام",
 	"Speech-to-Text Engine": "محرك تحويل الكلام إلى نص",
+	"Speed Rate": "",
 	"Stop Sequence": "وقف التسلسل",
+	"Stream Chat Response": "",
 	"STT Model": "",
 	"STT Settings": "STT اعدادات",
 	"Submit": "إرسال",

src/lib/i18n/locales/bg-BG/translation.json

@@ -73,7 +73,10 @@
 	"AUTOMATIC1111 Api Auth String": "",
 	"AUTOMATIC1111 Base URL": "AUTOMATIC1111 Базов URL",
 	"AUTOMATIC1111 Base URL is required.": "AUTOMATIC1111 Базов URL е задължителен.",
+	"Available list": "",
 	"available!": "наличен!",
+	"Azure AI Speech": "",
+	"Azure Region": "",
 	"Back": "Назад",
 	"Bad Response": "Невалиден отговор от API",
 	"Banners": "Банери",
@@ -94,6 +97,7 @@
 	"Chat Bubble UI": "UI за чат бублон",
 	"Chat Controls": "",
 	"Chat direction": "Направление на чата",
+	"Chat Overview": "",
 	"Chats": "Чатове",
 	"Check Again": "Проверете Още Веднъж",
 	"Check for updates": "Проверка за актуализации",
@@ -230,6 +234,7 @@
 	"Enable Message Rating": "",
 	"Enable New Sign Ups": "Вклюване на Нови Потребители",
 	"Enable Web Search": "Разрешаване на търсене в уеб",
+	"Enable Web Search Query Generation": "",
 	"Enabled": "",
 	"Engine": "",
 	"Ensure your CSV file includes 4 columns in this order: Name, Email, Password, Role.": "Уверете се, че вашият CSV файл включва 4 колони в следния ред: Име, Имейл, Парола, Роля.",
@@ -237,6 +242,7 @@
 	"Enter a detail about yourself for your LLMs to recall": "Въведете подробности за себе си, за да се herinnerат вашите LLMs",
 	"Enter api auth string (e.g. username:password)": "",
 	"Enter Brave Search API Key": "Въведете Brave Search API ключ",
+	"Enter CFG Scale (e.g. 7.0)": "",
 	"Enter Chunk Overlap": "Въведете Chunk Overlap",
 	"Enter Chunk Size": "Въведете Chunk Size",
 	"Enter Github Raw URL": "Въведете URL адреса на Github Raw",
@@ -247,6 +253,8 @@
 	"Enter Model ID": "",
 	"Enter model tag (e.g. {{modelTag}})": "Въведете таг на модел (напр. {{modelTag}})",
 	"Enter Number of Steps (e.g. 50)": "Въведете брой стъпки (напр. 50)",
+	"Enter Sampler (e.g. Euler a)": "",
+	"Enter Scheduler (e.g. Karras)": "",
 	"Enter Score": "Въведете оценка",
 	"Enter SearchApi API Key": "",
 	"Enter SearchApi Engine": "",
@@ -367,6 +375,7 @@
 	"Last Active": "Последни активни",
 	"Last Modified": "",
 	"Leave empty for unlimited": "",
+	"Leave empty to use the default prompt, or enter a custom prompt": "",
 	"Light": "Светъл",
 	"Listening...": "",
 	"LLMs can make mistakes. Verify important information.": "LLMs могат да правят грешки. Проверете важните данни.",
@@ -406,6 +415,8 @@
 	"Model {{modelId}} not found": "Моделът {{modelId}} не е намерен",
 	"Model {{modelName}} is not vision capable": "Моделът {{modelName}} не може да се вижда",
 	"Model {{name}} is now {{status}}": "Моделът {{name}} сега е {{status}}",
+	"Model {{name}} is now at the top": "",
+	"Model accepts image inputs": "",
 	"Model created successfully!": "",
 	"Model filesystem path detected. Model shortname is required for update, cannot continue.": "Открит е път до файловата система на модела. За актуализацията се изисква съкратено име на модела, не може да продължи.",
 	"Model ID": "ИД на модел",
@@ -417,6 +428,7 @@
 	"Modelfile Content": "Съдържание на модфайл",
 	"Models": "Модели",
 	"More": "Повече",
+	"Move to Top": "",
 	"Name": "Име",
 	"Name Tag": "Име Таг",
 	"Name your model": "Дайте име на вашия модел",
@@ -462,6 +474,8 @@
 	"OpenAI URL/Key required.": "OpenAI URL/Key е задължителен.",
 	"or": "или",
 	"Other": "Other",
+	"Output format": "",
+	"Overview": "",
 	"Password": "Парола",
 	"PDF document (.pdf)": "PDF документ (.pdf)",
 	"PDF Extract Images (OCR)": "PDF Extract Images (OCR)",
@@ -524,6 +538,7 @@
 	"Save": "Запис",
 	"Save & Create": "Запис & Създаване",
 	"Save & Update": "Запис & Актуализиране",
+	"Save As Copy": "",
 	"Save Tag": "",
 	"Saving chat logs directly to your browser's storage is no longer supported. Please take a moment to download and delete your chat logs by clicking the button below. Don't worry, you can easily re-import your chat logs to the backend through": "Запазването на чат логове директно в хранилището на вашия браузър вече не се поддържа. Моля, отделете малко време, за да изтеглите и изтриете чат логовете си, като щракнете върху бутона по-долу. Не се притеснявайте, можете лесно да импортирате отново чат логовете си в бекенда чрез",
 	"Scan": "Сканиране",
@@ -538,7 +553,6 @@
 	"Search Models": "Търсене на модели",
 	"Search Prompts": "Търси Промптове",
 	"Search Query Generation Prompt": "",
-	"Search Query Generation Prompt Length Threshold": "",
 	"Search Result Count": "Брой резултати от търсенето",
 	"Search Tools": "",
 	"SearchApi API Key": "",
@@ -567,16 +581,20 @@
 	"Send": "Изпрати",
 	"Send a Message": "Изпращане на Съобщение",
 	"Send message": "Изпращане на съобщение",
+	"Sends `stream_options: { include_usage: true }` in the request.\nSupported providers will return token usage information in the response when set.": "",
 	"September": "Септември",
 	"Serper API Key": "Serper API ключ",
 	"Serply API Key": "",
 	"Serpstack API Key": "Serpstack API ключ",
 	"Server connection verified": "Server connection verified",
 	"Set as default": "Задай по подразбиране",
+	"Set CFG Scale": "",
 	"Set Default Model": "Задай Модел По Подразбиране",
 	"Set embedding model (e.g. {{model}})": "Задай embedding model (e.g. {{model}})",
 	"Set Image Size": "Задай Размер на Изображението",
 	"Set reranking model (e.g. {{model}})": "Задай reranking model (e.g. {{model}})",
+	"Set Sampler": "",
+	"Set Scheduler": "",
 	"Set Steps": "Задай Стъпки",
 	"Set Task Model": "Задаване на модел на задача",
 	"Set Voice": "Задай Глас",
@@ -599,7 +617,9 @@
 	"Source": "Източник",
 	"Speech recognition error: {{error}}": "Speech recognition error: {{error}}",
 	"Speech-to-Text Engine": "Speech-to-Text Engine",
+	"Speed Rate": "",
 	"Stop Sequence": "Stop Sequence",
+	"Stream Chat Response": "",
 	"STT Model": "",
 	"STT Settings": "STT Настройки",
 	"Submit": "Изпращане",

src/lib/i18n/locales/bn-BD/translation.json

@@ -73,7 +73,10 @@
 	"AUTOMATIC1111 Api Auth String": "",
 	"AUTOMATIC1111 Base URL": "AUTOMATIC1111 বেজ ইউআরএল",
 	"AUTOMATIC1111 Base URL is required.": "AUTOMATIC1111 বেজ ইউআরএল আবশ্যক",
+	"Available list": "",
 	"available!": "উপলব্ধ!",
+	"Azure AI Speech": "",
+	"Azure Region": "",
 	"Back": "পেছনে",
 	"Bad Response": "খারাপ প্রতিক্রিয়া",
 	"Banners": "ব্যানার",
@@ -94,6 +97,7 @@
 	"Chat Bubble UI": "চ্যাট বাবল UI",
 	"Chat Controls": "",
 	"Chat direction": "চ্যাট দিকনির্দেশ",
+	"Chat Overview": "",
 	"Chats": "চ্যাটসমূহ",
 	"Check Again": "আবার চেক করুন",
 	"Check for updates": "নতুন আপডেট আছে কিনা চেক করুন",
@@ -230,6 +234,7 @@
 	"Enable Message Rating": "",
 	"Enable New Sign Ups": "নতুন সাইনআপ চালু করুন",
 	"Enable Web Search": "ওয়েব অনুসন্ধান সক্ষম করুন",
+	"Enable Web Search Query Generation": "",
 	"Enabled": "",
 	"Engine": "",
 	"Ensure your CSV file includes 4 columns in this order: Name, Email, Password, Role.": "আপনার সিএসভি ফাইলটিতে এই ক্রমে 4 টি কলাম অন্তর্ভুক্ত রয়েছে তা নিশ্চিত করুন: নাম, ইমেল, পাসওয়ার্ড, ভূমিকা।.",
@@ -237,6 +242,7 @@
 	"Enter a detail about yourself for your LLMs to recall": "আপনার এলএলএমগুলি স্মরণ করার জন্য নিজের সম্পর্কে একটি বিশদ লিখুন",
 	"Enter api auth string (e.g. username:password)": "",
 	"Enter Brave Search API Key": "সাহসী অনুসন্ধান API কী লিখুন",
+	"Enter CFG Scale (e.g. 7.0)": "",
 	"Enter Chunk Overlap": "চাঙ্ক ওভারল্যাপ লিখুন",
 	"Enter Chunk Size": "চাংক সাইজ লিখুন",
 	"Enter Github Raw URL": "গিটহাব কাঁচা URL লিখুন",
@@ -247,6 +253,8 @@
 	"Enter Model ID": "",
 	"Enter model tag (e.g. {{modelTag}})": "মডেল ট্যাগ লিখুন (e.g. {{modelTag}})",
 	"Enter Number of Steps (e.g. 50)": "ধাপের সংখ্যা দিন (যেমন: 50)",
+	"Enter Sampler (e.g. Euler a)": "",
+	"Enter Scheduler (e.g. Karras)": "",
 	"Enter Score": "স্কোর দিন",
 	"Enter SearchApi API Key": "",
 	"Enter SearchApi Engine": "",
@@ -367,6 +375,7 @@
 	"Last Active": "সর্বশেষ সক্রিয়",
 	"Last Modified": "",
 	"Leave empty for unlimited": "",
+	"Leave empty to use the default prompt, or enter a custom prompt": "",
 	"Light": "লাইট",
 	"Listening...": "",
 	"LLMs can make mistakes. Verify important information.": "LLM ভুল করতে পারে। গুরুত্বপূর্ণ তথ্য যাচাই করে নিন।",
@@ -406,6 +415,8 @@
 	"Model {{modelId}} not found": "{{modelId}} মডেল পাওয়া যায়নি",
 	"Model {{modelName}} is not vision capable": "মডেল {{modelName}} দৃষ্টি সক্ষম নয়",
 	"Model {{name}} is now {{status}}": "মডেল {{name}} এখন {{status}}",
+	"Model {{name}} is now at the top": "",
+	"Model accepts image inputs": "",
 	"Model created successfully!": "",
 	"Model filesystem path detected. Model shortname is required for update, cannot continue.": "মডেল ফাইলসিস্টেম পাথ পাওয়া গেছে। আপডেটের জন্য মডেলের শর্টনেম আবশ্যক, এগিয়ে যাওয়া যাচ্ছে না।",
 	"Model ID": "মডেল ID",
@@ -417,6 +428,7 @@
 	"Modelfile Content": "মডেলফাইল কনটেন্ট",
 	"Models": "মডেলসমূহ",
 	"More": "আরো",
+	"Move to Top": "",
 	"Name": "নাম",
 	"Name Tag": "নামের ট্যাগ",
 	"Name your model": "আপনার মডেলের নাম দিন",
@@ -462,6 +474,8 @@
 	"OpenAI URL/Key required.": "OpenAI URL/Key আবশ্যক",
 	"or": "অথবা",
 	"Other": "অন্যান্য",
+	"Output format": "",
+	"Overview": "",
 	"Password": "পাসওয়ার্ড",
 	"PDF document (.pdf)": "PDF ডকুমেন্ট (.pdf)",
 	"PDF Extract Images (OCR)": "পিডিএফ এর ছবি থেকে লেখা বের করুন (OCR)",
@@ -524,6 +538,7 @@
 	"Save": "সংরক্ষণ",
 	"Save & Create": "সংরক্ষণ এবং তৈরি করুন",
 	"Save & Update": "সংরক্ষণ এবং আপডেট করুন",
+	"Save As Copy": "",
 	"Save Tag": "",
 	"Saving chat logs directly to your browser's storage is no longer supported. Please take a moment to download and delete your chat logs by clicking the button below. Don't worry, you can easily re-import your chat logs to the backend through": "মাধ্যমে",
 	"Scan": "স্ক্যান",
@@ -538,7 +553,6 @@
 	"Search Models": "অনুসন্ধান মডেল",
 	"Search Prompts": "প্রম্পটসমূহ অনুসন্ধান করুন",
 	"Search Query Generation Prompt": "",
-	"Search Query Generation Prompt Length Threshold": "",
 	"Search Result Count": "অনুসন্ধানের ফলাফল গণনা",
 	"Search Tools": "",
 	"SearchApi API Key": "",
@@ -567,16 +581,20 @@
 	"Send": "পাঠান",
 	"Send a Message": "একটি মেসেজ পাঠান",
 	"Send message": "মেসেজ পাঠান",
+	"Sends `stream_options: { include_usage: true }` in the request.\nSupported providers will return token usage information in the response when set.": "",
 	"September": "সেপ্টেম্বর",
 	"Serper API Key": "Serper API Key",
 	"Serply API Key": "",
 	"Serpstack API Key": "Serpstack API Key",
 	"Server connection verified": "সার্ভার কানেকশন যাচাই করা হয়েছে",
 	"Set as default": "ডিফল্ট হিসেবে নির্ধারণ করুন",
+	"Set CFG Scale": "",
 	"Set Default Model": "ডিফল্ট মডেল নির্ধারণ করুন",
 	"Set embedding model (e.g. {{model}})": "ইমেম্বিং মডেল নির্ধারণ করুন (উদাহরণ {{model}})",
 	"Set Image Size": "ছবির সাইজ নির্ধারণ করুন",
 	"Set reranking model (e.g. {{model}})": "রি-র্যাংকিং মডেল নির্ধারণ করুন (উদাহরণ {{model}})",
+	"Set Sampler": "",
+	"Set Scheduler": "",
 	"Set Steps": "পরবর্তী ধাপসমূহ",
 	"Set Task Model": "টাস্ক মডেল সেট করুন",
 	"Set Voice": "কন্ঠস্বর নির্ধারণ করুন",
@@ -599,7 +617,9 @@
 	"Source": "উৎস",
 	"Speech recognition error: {{error}}": "স্পিচ রিকগনিশনে সমস্যা: {{error}}",
 	"Speech-to-Text Engine": "স্পিচ-টু-টেক্সট ইঞ্জিন",
+	"Speed Rate": "",
 	"Stop Sequence": "সিকোয়েন্স থামান",
+	"Stream Chat Response": "",
 	"STT Model": "",
 	"STT Settings": "STT সেটিংস",
 	"Submit": "সাবমিট",

src/lib/i18n/locales/ca-ES/translation.json

@@ -73,7 +73,10 @@
 	"AUTOMATIC1111 Api Auth String": "Cadena d'autenticació de l'API d'AUTOMATIC1111",
 	"AUTOMATIC1111 Base URL": "URL Base d'AUTOMATIC1111",
 	"AUTOMATIC1111 Base URL is required.": "Es requereix l'URL Base d'AUTOMATIC1111.",
+	"Available list": "",
 	"available!": "disponible!",
+	"Azure AI Speech": "",
+	"Azure Region": "",
 	"Back": "Enrere",
 	"Bad Response": "Resposta errònia",
 	"Banners": "Banners",
@@ -94,6 +97,7 @@
 	"Chat Bubble UI": "Chat Bubble UI",
 	"Chat Controls": "Controls de xat",
 	"Chat direction": "Direcció del xat",
+	"Chat Overview": "",
 	"Chats": "Xats",
 	"Check Again": "Comprovar-ho de nou",
 	"Check for updates": "Comprovar si hi ha actualitzacions",
@@ -230,6 +234,7 @@
 	"Enable Message Rating": "Permetre la qualificació de missatges",
 	"Enable New Sign Ups": "Permetre nous registres",
 	"Enable Web Search": "Activar la cerca web",
+	"Enable Web Search Query Generation": "Activa la generació de consultes de cerca web",
 	"Enabled": "Habilitat",
 	"Engine": "Motor",
 	"Ensure your CSV file includes 4 columns in this order: Name, Email, Password, Role.": "Assegura't que els teus fitxers CSV inclouen 4 columnes en aquest ordre: Nom, Correu electrònic, Contrasenya, Rol.",
@@ -237,6 +242,7 @@
 	"Enter a detail about yourself for your LLMs to recall": "Introdueix un detall sobre tu què els teus models de llenguatge puguin recordar",
 	"Enter api auth string (e.g. username:password)": "Entra la cadena d'autenticació api (p. ex. nom d'usuari:contrasenya)",
 	"Enter Brave Search API Key": "Introdueix la clau API de Brave Search",
+	"Enter CFG Scale (e.g. 7.0)": "Entra l'escala CFG (p.ex. 7.0)",
 	"Enter Chunk Overlap": "Introdueix la mida de solapament de blocs",
 	"Enter Chunk Size": "Introdueix la mida del bloc",
 	"Enter Github Raw URL": "Introdueix l'URL en brut de Github",
@@ -247,9 +253,11 @@
 	"Enter Model ID": "Introdueix l'identificador del model",
 	"Enter model tag (e.g. {{modelTag}})": "Introdueix l'etiqueta del model (p. ex. {{modelTag}})",
 	"Enter Number of Steps (e.g. 50)": "Introdueix el nombre de passos (p. ex. 50)",
+	"Enter Sampler (e.g. Euler a)": "Introdueix el mostrejador (p.ex. Euler a)",
+	"Enter Scheduler (e.g. Karras)": "Entra el programador (p.ex. Karras)",
 	"Enter Score": "Introdueix la puntuació",
-	"Enter SearchApi API Key": "",
-	"Enter SearchApi Engine": "",
+	"Enter SearchApi API Key": "Introdueix la clau API SearchApi",
+	"Enter SearchApi Engine": "Introdueix el motor SearchApi",
 	"Enter Searxng Query URL": "Introdueix l'URL de consulta de Searxng",
 	"Enter Serper API Key": "Introdueix la clau API Serper",
 	"Enter Serply API Key": "Introdueix la clau API Serply",
@@ -272,7 +280,7 @@
 	"Export All Chats (All Users)": "Exportar tots els xats (Tots els usuaris)",
 	"Export chat (.json)": "Exportar el xat (.json)",
 	"Export Chats": "Exportar els xats",
-	"Export Config to JSON File": "",
+	"Export Config to JSON File": "Exportar la configuració a un arxiu JSON",
 	"Export Documents Mapping": "Exportar el mapatge de documents",
 	"Export Functions": "Exportar funcions",
 	"Export LiteLLM config.yaml": "Exportar la configuració LiteLLM config.yaml",
@@ -337,7 +345,7 @@
 	"Image Settings": "Preferències d'imatges",
 	"Images": "Imatges",
 	"Import Chats": "Importar xats",
-	"Import Config from JSON File": "",
+	"Import Config from JSON File": "Importar la configuració des d'un arxiu JSON",
 	"Import Documents Mapping": "Importar el mapatge de documents",
 	"Import Functions": "Importar funcions",
 	"Import Models": "Importar models",
@@ -367,6 +375,7 @@
 	"Last Active": "Activitat recent",
 	"Last Modified": "Modificació",
 	"Leave empty for unlimited": "Deixar-ho buit per il·limitat",
+	"Leave empty to use the default prompt, or enter a custom prompt": "Deixa-ho en blanc per utilitzar la indicació predeterminada o introdueix una indicació personalitzada",
 	"Light": "Clar",
 	"Listening...": "Escoltant...",
 	"LLMs can make mistakes. Verify important information.": "Els models de llenguatge poden cometre errors. Verifica la informació important.",
@@ -406,6 +415,8 @@
 	"Model {{modelId}} not found": "No s'ha trobat el model {{modelId}}",
 	"Model {{modelName}} is not vision capable": "El model {{modelName}} no és capaç de visió",
 	"Model {{name}} is now {{status}}": "El model {{name}} ara és {{status}}",
+	"Model {{name}} is now at the top": "",
+	"Model accepts image inputs": "El model accepta entrades d'imatge",
 	"Model created successfully!": "Model creat correctament",
 	"Model filesystem path detected. Model shortname is required for update, cannot continue.": "S'ha detectat el camí del sistema de fitxers del model. És necessari un nom curt del model per actualitzar, no es pot continuar.",
 	"Model ID": "Identificador del model",
@@ -417,6 +428,7 @@
 	"Modelfile Content": "Contingut del Modelfile",
 	"Models": "Models",
 	"More": "Més",
+	"Move to Top": "",
 	"Name": "Nom",
 	"Name Tag": "Etiqueta de nom",
 	"Name your model": "Posa un nom al teu model",
@@ -462,6 +474,8 @@
 	"OpenAI URL/Key required.": "URL/Clau d'OpenAI requerides.",
 	"or": "o",
 	"Other": "Altres",
+	"Output format": "",
+	"Overview": "",
 	"Password": "Contrasenya",
 	"PDF document (.pdf)": "Document PDF (.pdf)",
 	"PDF Extract Images (OCR)": "Extreu imatges del PDF (OCR)",
@@ -524,6 +538,7 @@
 	"Save": "Desar",
 	"Save & Create": "Desar i crear",
 	"Save & Update": "Desar i actualitzar",
+	"Save As Copy": "",
 	"Save Tag": "Desar l'etiqueta",
 	"Saving chat logs directly to your browser's storage is no longer supported. Please take a moment to download and delete your chat logs by clicking the button below. Don't worry, you can easily re-import your chat logs to the backend through": "Desar els registres de xat directament a l'emmagatzematge del teu navegador ja no està suportat. Si us plau, descarregr i elimina els registres de xat fent clic al botó de sota. No et preocupis, pots tornar a importar fàcilment els teus registres de xat al backend a través de",
 	"Scan": "Escanejar",
@@ -538,11 +553,10 @@
 	"Search Models": "Cercar models",
 	"Search Prompts": "Cercar indicacions",
 	"Search Query Generation Prompt": "Indicació de cerca de generació de consultes",
-	"Search Query Generation Prompt Length Threshold": "Mida màxima de la indicació de cerca de generació de consultes",
 	"Search Result Count": "Recompte de resultats de cerca",
 	"Search Tools": "Cercar eines",
-	"SearchApi API Key": "",
-	"SearchApi Engine": "",
+	"SearchApi API Key": "Clau API de SearchApi",
+	"SearchApi Engine": "Motor de SearchApi",
 	"Searched {{count}} sites_one": "S'ha cercat {{count}} una pàgina",
 	"Searched {{count}} sites_many": "S'han cercat {{count}} pàgines",
 	"Searched {{count}} sites_other": "S'han cercat {{count}} pàgines",
@@ -568,16 +582,20 @@
 	"Send": "Enviar",
 	"Send a Message": "Enviar un missatge",
 	"Send message": "Enviar missatge",
+	"Sends `stream_options: { include_usage: true }` in the request.\nSupported providers will return token usage information in the response when set.": "Envia `stream_options: { include_usage: true }` a la sol·licitud.\nEls proveïdors compatibles retornaran la informació d'ús del token a la resposta quan s'estableixi.",
 	"September": "Setembre",
 	"Serper API Key": "Clau API de Serper",
 	"Serply API Key": "Clau API de Serply",
 	"Serpstack API Key": "Clau API de Serpstack",
 	"Server connection verified": "Connexió al servidor verificada",
 	"Set as default": "Establir com a predeterminat",
+	"Set CFG Scale": "Establir l'escala CFG",
 	"Set Default Model": "Establir el model predeterminat",
 	"Set embedding model (e.g. {{model}})": "Establir el model d'incrustació (p.ex. {{model}})",
 	"Set Image Size": "Establir la mida de la image",
 	"Set reranking model (e.g. {{model}})": "Establir el model de reavaluació (p.ex. {{model}})",
+	"Set Sampler": "Establir el mostrejador",
+	"Set Scheduler": "Establir el programador",
 	"Set Steps": "Establir el nombre de passos",
 	"Set Task Model": "Establir el model de tasca",
 	"Set Voice": "Establir la veu",
@@ -600,7 +618,9 @@
 	"Source": "Font",
 	"Speech recognition error: {{error}}": "Error de reconeixement de veu: {{error}}",
 	"Speech-to-Text Engine": "Motor de veu a text",
+	"Speed Rate": "",
 	"Stop Sequence": "Atura la seqüència",
+	"Stream Chat Response": "",
 	"STT Model": "Model SST",
 	"STT Settings": "Preferències de STT",
 	"Submit": "Enviar",

src/lib/i18n/locales/ceb-PH/translation.json

@@ -73,7 +73,10 @@
 	"AUTOMATIC1111 Api Auth String": "",
 	"AUTOMATIC1111 Base URL": "Base URL AUTOMATIC1111",
 	"AUTOMATIC1111 Base URL is required.": "Ang AUTOMATIC1111 base URL gikinahanglan.",
+	"Available list": "",
 	"available!": "magamit!",
+	"Azure AI Speech": "",
+	"Azure Region": "",
 	"Back": "Balik",
 	"Bad Response": "",
 	"Banners": "",
@@ -94,6 +97,7 @@
 	"Chat Bubble UI": "",
 	"Chat Controls": "",
 	"Chat direction": "",
+	"Chat Overview": "",
 	"Chats": "Mga panaghisgot",
 	"Check Again": "Susiha pag-usab",
 	"Check for updates": "Susiha ang mga update",
@@ -230,6 +234,7 @@
 	"Enable Message Rating": "",
 	"Enable New Sign Ups": "I-enable ang bag-ong mga rehistro",
 	"Enable Web Search": "",
+	"Enable Web Search Query Generation": "",
 	"Enabled": "",
 	"Engine": "",
 	"Ensure your CSV file includes 4 columns in this order: Name, Email, Password, Role.": "",
@@ -237,6 +242,7 @@
 	"Enter a detail about yourself for your LLMs to recall": "",
 	"Enter api auth string (e.g. username:password)": "",
 	"Enter Brave Search API Key": "",
+	"Enter CFG Scale (e.g. 7.0)": "",
 	"Enter Chunk Overlap": "Pagsulod sa block overlap",
 	"Enter Chunk Size": "Isulod ang block size",
 	"Enter Github Raw URL": "",
@@ -247,6 +253,8 @@
 	"Enter Model ID": "",
 	"Enter model tag (e.g. {{modelTag}})": "Pagsulod sa template tag (e.g. {{modelTag}})",
 	"Enter Number of Steps (e.g. 50)": "Pagsulod sa gidaghanon sa mga lakang (e.g. 50)",
+	"Enter Sampler (e.g. Euler a)": "",
+	"Enter Scheduler (e.g. Karras)": "",
 	"Enter Score": "",
 	"Enter SearchApi API Key": "",
 	"Enter SearchApi Engine": "",
@@ -367,6 +375,7 @@
 	"Last Active": "",
 	"Last Modified": "",
 	"Leave empty for unlimited": "",
+	"Leave empty to use the default prompt, or enter a custom prompt": "",
 	"Light": "Kahayag",
 	"Listening...": "",
 	"LLMs can make mistakes. Verify important information.": "Ang mga LLM mahimong masayop. ",
@@ -406,6 +415,8 @@
 	"Model {{modelId}} not found": "Modelo {{modelId}} wala makit-an",
 	"Model {{modelName}} is not vision capable": "",
 	"Model {{name}} is now {{status}}": "",
+	"Model {{name}} is now at the top": "",
+	"Model accepts image inputs": "",
 	"Model created successfully!": "",
 	"Model filesystem path detected. Model shortname is required for update, cannot continue.": "",
 	"Model ID": "",
@@ -417,6 +428,7 @@
 	"Modelfile Content": "Mga sulod sa template file",
 	"Models": "Mga modelo",
 	"More": "",
+	"Move to Top": "",
 	"Name": "Ngalan",
 	"Name Tag": "Tag sa ngalan",
 	"Name your model": "",
@@ -462,6 +474,8 @@
 	"OpenAI URL/Key required.": "",
 	"or": "O",
 	"Other": "",
+	"Output format": "",
+	"Overview": "",
 	"Password": "Password",
 	"PDF document (.pdf)": "",
 	"PDF Extract Images (OCR)": "PDF Image Extraction (OCR)",
@@ -524,6 +538,7 @@
 	"Save": "Tipigi",
 	"Save & Create": "I-save ug Paghimo",
 	"Save & Update": "I-save ug I-update",
+	"Save As Copy": "",
 	"Save Tag": "",
 	"Saving chat logs directly to your browser's storage is no longer supported. Please take a moment to download and delete your chat logs by clicking the button below. Don't worry, you can easily re-import your chat logs to the backend through": "Ang pag-save sa mga chat log direkta sa imong browser storage dili na suportado. ",
 	"Scan": "Aron ma-scan",
@@ -538,7 +553,6 @@
 	"Search Models": "",
 	"Search Prompts": "Pangitaa ang mga prompt",
 	"Search Query Generation Prompt": "",
-	"Search Query Generation Prompt Length Threshold": "",
 	"Search Result Count": "",
 	"Search Tools": "",
 	"SearchApi API Key": "",
@@ -567,16 +581,20 @@
 	"Send": "",
 	"Send a Message": "Magpadala ug mensahe",
 	"Send message": "Magpadala ug mensahe",
+	"Sends `stream_options: { include_usage: true }` in the request.\nSupported providers will return token usage information in the response when set.": "",
 	"September": "",
 	"Serper API Key": "",
 	"Serply API Key": "",
 	"Serpstack API Key": "",
 	"Server connection verified": "Gipamatud-an nga koneksyon sa server",
 	"Set as default": "Define pinaagi sa default",
+	"Set CFG Scale": "",
 	"Set Default Model": "Ibutang ang default template",
 	"Set embedding model (e.g. {{model}})": "",
 	"Set Image Size": "Ibutang ang gidak-on sa hulagway",
 	"Set reranking model (e.g. {{model}})": "",
+	"Set Sampler": "",
+	"Set Scheduler": "",
 	"Set Steps": "Ipasabot ang mga lakang",
 	"Set Task Model": "",
 	"Set Voice": "Ibutang ang tingog",
@@ -599,7 +617,9 @@
 	"Source": "Tinubdan",
 	"Speech recognition error: {{error}}": "Sayop sa pag-ila sa tingog: {{error}}",
 	"Speech-to-Text Engine": "Engine sa pag-ila sa tingog",
+	"Speed Rate": "",
 	"Stop Sequence": "Pagkasunod-sunod sa pagsira",
+	"Stream Chat Response": "",
 	"STT Model": "",
 	"STT Settings": "Mga setting sa STT",
 	"Submit": "Isumite",

src/lib/i18n/locales/de-DE/translation.json

@@ -73,7 +73,10 @@
 	"AUTOMATIC1111 Api Auth String": "AUTOMATIC1111-API-Authentifizierungszeichenfolge",
 	"AUTOMATIC1111 Base URL": "AUTOMATIC1111-Basis-URL",
 	"AUTOMATIC1111 Base URL is required.": "AUTOMATIC1111-Basis-URL ist erforderlich.",
+	"Available list": "",
 	"available!": "Verfügbar!",
+	"Azure AI Speech": "",
+	"Azure Region": "",
 	"Back": "Zurück",
 	"Bad Response": "Schlechte Antwort",
 	"Banners": "Banner",
@@ -94,6 +97,7 @@
 	"Chat Bubble UI": "Chat Bubble UI",
 	"Chat Controls": "",
 	"Chat direction": "Textrichtung",
+	"Chat Overview": "",
 	"Chats": "Unterhaltungen",
 	"Check Again": "Erneut überprüfen",
 	"Check for updates": "Nach Updates suchen",
@@ -230,6 +234,7 @@
 	"Enable Message Rating": "",
 	"Enable New Sign Ups": "Registrierung erlauben",
 	"Enable Web Search": "Websuche aktivieren",
+	"Enable Web Search Query Generation": "",
 	"Enabled": "",
 	"Engine": "Engine",
 	"Ensure your CSV file includes 4 columns in this order: Name, Email, Password, Role.": "Stellen Sie sicher, dass Ihre CSV-Datei 4 Spalten in dieser Reihenfolge enthält: Name, E-Mail, Passwort, Rolle.",
@@ -237,6 +242,7 @@
 	"Enter a detail about yourself for your LLMs to recall": "Geben Sie ein Detail über sich selbst ein, das Ihre Sprachmodelle (LLMs) sich merken sollen",
 	"Enter api auth string (e.g. username:password)": "Geben Sie die API-Authentifizierungszeichenfolge ein (z. B. Benutzername:Passwort)",
 	"Enter Brave Search API Key": "Geben Sie den Brave Search API-Schlüssel ein",
+	"Enter CFG Scale (e.g. 7.0)": "",
 	"Enter Chunk Overlap": "Geben Sie die Blocküberlappung ein",
 	"Enter Chunk Size": "Geben Sie die Blockgröße ein",
 	"Enter Github Raw URL": "Geben Sie die Github Raw-URL ein",
@@ -247,6 +253,8 @@
 	"Enter Model ID": "",
 	"Enter model tag (e.g. {{modelTag}})": "Gebn Sie den Model-Tag ein",
 	"Enter Number of Steps (e.g. 50)": "Geben Sie die Anzahl an Schritten ein (z. B. 50)",
+	"Enter Sampler (e.g. Euler a)": "",
+	"Enter Scheduler (e.g. Karras)": "",
 	"Enter Score": "Punktzahl eingeben",
 	"Enter SearchApi API Key": "",
 	"Enter SearchApi Engine": "",
@@ -367,6 +375,7 @@
 	"Last Active": "Zuletzt aktiv",
 	"Last Modified": "Zuletzt bearbeitet",
 	"Leave empty for unlimited": "",
+	"Leave empty to use the default prompt, or enter a custom prompt": "",
 	"Light": "Hell",
 	"Listening...": "Höre zu...",
 	"LLMs can make mistakes. Verify important information.": "LLMs können Fehler machen. Überprüfe wichtige Informationen.",
@@ -406,6 +415,8 @@
 	"Model {{modelId}} not found": "Modell {{modelId}} nicht gefunden",
 	"Model {{modelName}} is not vision capable": "Das Modell {{modelName}} ist nicht für die Bildverarbeitung geeignet",
 	"Model {{name}} is now {{status}}": "Modell {{name}} ist jetzt {{status}}",
+	"Model {{name}} is now at the top": "",
+	"Model accepts image inputs": "",
 	"Model created successfully!": "Modell erfolgreich erstellt!",
 	"Model filesystem path detected. Model shortname is required for update, cannot continue.": "Modell-Dateisystempfad erkannt. Modellkurzname ist für das Update erforderlich, Fortsetzung nicht möglich.",
 	"Model ID": "Modell-ID",
@@ -417,6 +428,7 @@
 	"Modelfile Content": "Modelfile-Inhalt",
 	"Models": "Modelle",
 	"More": "Mehr",
+	"Move to Top": "",
 	"Name": "Name",
 	"Name Tag": "Namens-Tag",
 	"Name your model": "Benennen Sie Ihr Modell",
@@ -462,6 +474,8 @@
 	"OpenAI URL/Key required.": "OpenAI-URL/Schlüssel erforderlich.",
 	"or": "oder",
 	"Other": "Andere",
+	"Output format": "",
+	"Overview": "",
 	"Password": "Passwort",
 	"PDF document (.pdf)": "PDF-Dokument (.pdf)",
 	"PDF Extract Images (OCR)": "Text von Bildern aus PDFs extrahieren (OCR)",
@@ -524,6 +538,7 @@
 	"Save": "Speichern",
 	"Save & Create": "Erstellen",
 	"Save & Update": "Aktualisieren",
+	"Save As Copy": "",
 	"Save Tag": "Tag speichern",
 	"Saving chat logs directly to your browser's storage is no longer supported. Please take a moment to download and delete your chat logs by clicking the button below. Don't worry, you can easily re-import your chat logs to the backend through": "Das direkte Speichern von Unterhaltungen im Browser-Speicher wird nicht mehr unterstützt. Bitte nehmen Sie einen Moment Zeit, um Ihre Unterhaltungen zu exportieren und zu löschen, indem Sie auf die Schaltfläche unten klicken. Keine Sorge, Sie können Ihre Unterhaltungen problemlos über das Backend wieder importieren.",
 	"Scan": "Scannen",
@@ -538,7 +553,6 @@
 	"Search Models": "Modelle durchsuchen...",
 	"Search Prompts": "Prompts durchsuchen...",
 	"Search Query Generation Prompt": "Suchanfragengenerierungsvorlage",
-	"Search Query Generation Prompt Length Threshold": "Längenschwelle für Suchanfragengenerierung",
 	"Search Result Count": "Anzahl der Suchergebnisse",
 	"Search Tools": "Werkzeuge durchsuchen...",
 	"SearchApi API Key": "",
@@ -567,16 +581,20 @@
 	"Send": "Senden",
 	"Send a Message": "Eine Nachricht senden",
 	"Send message": "Nachricht senden",
+	"Sends `stream_options: { include_usage: true }` in the request.\nSupported providers will return token usage information in the response when set.": "",
 	"September": "September",
 	"Serper API Key": "Serper-API-Schlüssel",
 	"Serply API Key": "Serply-API-Schlüssel",
 	"Serpstack API Key": "Serpstack-API-Schlüssel",
 	"Server connection verified": "Serververbindung überprüft",
 	"Set as default": "Als Standard festlegen",
+	"Set CFG Scale": "",
 	"Set Default Model": "Standardmodell festlegen",
 	"Set embedding model (e.g. {{model}})": "Einbettungsmodell festlegen (z. B. {{model}})",
 	"Set Image Size": "Bildgröße festlegen",
 	"Set reranking model (e.g. {{model}})": "Rerankingmodell festlegen (z. B. {{model}})",
+	"Set Sampler": "",
+	"Set Scheduler": "",
 	"Set Steps": "Schrittgröße festlegen",
 	"Set Task Model": "Aufgabenmodell festlegen",
 	"Set Voice": "Stimme festlegen",
@@ -599,7 +617,9 @@
 	"Source": "Quelle",
 	"Speech recognition error: {{error}}": "Spracherkennungsfehler: {{error}}",
 	"Speech-to-Text Engine": "Sprache-zu-Text-Engine",
+	"Speed Rate": "",
 	"Stop Sequence": "Stop-Sequenz",
+	"Stream Chat Response": "",
 	"STT Model": "STT-Modell",
 	"STT Settings": "STT-Einstellungen",
 	"Submit": "Senden",

src/lib/i18n/locales/dg-DG/translation.json

@@ -73,7 +73,10 @@
 	"AUTOMATIC1111 Api Auth String": "",
 	"AUTOMATIC1111 Base URL": "AUTOMATIC1111 Base URL",
 	"AUTOMATIC1111 Base URL is required.": "AUTOMATIC1111 Base URL is required.",
+	"Available list": "",
 	"available!": "available! So excite!",
+	"Azure AI Speech": "",
+	"Azure Region": "",
 	"Back": "Back",
 	"Bad Response": "",
 	"Banners": "",
@@ -94,6 +97,7 @@
 	"Chat Bubble UI": "",
 	"Chat Controls": "",
 	"Chat direction": "",
+	"Chat Overview": "",
 	"Chats": "Chats",
 	"Check Again": "Check Again",
 	"Check for updates": "Check for updates",
@@ -230,6 +234,7 @@
 	"Enable Message Rating": "",
 	"Enable New Sign Ups": "Enable New Bark Ups",
 	"Enable Web Search": "",
+	"Enable Web Search Query Generation": "",
 	"Enabled": "",
 	"Engine": "",
 	"Ensure your CSV file includes 4 columns in this order: Name, Email, Password, Role.": "",
@@ -237,6 +242,7 @@
 	"Enter a detail about yourself for your LLMs to recall": "",
 	"Enter api auth string (e.g. username:password)": "",
 	"Enter Brave Search API Key": "",
+	"Enter CFG Scale (e.g. 7.0)": "",
 	"Enter Chunk Overlap": "Enter Overlap of Chunks",
 	"Enter Chunk Size": "Enter Size of Chunk",
 	"Enter Github Raw URL": "",
@@ -247,6 +253,8 @@
 	"Enter Model ID": "",
 	"Enter model tag (e.g. {{modelTag}})": "Enter model doge tag (e.g. {{modelTag}})",
 	"Enter Number of Steps (e.g. 50)": "Enter Number of Steps (e.g. 50)",
+	"Enter Sampler (e.g. Euler a)": "",
+	"Enter Scheduler (e.g. Karras)": "",
 	"Enter Score": "",
 	"Enter SearchApi API Key": "",
 	"Enter SearchApi Engine": "",
@@ -367,6 +375,7 @@
 	"Last Active": "",
 	"Last Modified": "",
 	"Leave empty for unlimited": "",
+	"Leave empty to use the default prompt, or enter a custom prompt": "",
 	"Light": "Light",
 	"Listening...": "",
 	"LLMs can make mistakes. Verify important information.": "LLMs can make borks. Verify important info.",
@@ -406,6 +415,8 @@
 	"Model {{modelId}} not found": "Model {{modelId}} not found",
 	"Model {{modelName}} is not vision capable": "",
 	"Model {{name}} is now {{status}}": "",
+	"Model {{name}} is now at the top": "",
+	"Model accepts image inputs": "",
 	"Model created successfully!": "",
 	"Model filesystem path detected. Model shortname is required for update, cannot continue.": "Model filesystem bark detected. Model shortname is required for update, cannot continue.",
 	"Model ID": "",
@@ -417,6 +428,7 @@
 	"Modelfile Content": "Modelfile Content",
 	"Models": "Wowdels",
 	"More": "",
+	"Move to Top": "",
 	"Name": "Name",
 	"Name Tag": "Name Tag",
 	"Name your model": "",
@@ -462,6 +474,8 @@
 	"OpenAI URL/Key required.": "",
 	"or": "or",
 	"Other": "",
+	"Output format": "",
+	"Overview": "",
 	"Password": "Barkword",
 	"PDF document (.pdf)": "",
 	"PDF Extract Images (OCR)": "PDF Extract Wowmages (OCR)",
@@ -524,6 +538,7 @@
 	"Save": "Save much wow",
 	"Save & Create": "Save & Create much create",
 	"Save & Update": "Save & Update much update",
+	"Save As Copy": "",
 	"Save Tag": "",
 	"Saving chat logs directly to your browser's storage is no longer supported. Please take a moment to download and delete your chat logs by clicking the button below. Don't worry, you can easily re-import your chat logs to the backend through": "Saving chat logs in browser storage not support anymore. Pls download and delete your chat logs by click button below. Much easy re-import to backend through",
 	"Scan": "Scan much scan",
@@ -538,7 +553,6 @@
 	"Search Models": "",
 	"Search Prompts": "Search Prompts much wow",
 	"Search Query Generation Prompt": "",
-	"Search Query Generation Prompt Length Threshold": "",
 	"Search Result Count": "",
 	"Search Tools": "",
 	"SearchApi API Key": "",
@@ -569,16 +583,20 @@
 	"Send": "",
 	"Send a Message": "Send a Message much message",
 	"Send message": "Send message very send",
+	"Sends `stream_options: { include_usage: true }` in the request.\nSupported providers will return token usage information in the response when set.": "",
 	"September": "",
 	"Serper API Key": "",
 	"Serply API Key": "",
 	"Serpstack API Key": "",
 	"Server connection verified": "Server connection verified much secure",
 	"Set as default": "Set as default very default",
+	"Set CFG Scale": "",
 	"Set Default Model": "Set Default Model much model",
 	"Set embedding model (e.g. {{model}})": "",
 	"Set Image Size": "Set Image Size very size",
 	"Set reranking model (e.g. {{model}})": "",
+	"Set Sampler": "",
+	"Set Scheduler": "",
 	"Set Steps": "Set Steps so many steps",
 	"Set Task Model": "",
 	"Set Voice": "Set Voice so speak",
@@ -601,7 +619,9 @@
 	"Source": "Source",
 	"Speech recognition error: {{error}}": "Speech recognition error: {{error}} so error",
 	"Speech-to-Text Engine": "Speech-to-Text Engine much speak",
+	"Speed Rate": "",
 	"Stop Sequence": "Stop Sequence much stop",
+	"Stream Chat Response": "",
 	"STT Model": "",
 	"STT Settings": "STT Settings very settings",
 	"Submit": "Submit much submit",

src/lib/i18n/locales/en-GB/translation.json

@@ -73,7 +73,10 @@
 	"AUTOMATIC1111 Api Auth String": "",
 	"AUTOMATIC1111 Base URL": "",
 	"AUTOMATIC1111 Base URL is required.": "",
+	"Available list": "",
 	"available!": "",
+	"Azure AI Speech": "",
+	"Azure Region": "",
 	"Back": "",
 	"Bad Response": "",
 	"Banners": "",
@@ -94,6 +97,7 @@
 	"Chat Bubble UI": "",
 	"Chat Controls": "",
 	"Chat direction": "",
+	"Chat Overview": "",
 	"Chats": "",
 	"Check Again": "",
 	"Check for updates": "",
@@ -230,6 +234,7 @@
 	"Enable Message Rating": "",
 	"Enable New Sign Ups": "",
 	"Enable Web Search": "",
+	"Enable Web Search Query Generation": "",
 	"Enabled": "",
 	"Engine": "",
 	"Ensure your CSV file includes 4 columns in this order: Name, Email, Password, Role.": "",
@@ -237,6 +242,7 @@
 	"Enter a detail about yourself for your LLMs to recall": "",
 	"Enter api auth string (e.g. username:password)": "",
 	"Enter Brave Search API Key": "",
+	"Enter CFG Scale (e.g. 7.0)": "",
 	"Enter Chunk Overlap": "",
 	"Enter Chunk Size": "",
 	"Enter Github Raw URL": "",
@@ -247,6 +253,8 @@
 	"Enter Model ID": "",
 	"Enter model tag (e.g. {{modelTag}})": "",
 	"Enter Number of Steps (e.g. 50)": "",
+	"Enter Sampler (e.g. Euler a)": "",
+	"Enter Scheduler (e.g. Karras)": "",
 	"Enter Score": "",
 	"Enter SearchApi API Key": "",
 	"Enter SearchApi Engine": "",
@@ -367,6 +375,7 @@
 	"Last Active": "",
 	"Last Modified": "",
 	"Leave empty for unlimited": "",
+	"Leave empty to use the default prompt, or enter a custom prompt": "",
 	"Light": "",
 	"Listening...": "",
 	"LLMs can make mistakes. Verify important information.": "",
@@ -406,6 +415,8 @@
 	"Model {{modelId}} not found": "",
 	"Model {{modelName}} is not vision capable": "",
 	"Model {{name}} is now {{status}}": "",
+	"Model {{name}} is now at the top": "",
+	"Model accepts image inputs": "",
 	"Model created successfully!": "",
 	"Model filesystem path detected. Model shortname is required for update, cannot continue.": "",
 	"Model ID": "",
@@ -417,6 +428,7 @@
 	"Modelfile Content": "",
 	"Models": "",
 	"More": "",
+	"Move to Top": "",
 	"Name": "",
 	"Name Tag": "",
 	"Name your model": "",
@@ -462,6 +474,8 @@
 	"OpenAI URL/Key required.": "",
 	"or": "",
 	"Other": "",
+	"Output format": "",
+	"Overview": "",
 	"Password": "",
 	"PDF document (.pdf)": "",
 	"PDF Extract Images (OCR)": "",
@@ -524,6 +538,7 @@
 	"Save": "",
 	"Save & Create": "",
 	"Save & Update": "",
+	"Save As Copy": "",
 	"Save Tag": "",
 	"Saving chat logs directly to your browser's storage is no longer supported. Please take a moment to download and delete your chat logs by clicking the button below. Don't worry, you can easily re-import your chat logs to the backend through": "",
 	"Scan": "",
@@ -538,7 +553,6 @@
 	"Search Models": "",
 	"Search Prompts": "",
 	"Search Query Generation Prompt": "",
-	"Search Query Generation Prompt Length Threshold": "",
 	"Search Result Count": "",
 	"Search Tools": "",
 	"SearchApi API Key": "",
@@ -567,16 +581,20 @@
 	"Send": "",
 	"Send a Message": "",
 	"Send message": "",
+	"Sends `stream_options: { include_usage: true }` in the request.\nSupported providers will return token usage information in the response when set.": "",
 	"September": "",
 	"Serper API Key": "",
 	"Serply API Key": "",
 	"Serpstack API Key": "",
 	"Server connection verified": "",
 	"Set as default": "",
+	"Set CFG Scale": "",
 	"Set Default Model": "",
 	"Set embedding model (e.g. {{model}})": "",
 	"Set Image Size": "",
 	"Set reranking model (e.g. {{model}})": "",
+	"Set Sampler": "",
+	"Set Scheduler": "",
 	"Set Steps": "",
 	"Set Task Model": "",
 	"Set Voice": "",
@@ -599,7 +617,9 @@
 	"Source": "",
 	"Speech recognition error: {{error}}": "",
 	"Speech-to-Text Engine": "",
+	"Speed Rate": "",
 	"Stop Sequence": "",
+	"Stream Chat Response": "",
 	"STT Model": "",
 	"STT Settings": "",
 	"Submit": "",

src/lib/i18n/locales/en-US/translation.json

@@ -73,7 +73,10 @@
 	"AUTOMATIC1111 Api Auth String": "",
 	"AUTOMATIC1111 Base URL": "",
 	"AUTOMATIC1111 Base URL is required.": "",
+	"Available list": "",
 	"available!": "",
+	"Azure AI Speech": "",
+	"Azure Region": "",
 	"Back": "",
 	"Bad Response": "",
 	"Banners": "",
@@ -94,6 +97,7 @@
 	"Chat Bubble UI": "",
 	"Chat Controls": "",
 	"Chat direction": "",
+	"Chat Overview": "",
 	"Chats": "",
 	"Check Again": "",
 	"Check for updates": "",
@@ -230,6 +234,7 @@
 	"Enable Message Rating": "",
 	"Enable New Sign Ups": "",
 	"Enable Web Search": "",
+	"Enable Web Search Query Generation": "",
 	"Enabled": "",
 	"Engine": "",
 	"Ensure your CSV file includes 4 columns in this order: Name, Email, Password, Role.": "",
@@ -237,6 +242,7 @@
 	"Enter a detail about yourself for your LLMs to recall": "",
 	"Enter api auth string (e.g. username:password)": "",
 	"Enter Brave Search API Key": "",
+	"Enter CFG Scale (e.g. 7.0)": "",
 	"Enter Chunk Overlap": "",
 	"Enter Chunk Size": "",
 	"Enter Github Raw URL": "",
@@ -247,6 +253,8 @@
 	"Enter Model ID": "",
 	"Enter model tag (e.g. {{modelTag}})": "",
 	"Enter Number of Steps (e.g. 50)": "",
+	"Enter Sampler (e.g. Euler a)": "",
+	"Enter Scheduler (e.g. Karras)": "",
 	"Enter Score": "",
 	"Enter SearchApi API Key": "",
 	"Enter SearchApi Engine": "",
@@ -367,6 +375,7 @@
 	"Last Active": "",
 	"Last Modified": "",
 	"Leave empty for unlimited": "",
+	"Leave empty to use the default prompt, or enter a custom prompt": "",
 	"Light": "",
 	"Listening...": "",
 	"LLMs can make mistakes. Verify important information.": "",
@@ -406,6 +415,8 @@
 	"Model {{modelId}} not found": "",
 	"Model {{modelName}} is not vision capable": "",
 	"Model {{name}} is now {{status}}": "",
+	"Model {{name}} is now at the top": "",
+	"Model accepts image inputs": "",
 	"Model created successfully!": "",
 	"Model filesystem path detected. Model shortname is required for update, cannot continue.": "",
 	"Model ID": "",
@@ -417,6 +428,7 @@
 	"Modelfile Content": "",
 	"Models": "",
 	"More": "",
+	"Move to Top": "",
 	"Name": "",
 	"Name Tag": "",
 	"Name your model": "",
@@ -462,6 +474,8 @@
 	"OpenAI URL/Key required.": "",
 	"or": "",
 	"Other": "",
+	"Output format": "",
+	"Overview": "",
 	"Password": "",
 	"PDF document (.pdf)": "",
 	"PDF Extract Images (OCR)": "",
@@ -524,6 +538,7 @@
 	"Save": "",
 	"Save & Create": "",
 	"Save & Update": "",
+	"Save As Copy": "",
 	"Save Tag": "",
 	"Saving chat logs directly to your browser's storage is no longer supported. Please take a moment to download and delete your chat logs by clicking the button below. Don't worry, you can easily re-import your chat logs to the backend through": "",
 	"Scan": "",
@@ -538,7 +553,6 @@
 	"Search Models": "",
 	"Search Prompts": "",
 	"Search Query Generation Prompt": "",
-	"Search Query Generation Prompt Length Threshold": "",
 	"Search Result Count": "",
 	"Search Tools": "",
 	"SearchApi API Key": "",
@@ -567,16 +581,20 @@
 	"Send": "",
 	"Send a Message": "",
 	"Send message": "",
+	"Sends `stream_options: { include_usage: true }` in the request.\nSupported providers will return token usage information in the response when set.": "",
 	"September": "",
 	"Serper API Key": "",
 	"Serply API Key": "",
 	"Serpstack API Key": "",
 	"Server connection verified": "",
 	"Set as default": "",
+	"Set CFG Scale": "",
 	"Set Default Model": "",
 	"Set embedding model (e.g. {{model}})": "",
 	"Set Image Size": "",
 	"Set reranking model (e.g. {{model}})": "",
+	"Set Sampler": "",
+	"Set Scheduler": "",
 	"Set Steps": "",
 	"Set Task Model": "",
 	"Set Voice": "",
@@ -599,7 +617,9 @@
 	"Source": "",
 	"Speech recognition error: {{error}}": "",
 	"Speech-to-Text Engine": "",
+	"Speed Rate": "",
 	"Stop Sequence": "",
+	"Stream Chat Response": "",
 	"STT Model": "",
 	"STT Settings": "",
 	"Submit": "",

src/lib/i18n/locales/es-ES/translation.json

@@ -73,7 +73,10 @@
 	"AUTOMATIC1111 Api Auth String": "",
 	"AUTOMATIC1111 Base URL": "Dirección URL de AUTOMATIC1111",
 	"AUTOMATIC1111 Base URL is required.": "La dirección URL de AUTOMATIC1111 es requerida.",
+	"Available list": "",
 	"available!": "¡disponible!",
+	"Azure AI Speech": "",
+	"Azure Region": "",
 	"Back": "Volver",
 	"Bad Response": "Respuesta incorrecta",
 	"Banners": "Banners",
@@ -94,6 +97,7 @@
 	"Chat Bubble UI": "Burbuja de chat UI",
 	"Chat Controls": "",
 	"Chat direction": "Dirección del Chat",
+	"Chat Overview": "",
 	"Chats": "Chats",
 	"Check Again": "Verifica de nuevo",
 	"Check for updates": "Verificar actualizaciones",
@@ -230,6 +234,7 @@
 	"Enable Message Rating": "",
 	"Enable New Sign Ups": "Habilitar Nuevos Registros",
 	"Enable Web Search": "Habilitar la búsqueda web",
+	"Enable Web Search Query Generation": "",
 	"Enabled": "",
 	"Engine": "",
 	"Ensure your CSV file includes 4 columns in this order: Name, Email, Password, Role.": "Asegúrese de que su archivo CSV incluya 4 columnas en este orden: Nombre, Correo Electrónico, Contraseña, Rol.",
@@ -237,6 +242,7 @@
 	"Enter a detail about yourself for your LLMs to recall": "Ingrese un detalle sobre usted para que sus LLMs recuerden",
 	"Enter api auth string (e.g. username:password)": "",
 	"Enter Brave Search API Key": "Ingresa la clave de API de Brave Search",
+	"Enter CFG Scale (e.g. 7.0)": "",
 	"Enter Chunk Overlap": "Ingresar superposición de fragmentos",
 	"Enter Chunk Size": "Ingrese el tamaño del fragmento",
 	"Enter Github Raw URL": "Ingresa la URL sin procesar de Github",
@@ -247,6 +253,8 @@
 	"Enter Model ID": "",
 	"Enter model tag (e.g. {{modelTag}})": "Ingrese la etiqueta del modelo (p.ej. {{modelTag}})",
 	"Enter Number of Steps (e.g. 50)": "Ingrese el número de pasos (p.ej., 50)",
+	"Enter Sampler (e.g. Euler a)": "",
+	"Enter Scheduler (e.g. Karras)": "",
 	"Enter Score": "Ingrese la puntuación",
 	"Enter SearchApi API Key": "",
 	"Enter SearchApi Engine": "",
@@ -367,6 +375,7 @@
 	"Last Active": "Última Actividad",
 	"Last Modified": "Modificado por última vez",
 	"Leave empty for unlimited": "",
+	"Leave empty to use the default prompt, or enter a custom prompt": "",
 	"Light": "Claro",
 	"Listening...": "Escuchando...",
 	"LLMs can make mistakes. Verify important information.": "Los LLM pueden cometer errores. Verifica la información importante.",
@@ -406,6 +415,8 @@
 	"Model {{modelId}} not found": "El modelo {{modelId}} no fue encontrado",
 	"Model {{modelName}} is not vision capable": "El modelo {{modelName}} no es capaz de ver",
 	"Model {{name}} is now {{status}}": "El modelo {{name}} ahora es {{status}}",
+	"Model {{name}} is now at the top": "",
+	"Model accepts image inputs": "",
 	"Model created successfully!": "Modelo creado correctamente!",
 	"Model filesystem path detected. Model shortname is required for update, cannot continue.": "Se detectó la ruta del sistema de archivos del modelo. Se requiere el nombre corto del modelo para la actualización, no se puede continuar.",
 	"Model ID": "ID del modelo",
@@ -417,6 +428,7 @@
 	"Modelfile Content": "Contenido del Modelfile",
 	"Models": "Modelos",
 	"More": "Más",
+	"Move to Top": "",
 	"Name": "Nombre",
 	"Name Tag": "Nombre de etiqueta",
 	"Name your model": "Asigne un nombre a su modelo",
@@ -462,6 +474,8 @@
 	"OpenAI URL/Key required.": "URL/Clave de OpenAI es requerida.",
 	"or": "o",
 	"Other": "Otro",
+	"Output format": "",
+	"Overview": "",
 	"Password": "Contraseña",
 	"PDF document (.pdf)": "PDF document (.pdf)",
 	"PDF Extract Images (OCR)": "Extraer imágenes de PDF (OCR)",
@@ -524,6 +538,7 @@
 	"Save": "Guardar",
 	"Save & Create": "Guardar y Crear",
 	"Save & Update": "Guardar y Actualizar",
+	"Save As Copy": "",
 	"Save Tag": "",
 	"Saving chat logs directly to your browser's storage is no longer supported. Please take a moment to download and delete your chat logs by clicking the button below. Don't worry, you can easily re-import your chat logs to the backend through": "Ya no se admite guardar registros de chat directamente en el almacenamiento de su navegador. Tómese un momento para descargar y eliminar sus registros de chat haciendo clic en el botón a continuación. No te preocupes, puedes volver a importar fácilmente tus registros de chat al backend a través de",
 	"Scan": "Escanear",
@@ -538,7 +553,6 @@
 	"Search Models": "Modelos de búsqueda",
 	"Search Prompts": "Buscar Prompts",
 	"Search Query Generation Prompt": "Búsqueda de consulta de generación de prompts",
-	"Search Query Generation Prompt Length Threshold": "Nivel de longitud de Búsqueda de consulta de generación de prompts",
 	"Search Result Count": "Recuento de resultados de búsqueda",
 	"Search Tools": "Búsqueda de herramientas",
 	"SearchApi API Key": "",
@@ -568,16 +582,20 @@
 	"Send": "Enviar",
 	"Send a Message": "Enviar un Mensaje",
 	"Send message": "Enviar Mensaje",
+	"Sends `stream_options: { include_usage: true }` in the request.\nSupported providers will return token usage information in the response when set.": "",
 	"September": "Septiembre",
 	"Serper API Key": "Clave API de Serper",
 	"Serply API Key": "Clave API de Serply",
 	"Serpstack API Key": "Clave API de Serpstack",
 	"Server connection verified": "Conexión del servidor verificada",
 	"Set as default": "Establecer por defecto",
+	"Set CFG Scale": "",
 	"Set Default Model": "Establecer modelo predeterminado",
 	"Set embedding model (e.g. {{model}})": "Establecer modelo de embedding (ej. {{model}})",
 	"Set Image Size": "Establecer tamaño de imagen",
 	"Set reranking model (e.g. {{model}})": "Establecer modelo de reranking (ej. {{model}})",
+	"Set Sampler": "",
+	"Set Scheduler": "",
 	"Set Steps": "Establecer Pasos",
 	"Set Task Model": "Establecer modelo de tarea",
 	"Set Voice": "Establecer la voz",
@@ -600,7 +618,9 @@
 	"Source": "Fuente",
 	"Speech recognition error: {{error}}": "Error de reconocimiento de voz: {{error}}",
 	"Speech-to-Text Engine": "Motor de voz a texto",
+	"Speed Rate": "",
 	"Stop Sequence": "Detener secuencia",
+	"Stream Chat Response": "",
 	"STT Model": "Modelo STT",
 	"STT Settings": "Configuraciones de STT",
 	"Submit": "Enviar",