Merge pull request #6460 from open-webui/dev

0.3.35
doc: changelog
2024-10-26 13:17:35 -07:00 · 2024-10-26 13:16:49 -07:00 · 2024-10-26 13:05:54 -07:00 · 2024-10-26 13:05:37 -07:00 · 2024-10-26 12:58:14 -07:00 · 2024-10-26 12:56:37 -07:00
358 changed files with 36443 additions and 11184 deletions
@@ -8,6 +8,20 @@ assignees: ''

 # Bug Report

+## Important Notes
+
+- **Before submitting a bug report**: Please check the Issues or Discussions section to see if a similar issue or feature request has already been posted. It's likely we're already tracking it! If you’re unsure, start a discussion post first. This will help us efficiently focus on improving the project.
+
+- **Collaborate respectfully**: We value a constructive attitude, so please be mindful of your communication. If negativity is part of your approach, our capacity to engage may be limited. We’re here to help if you’re open to learning and communicating positively. Remember, Open WebUI is a volunteer-driven project managed by a single maintainer and supported by contributors who also have full-time jobs. We appreciate your time and ask that you respect ours.
+
+- **Contributing**: If you encounter an issue, we highly encourage you to submit a pull request or fork the project. We actively work to prevent contributor burnout to maintain the quality and continuity of Open WebUI.
+
+- **Bug reproducibility**: If a bug cannot be reproduced with a `:main` or `:dev` Docker setup, or a pip install with Python 3.11, it may require additional help from the community. In such cases, we will move it to the "issues" Discussions section due to our limited resources. We encourage the community to assist with these issues. Remember, it’s not that the issue doesn’t exist; we need your help!
+
+Note: Please remove the notes above when submitting your post. Thank you for your understanding and support!
+
+---
+
 ## Installation Method

 [Describe the method you used to install the project, e.g., git clone, Docker, pip, etc.]
@@ -6,6 +6,22 @@ labels: ''
 assignees: ''
 ---

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

@@ -23,7 +23,7 @@ jobs:
      - uses: actions/checkout@v4

      - name: Set up Python
-        uses: actions/setup-python@v4
+        uses: actions/setup-python@v5
        with:
          python-version: ${{ matrix.python-version }}

@@ -21,7 +21,7 @@ jobs:
      - name: Setup Node.js
        uses: actions/setup-node@v4
        with:
-          node-version: '20' # Or specify any other version you want to use
+          node-version: '22' # Or specify any other version you want to use

      - name: Install Dependencies
        run: npm install
@@ -48,7 +48,7 @@ jobs:
      - name: Setup Node.js
        uses: actions/setup-node@v4
        with:
-          node-version: '20'
+          node-version: '22'

      - name: Install Dependencies
        run: npm ci
@@ -85,7 +85,7 @@ jobs:
  #     - uses: actions/checkout@v4

  #     - name: Set up Python
-  #       uses: actions/setup-python@v4
+  #       uses: actions/setup-python@v5
  #       with:
  #         python-version: ${{ matrix.python-version }}

@@ -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=$!
          # Wait up to 40 seconds for the server to start
          for i in {1..40}; do
@@ -182,9 +182,12 @@ jobs:
          WEBUI_SECRET_KEY: secret-key
          GLOBAL_LOG_LEVEL: debug
          DATABASE_URL: postgresql://postgres:postgres@localhost:5432/postgres
+          DATABASE_POOL_SIZE: 10
+          DATABASE_POOL_MAX_OVERFLOW: 10
+          DATABASE_POOL_TIMEOUT: 30
        run: |
          cd backend
-          uvicorn main:app --port "8081" --forwarded-allow-ips '*' &
+          uvicorn open_webui.main:app --port "8081" --forwarded-allow-ips '*' &
          UVICORN_PID=$!
          # Wait up to 20 seconds for the server to start
          for i in {1..20}; do
@@ -230,7 +233,7 @@ 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
@@ -16,7 +16,7 @@ jobs:
    steps:
      - uses: actions/checkout@v4
      - name: Use Python
-        uses: actions/setup-python@v4
+        uses: actions/setup-python@v5
      - name: Use Bun
        uses: oven-sh/setup-bun@v1
      - name: Install dependencies
@@ -5,6 +5,313 @@ 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).

+## [0.3.35] - 2024-10-26
+
+### Added
+
+- **📁 Robust File Handling**: Enhanced file input handling for chat. If the content extraction fails or is empty, users will now receive a clear warning, preventing silent failures and ensuring you always know what's happening with your uploads.
+- **🌍 New Language Support**: Introduced Hungarian translations and updated French translations, expanding the platform's language accessibility for a more global user base.
+
+### Fixed
+
+- **📚 Knowledge Base Loading Issue**: Resolved a critical bug where the Knowledge Base was not loading, ensuring smooth access to your stored documents and improving information retrieval in RAG-enhanced workflows.
+- **🛠️ Tool Parameters Issue**: Fixed an error where tools were not functioning correctly when required parameters were missing, ensuring reliable tool performance and more efficient task completions.
+- **🔗 Merged Response Loss in Multi-Model Chats**: Addressed an issue where responses in multi-model chat workflows were being deleted after follow-up queries, improving consistency and ensuring smoother interactions across models.
+
+## [0.3.34] - 2024-10-26
+
+### Added
+
+- **🔧 Feedback Export Enhancements**: Feedback history data can now be exported to JSON, allowing for seamless integration in RLHF processing and further analysis.
+- **🗂️ Embedding Model Lazy Loading**: Search functionality for leaderboard reranking is now more efficient, as embedding models are lazy-loaded only when needed, optimizing performance.
+- **🎨 Rich Text Input Toggle**: Users can now switch back to legacy textarea input for chat if they prefer simpler text input, though rich text is still the default until deprecation.
+- **🛠️ Improved Tool Calling Mechanism**: Enhanced method for parsing and calling tools, improving the reliability and robustness of tool function calls.
+- **🌐 Globalization Enhancements**: Updates to internationalization (i18n) support, further refining multi-language compatibility and accuracy.
+
+### Fixed
+
+- **🖥️ Folder Rename Fix for Firefox**: Addressed a persistent issue where users could not rename folders by pressing enter in Firefox, now ensuring seamless folder management across browsers.
+- **🔠 Tiktoken Model Text Splitter Issue**: Resolved an issue where the tiktoken text splitter wasn’t working in Docker installations, restoring full functionality for tokenized text editing.
+- **💼 S3 File Upload Issue**: Fixed a problem affecting S3 file uploads, ensuring smooth operations for those who store files on cloud storage.
+- **🔒 Strict-Transport-Security Crash**: Resolved a crash when setting the Strict-Transport-Security (HSTS) header, improving stability and security enhancements.
+- **🚫 OIDC Boolean Access Fix**: Addressed an issue with boolean values not being accessed correctly during OIDC logins, ensuring login reliability.
+- **⚙️ Rich Text Paste Behavior**: Refined paste behavior in rich text input to make it smoother and more intuitive when pasting various content types.
+- **🔨 Model Exclusion for Arena Fix**: Corrected the filter function that was not properly excluding models from the arena, improving model management.
+- **🏷️ "Tags Generation Prompt" Fix**: Addressed an issue preventing custom "tags generation prompts" from registering properly, ensuring custom prompt work seamlessly.
+
+## [0.3.33] - 2024-10-24
+
+### Added
+
+- **🏆 Evaluation Leaderboard**: Easily track your performance through a new leaderboard system where your ratings contribute to a real-time ranking based on the Elo system. Sibling responses (regenerations, many model chats) are required for your ratings to count in the leaderboard. Additionally, you can opt-in to share your feedback history and be part of the community-wide leaderboard. Expect further improvements as we refine the algorithm—help us build the best community leaderboard!
+- **⚔️ Arena Model Evaluation**: Enable blind A/B testing of models directly from Admin Settings > Evaluation for a true side-by-side comparison. Ideal for pinpointing the best model for your needs.
+- **🎯 Topic-Based Leaderboard**: Discover more accurate rankings with experimental topic-based reranking, which adjusts leaderboard standings based on tag similarity in feedback. Get more relevant insights based on specific topics!
+- **📁 Folders Support for Chats**: Organize your chats better by grouping them into folders. Drag and drop chats between folders and export them seamlessly for easy sharing or analysis.
+- **📤 Easy Chat Import via Drag & Drop**: Save time by simply dragging and dropping chat exports (JSON) directly onto the sidebar to import them into your workspace—streamlined, efficient, and intuitive!
+- **📚 Enhanced Knowledge Collection**: Now, you can reference individual files from a knowledge collection—ideal for more precise Retrieval-Augmented Generations (RAG) queries and document analysis.
+- **🏷️ Enhanced Tagging System**: Tags now take up less space! Utilize the new 'tag:' query system to manage, search, and organize your conversations more effectively without cluttering the interface.
+- **🧠 Auto-Tagging for Chats**: Your conversations are now automatically tagged for improved organization, mirroring the efficiency of auto-generated titles.
+- **🔍 Backend Chat Query System**: Chat filtering has become more efficient, now handled through the backend\*\* instead of your browser, improving search performance and accuracy.
+- **🎮 Revamped Playground**: Experience a refreshed and optimized Playground for smoother testing, tweaks, and experimentation of your models and tools.
+- **🧩 Token-Based Text Splitter**: Introducing token-based text splitting (tiktoken), giving you more precise control over how text is processed. Previously, only character-based splitting was available.
+- **🔢 Ollama Batch Embeddings**: Leverage new batch embedding support for improved efficiency and performance with Ollama embedding models.
+- **🔍 Enhanced Add Text Content Modal**: Enjoy a cleaner, more intuitive workflow for adding and curating knowledge content with an upgraded input modal from our Knowledge workspace.
+- **🖋️ Rich Text Input for Chats**: Make your chat inputs more dynamic with support for rich text formatting. Your conversations just got a lot more polished and professional.
+- **⚡ Faster Whisper Model Configurability**: Customize your local faster whisper model directly from the WebUI.
+- **☁️ Experimental S3 Support**: Enable stateless WebUI instances with S3 support, greatly enhancing scalability and balancing heavy workloads.
+- **🔕 Disable Update Toast**: Now you can streamline your workspace even further—choose to disable update notifications for a more focused experience.
+- **🌟 RAG Citation Relevance Percentage**: Easily assess citation accuracy with the addition of relevance percentages in RAG results.
+- **⚙️ Mermaid Copy Button**: Mermaid diagrams now come with a handy copy button, simplifying the extraction and use of diagram contents directly in your workflow.
+- **🎨 UI Redesign**: Major interface redesign that will make navigation smoother, keep your focus where it matters, and ensure a modern look.
+
+### Fixed
+
+- **🎙️ Voice Note Mic Stopping Issue**: Fixed the issue where the microphone stayed active after ending a voice note recording, ensuring your audio workflow runs smoothly.
+
+### Removed
+
+- **👋 Goodbye Sidebar Tags**: Sidebar tag clutter is gone. We’ve shifted tag buttons to more effective query-based tag filtering for a sleeker, more agile interface.
+
+## [0.3.32] - 2024-10-06
+
+### Added
+
+- **🔢 Workspace Enhancements**: Added a display count for models, prompts, tools, and functions in the workspace, providing a clear overview and easier management.
+
+### Fixed
+
+- **🖥️ Web and YouTube Attachment Fix**: Resolved an issue where attaching web links and YouTube videos was malfunctioning, ensuring seamless integration and display within chats.
+- **📞 Call Mode Activation on Landing Page**: Fixed a bug where call mode was not operational from the landing page.
+
+### Changed
+
+- **🔄 URL Parameter Refinement**: Updated the 'tool_ids' URL parameter to 'tools' or 'tool-ids' for more intuitive and consistent user experience.
+- **🎨 Floating Buttons Styling Update**: Refactored the styling of floating buttons to intelligently adjust to the left side when there isn't enough room on the right, improving interface usability and aesthetic.
+- **🔧 Enhanced Accessibility for Floating Buttons**: Implemented the ability to close floating buttons with the 'Esc' key, making workflow smoother and more efficient for users navigating via keyboard.
+- **🖇️ Updated Information URL**: Information URLs now direct users to a general release page rather than a version-specific URL, ensuring access to the latest and relevant details all in one place.
+- **📦 Library Dependencies Update**: Upgraded dependencies to ensure compatibility and performance optimization for pip installs.
+
+## [0.3.31] - 2024-10-06
+
+### Added
+
+- **📚 Knowledge Feature**: Reimagined documents feature, now more performant with a better UI for enhanced organization; includes streamlined API integration for Retrieval-Augmented Generation (RAG). Detailed documentation forthcoming: https://docs.openwebui.com/
+- **🌐 New Landing Page**: Freshly designed landing page; toggle between the new UI and the classic chat UI from Settings > Interface for a personalized experience.
+- **📁 Full Document Retrieval Mode**: Toggle between full document retrieval or traditional snippets by clicking on the file item. This mode enhances document capabilities and supports comprehensive tasks like summarization by utilizing the entire content instead of RAG.
+- **📄 Extracted File Content Display**: View extracted content directly by clicking on the file item, simplifying file analysis.
+- **🎨 Artifacts Feature**: Render web content and SVGs directly in the interface, supporting quick iterations and live changes.
+- **🖊️ Editable Code Blocks**: Supercharged code blocks now allow live editing directly in the LLM response, with live reloads supported by artifacts.
+- **🔧 Code Block Enhancements**: Introduced a floating copy button in code blocks to facilitate easier code copying without scrolling.
+- **🔍 SVG Pan/Zoom**: Enhanced interaction with SVG images, including Mermaid diagrams, via new pan and zoom capabilities.
+- **🔍 Text Select Quick Actions**: New floating buttons appear when text is highlighted in LLM responses, offering deeper interactions like "Ask a Question" or "Explain".
+- **🗃️ Database Pool Configuration**: Enhanced database handling to support scalable user growth.
+- **🔊 Experimental Audio Compression**: Compress audio files to navigate around the 25MB limit for OpenAI's speech-to-text processing.
+- **🔍 Query Embedding**: Adjusted embedding behavior to enhance system performance by not repeating query embedding.
+- **💾 Lazy Load Optimizations**: Implemented lazy loading of large dependencies to minimize initial memory usage, boosting performance.
+- **🍏 Apple Touch Icon Support**: Optimizes the display of icons for web bookmarks on Apple mobile devices.
+- **🔽 Expandable Content Markdown Support**: Introducing 'details', 'summary' tag support for creating expandable content sections in markdown, facilitating cleaner, organized documentation and interactive content display.
+
+### Fixed
+
+- **🔘 Action Button Issue**: Resolved a bug where action buttons were not functioning, enhancing UI reliability.
+- **🔄 Multi-Model Chat Loop**: Fixed an infinite loop issue in multi-model chat environments, ensuring smoother chat operations.
+- **📄 Chat PDF/TXT Export Issue**: Resolved problems with exporting chat logs to PDF and TXT formats.
+- **🔊 Call to Text-to-Speech Issues**: Rectified problems with text-to-speech functions to improve audio interactions.
+
+### Changed
+
+- **⚙️ Endpoint Renaming**: Renamed 'rag' endpoints to 'retrieval' for clearer function description.
+- **🎨 Styling and Interface Updates**: Multiple refinements across the platform to enhance visual appeal and user interaction.
+
+### Removed
+
+- **🗑️ Deprecated 'DOCS_DIR'**: Removed the outdated 'docs_dir' variable in favor of more direct file management solutions, with direct file directory syncing and API uploads for a more integrated experience.
+
+## [0.3.30] - 2024-09-26
+
+### Fixed
+
+- **🍞 Update Available Toast Dismissal**: Enhanced user experience by ensuring that once the update available notification is dismissed, it won't reappear for 24 hours.
+- **📋 Ollama /embed Form Data**: Adjusted the integration inaccuracies in the /embed form data to ensure it perfectly matches with Ollama's specifications.
+- **🔧 O1 Max Completion Tokens Issue**: Resolved compatibility issues with OpenAI's o1 models max_completion_tokens param to ensure smooth operation.
+- **🔄 Pip Install Database Issue**: Fixed a critical issue where database changes during pip installations were reverting and not saving chat logs, now ensuring data persistence and reliability in chat operations.
+- **🏷️ Chat Rename Tab Update**: Fixed the functionality to change the web browser's tab title simultaneously when a chat is renamed, keeping tab titles consistent.
+
+## [0.3.29] - 2023-09-25
+
+### Fixed
+
+- **🔧 KaTeX Rendering Improvement**: Resolved specific corner cases in KaTeX rendering to enhance the display of complex mathematical notation.
+- **📞 'Call' URL Parameter Fix**: Corrected functionality for 'call' URL search parameter ensuring reliable activation of voice calls through URL triggers.
+- **🔄 Configuration Reset Fix**: Fixed the RESET_CONFIG_ON_START to ensure settings revert to default correctly upon each startup, improving reliability in configuration management.
+- **🌍 Filter Outlet Hook Fix**: Addressed issues in the filter outlet hook, ensuring all filter functions operate as intended.
+
+## [0.3.28] - 2024-09-24
+
+### Fixed
+
+- **🔍 Web Search Functionality**: Corrected an issue where the web search option was not functioning properly.
+
+## [0.3.27] - 2024-09-24
+
+### Fixed
+
+- **🔄 Periodic Cleanup Error Resolved**: Fixed a critical RuntimeError related to the 'periodic_usage_pool_cleanup' coroutine, ensuring smooth and efficient performance post-pip install, correcting a persisting issue from version 0.3.26.
+- **📊 Enhanced LaTeX Rendering**: Improved rendering for LaTeX content, enhancing clarity and visual presentation in documents and mathematical models.
+
+## [0.3.26] - 2024-09-24
+
+### Fixed
+
+- **🔄 Event Loop Error Resolution**: Addressed a critical error where a missing running event loop caused 'periodic_usage_pool_cleanup' to fail with pip installs. This fix ensures smoother and more reliable updates and installations, enhancing overall system stability.
+
+## [0.3.25] - 2024-09-24
+
+### Fixed
+
+- **🖼️ Image Generation Functionality**: Resolved an issue where image generation was not functioning, restoring full capability for visual content creation.
+- **⚖️ Rate Response Corrections**: Addressed a problem where rate responses were not working, ensuring reliable feedback mechanisms are operational.
+
+## [0.3.24] - 2024-09-24
+
+### Added
+
+- **🚀 Rendering Optimization**: Significantly improved message rendering performance, enhancing user experience and webui responsiveness.
+- **💖 Favorite Response Feature in Chat Overview**: Users can now mark responses as favorite directly from the chat overview, enhancing ease of retrieval and organization of preferred responses.
+- **💬 Create Message Pairs with Shortcut**: Implemented creation of new message pairs using Cmd/Ctrl+Shift+Enter, making conversation editing faster and more intuitive.
+- **🌍 Expanded User Prompt Variables**: Added weekday, timezone, and language information variables to user prompts to match system prompt variables.
+- **🎵 Enhanced Audio Support**: Now includes support for 'audio/x-m4a' files, broadening compatibility with audio content within the platform.
+- **🔏 Model URL Search Parameter**: Added an ability to select a model directly via URL parameters, streamlining navigation and model access.
+- **📄 Enhanced PDF Citations**: PDF citations now open at the associated page, streamlining reference checks and document handling.
+- **🔧Use of Redis in Sockets**: Enhanced socket implementation to fully support Redis, enabling effective stateless instances suitable for scalable load balancing.
+- **🌍 Stream Individual Model Responses**: Allows specific models to have individualized streaming settings, enhancing performance and customization.
+- **🕒 Display Model Hash and Last Modified Timestamp for Ollama Models**: Provides critical model details directly in the Models workspace for enhanced tracking.
+- **❗ Update Info Notification for Admins**: Ensures administrators receive immediate updates upon login, keeping them informed of the latest changes and system statuses.
+
+### Fixed
+
+- **🗑️ Temporary File Handling On Windows**: Fixed an issue causing errors when accessing a temporary file being used by another process, Tools & Functions should now work as intended.
+- **🔓 Authentication Toggle Issue**: Resolved the malfunction where setting 'WEBUI_AUTH=False' did not appropriately disable authentication, ensuring that user experience and system security settings function as configured.
+- **🔧 Save As Copy Issue for Many Model Chats**: Resolved an error preventing users from save messages as copies in many model chats.
+- **🔒 Sidebar Closure on Mobile**: Resolved an issue where the mobile sidebar remained open after menu engagement, improving user interface responsivity and comfort.
+- **🛡️ Tooltip XSS Vulnerability**: Resolved a cross-site scripting (XSS) issue within tooltips, ensuring enhanced security and data integrity during user interactions.
+
+### Changed
+
+- **↩️ Deprecated Interface Stream Response Settings**: Moved to advanced parameters to streamline interface settings and enhance user clarity.
+- **⚙️ Renamed 'speedRate' to 'playbackRate'**: Standardizes terminology, improving usability and understanding in media settings.
+
+## [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
@@ -1,6 +1,6 @@
 # syntax=docker/dockerfile:1
 # Initialize device type args
-# use build args in the docker build commmand with --build-arg="BUILDARG=true"
+# use build args in the docker build command with --build-arg="BUILDARG=true"
 ARG USE_CUDA=false
 ARG USE_OLLAMA=false
 # Tested with cu117 for CUDA 11 and cu121 for CUDA 12 (default)
@@ -11,13 +11,17 @@ ARG USE_CUDA_VER=cu121
 # IMPORTANT: If you change the embedding model (sentence-transformers/all-MiniLM-L6-v2) and vice versa, you aren't able to use RAG Chat with your previous documents loaded in the WebUI! You need to re-embed them.
 ARG USE_EMBEDDING_MODEL=sentence-transformers/all-MiniLM-L6-v2
 ARG USE_RERANKING_MODEL=""
+
+# Tiktoken encoding name; models to use can be found at https://huggingface.co/models?library=tiktoken
+ARG USE_TIKTOKEN_ENCODING_NAME="cl100k_base"
+
 ARG BUILD_HASH=dev-build
 # Override at your own risk - non-root configurations are untested
 ARG UID=0
 ARG GID=0

 ######## WebUI frontend ########
-FROM --platform=$BUILDPLATFORM node:21-alpine3.19 as build
+FROM --platform=$BUILDPLATFORM node:22-alpine3.20 AS build
 ARG BUILD_HASH

 WORKDIR /app
@@ -30,7 +34,7 @@ ENV APP_BUILD_HASH=${BUILD_HASH}
 RUN npm run build

 ######## WebUI backend ########
-FROM python:3.11-slim-bookworm as base
+FROM python:3.11-slim-bookworm AS base

 # Use args
 ARG USE_CUDA
@@ -72,13 +76,21 @@ ENV RAG_EMBEDDING_MODEL="$USE_EMBEDDING_MODEL_DOCKER" \
    RAG_RERANKING_MODEL="$USE_RERANKING_MODEL_DOCKER" \
    SENTENCE_TRANSFORMERS_HOME="/app/backend/data/cache/embedding/models"

+## Tiktoken model settings ##
+ENV TIKTOKEN_ENCODING_NAME="cl100k_base" \
+    TIKTOKEN_CACHE_DIR="/app/backend/data/cache/tiktoken"
+
 ## 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

-ENV HOME /root
+ENV HOME=/root
 # Create user and group if not root
 RUN if [ $UID -ne 0 ]; then \
    if [ $GID -ne 0 ]; then \
@@ -96,7 +108,7 @@ 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 +121,7 @@ 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 && \
@@ -127,11 +139,13 @@ RUN pip3 install uv && \
    uv pip install --system -r requirements.txt --no-cache-dir && \
    python -c "import os; from sentence_transformers import SentenceTransformer; SentenceTransformer(os.environ['RAG_EMBEDDING_MODEL'], device='cpu')" && \
    python -c "import os; from faster_whisper import WhisperModel; WhisperModel(os.environ['WHISPER_MODEL'], device='cpu', compute_type='int8', download_root=os.environ['WHISPER_MODEL_DIR'])"; \
+    python -c "import os; import tiktoken; tiktoken.get_encoding(os.environ['TIKTOKEN_ENCODING_NAME'])"; \
    else \
    pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu --no-cache-dir && \
    uv pip install --system -r requirements.txt --no-cache-dir && \
    python -c "import os; from sentence_transformers import SentenceTransformer; SentenceTransformer(os.environ['RAG_EMBEDDING_MODEL'], device='cpu')" && \
    python -c "import os; from faster_whisper import WhisperModel; WhisperModel(os.environ['WHISPER_MODEL'], device='cpu', compute_type='int8', download_root=os.environ['WHISPER_MODEL_DIR'])"; \
+    python -c "import os; import tiktoken; tiktoken.get_encoding(os.environ['TIKTOKEN_ENCODING_NAME'])"; \
    fi; \
    chown -R $UID:$GID /app/backend/data/

@@ -157,5 +171,6 @@ USER $UID:$GID

 ARG BUILD_HASH
 ENV WEBUI_BUILD_VERSION=${BUILD_HASH}
+ENV DOCKER=true

 CMD [ "bash", "start.sh"]
@@ -1,4 +1,4 @@
-# Open WebUI (Formerly Ollama WebUI) 👋
+# Open WebUI 👋

 ![GitHub stars](https://img.shields.io/github/stars/open-webui/open-webui?style=social)
 ![GitHub forks](https://img.shields.io/github/forks/open-webui/open-webui?style=social)
@@ -59,11 +59,31 @@ Don't forget to explore our sibling project, [Open WebUI Community](https://open

 ## How to Install 🚀

-> [!NOTE]  
-> Please note that for certain Docker environments, additional configurations might be needed. If you encounter any connection issues, our detailed guide on [Open WebUI Documentation](https://docs.openwebui.com/) is ready to assist you.
+### Installation via Python pip 🐍
+
+Open WebUI can be installed using pip, the Python package installer. Before proceeding, ensure you're using **Python 3.11** to avoid compatibility issues.
+
+1. **Install Open WebUI**:
+   Open your terminal and run the following command to install Open WebUI:
+
+   ```bash
+   pip install open-webui
+   ```
+
+2. **Running Open WebUI**:
+   After installation, you can start Open WebUI by executing:
+
+   ```bash
+   open-webui serve
+   ```
+
+This will start the Open WebUI server, which you can access at [http://localhost:8080](http://localhost:8080)

 ### Quick Start with Docker 🐳

+> [!NOTE]  
+> Please note that for certain Docker environments, additional configurations might be needed. If you encounter any connection issues, our detailed guide on [Open WebUI Documentation](https://docs.openwebui.com/) is ready to assist you.
+
 > [!WARNING]
 > When using Docker to install Open WebUI, make sure to include the `-v open-webui:/app/backend/data` in your Docker command. This step is crucial as it ensures your database is properly mounted and prevents any loss of data.

@@ -86,7 +106,7 @@ Don't forget to explore our sibling project, [Open WebUI Community](https://open
  docker run -d -p 3000:8080 -e OLLAMA_BASE_URL=https://example.com -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main
  ```

-  - **To run Open WebUI with Nvidia GPU support**, use this command:
+- **To run Open WebUI with Nvidia GPU support**, use this command:

  ```bash
  docker run -d -p 3000:8080 --gpus all --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:cuda
@@ -150,7 +170,7 @@ docker run --rm --volume /var/run/docker.sock:/var/run/docker.sock containrrr/wa

 In the last part of the command, replace `open-webui` with your container name if it is different.

-Check our Migration Guide available in our [Open WebUI Documentation](https://docs.openwebui.com/migration/).
+Check our Migration Guide available in our [Open WebUI Documentation](https://docs.openwebui.com/tutorials/migration/).

 ### Using the Dev Branch 🌙

@@ -200,4 +220,4 @@ If you have any questions, suggestions, or need assistance, please open an issue

 ---

-Created by [Timothy J. Baek](https://github.com/tjbck) - Let's make Open WebUI even more amazing together! 💪
+Created by [Timothy Jaeryang Baek](https://github.com/tjbck) - Let's make Open WebUI even more amazing together! 💪
@@ -18,7 +18,7 @@ If you're experiencing connection issues, it’s often due to the WebUI docker c
 docker run -d --network=host -v open-webui:/app/backend/data -e OLLAMA_BASE_URL=http://127.0.0.1:11434 --name open-webui --restart always ghcr.io/open-webui/open-webui:main
 ```

-### Error on Slow Reponses for Ollama
+### Error on Slow Responses for Ollama

 Open WebUI has a default timeout of 5 minutes for Ollama to finish generating the response. If needed, this can be adjusted via the environment variable AIOHTTP_CLIENT_TIMEOUT, which sets the timeout in seconds.

@@ -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
@@ -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 +39,20 @@ 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="*")
@@ -5,6 +5,8 @@ import os
 import uuid
 from functools import lru_cache
 from pathlib import Path
+from pydub import AudioSegment
+from pydub.silence import split_on_silence

 import requests
 from open_webui.config import (
@@ -19,21 +21,28 @@ 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
 from pydantic import BaseModel
-from open_webui.utils.utils import get_admin_user, get_current_user, get_verified_user
+from open_webui.utils.utils import get_admin_user, get_verified_user
+
+# Constants
+MAX_FILE_SIZE_MB = 25
+MAX_FILE_SIZE = MAX_FILE_SIZE_MB * 1024 * 1024  # Convert MB to bytes
+

 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["AUDIO"])
@@ -54,6 +63,9 @@ app.state.config.STT_OPENAI_API_KEY = AUDIO_STT_OPENAI_API_KEY
 app.state.config.STT_ENGINE = AUDIO_STT_ENGINE
 app.state.config.STT_MODEL = AUDIO_STT_MODEL

+app.state.config.WHISPER_MODEL = WHISPER_MODEL
+app.state.faster_whisper_model = None
+
 app.state.config.TTS_OPENAI_API_BASE_URL = AUDIO_TTS_OPENAI_API_BASE_URL
 app.state.config.TTS_OPENAI_API_KEY = AUDIO_TTS_OPENAI_API_KEY
 app.state.config.TTS_ENGINE = AUDIO_TTS_ENGINE
@@ -62,6 +74,9 @@ 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}")
@@ -70,6 +85,31 @@ SPEECH_CACHE_DIR = Path(CACHE_DIR).joinpath("./audio/speech/")
 SPEECH_CACHE_DIR.mkdir(parents=True, exist_ok=True)


+def set_faster_whisper_model(model: str, auto_update: bool = False):
+    if model and app.state.config.STT_ENGINE == "":
+        from faster_whisper import WhisperModel
+
+        faster_whisper_kwargs = {
+            "model_size_or_path": model,
+            "device": whisper_device_type,
+            "compute_type": "int8",
+            "download_root": WHISPER_MODEL_DIR,
+            "local_files_only": not auto_update,
+        }
+
+        try:
+            app.state.faster_whisper_model = WhisperModel(**faster_whisper_kwargs)
+        except Exception:
+            log.warning(
+                "WhisperModel initialization failed, attempting download with local_files_only=False"
+            )
+            faster_whisper_kwargs["local_files_only"] = False
+            app.state.faster_whisper_model = WhisperModel(**faster_whisper_kwargs)
+
+    else:
+        app.state.faster_whisper_model = None
+
+
 class TTSConfigForm(BaseModel):
    OPENAI_API_BASE_URL: str
    OPENAI_API_KEY: str
@@ -78,6 +118,8 @@ class TTSConfigForm(BaseModel):
    MODEL: str
    VOICE: str
    SPLIT_ON: str
+    AZURE_SPEECH_REGION: str
+    AZURE_SPEECH_OUTPUT_FORMAT: str


 class STTConfigForm(BaseModel):
@@ -85,6 +127,7 @@ class STTConfigForm(BaseModel):
    OPENAI_API_KEY: str
    ENGINE: str
    MODEL: str
+    WHISPER_MODEL: str


 class AudioConfigUpdateForm(BaseModel):
@@ -130,12 +173,15 @@ 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,
            "OPENAI_API_KEY": app.state.config.STT_OPENAI_API_KEY,
            "ENGINE": app.state.config.STT_ENGINE,
            "MODEL": app.state.config.STT_MODEL,
+            "WHISPER_MODEL": app.state.config.WHISPER_MODEL,
        },
    }

@@ -151,11 +197,17 @@ 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
    app.state.config.STT_ENGINE = form_data.stt.ENGINE
    app.state.config.STT_MODEL = form_data.stt.MODEL
+    app.state.config.WHISPER_MODEL = form_data.stt.WHISPER_MODEL
+    set_faster_whisper_model(form_data.stt.WHISPER_MODEL, WHISPER_MODEL_AUTO_UPDATE)

    return {
        "tts": {
@@ -166,12 +218,15 @@ 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,
            "OPENAI_API_KEY": app.state.config.STT_OPENAI_API_KEY,
            "ENGINE": app.state.config.STT_ENGINE,
            "MODEL": app.state.config.STT_MODEL,
+            "WHISPER_MODEL": app.state.config.WHISPER_MODEL,
        },
    }

@@ -301,68 +356,96 @@ 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")

-@app.post("/transcriptions")
-def transcribe(
-    file: UploadFile = File(...),
-    user=Depends(get_current_user),
-):
-    log.info(f"file.content_type: {file.content_type}")
+        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"

-    if file.content_type not in ["audio/mpeg", "audio/wav"]:
-        raise HTTPException(
-            status_code=status.HTTP_400_BAD_REQUEST,
-            detail=ERROR_MESSAGES.FILE_NOT_SUPPORTED,
-        )
+        headers = {
+            "Ocp-Apim-Subscription-Key": app.state.config.TTS_API_KEY,
+            "Content-Type": "application/ssml+xml",
+            "X-Microsoft-OutputFormat": output_format,
+        }

-    try:
-        ext = file.filename.split(".")[-1]
+        data = f"""<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xml:lang="{locale}">
+                <voice name="{language}">{payload["input"]}</voice>
+            </speak>"""

-        id = uuid.uuid4()
-        filename = f"{id}.{ext}"
+        response = requests.post(url, headers=headers, data=data)

-        file_dir = f"{CACHE_DIR}/audio/transcriptions"
-        os.makedirs(file_dir, exist_ok=True)
-        file_path = f"{file_dir}/{filename}"
-
-        print(filename)
-
-        contents = file.file.read()
-        with open(file_path, "wb") as f:
-            f.write(contents)
-            f.close()
-
-        if app.state.config.STT_ENGINE == "":
-            from faster_whisper import WhisperModel
-
-            whisper_kwargs = {
-                "model_size_or_path": WHISPER_MODEL,
-                "device": whisper_device_type,
-                "compute_type": "int8",
-                "download_root": WHISPER_MODEL_DIR,
-                "local_files_only": not WHISPER_MODEL_AUTO_UPDATE,
-            }
-
-            log.debug(f"whisper_kwargs: {whisper_kwargs}")
-
-            try:
-                model = WhisperModel(**whisper_kwargs)
-            except Exception:
-                log.warning(
-                    "WhisperModel initialization failed, attempting download with local_files_only=False"
-                )
-                whisper_kwargs["local_files_only"] = False
-                model = WhisperModel(**whisper_kwargs)
-
-            segments, info = model.transcribe(file_path, beam_size=5)
-            log.info(
-                "Detected language '%s' with probability %f"
-                % (info.language, info.language_probability)
+        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}"
            )

-            transcript = "".join([segment.text for segment in list(segments)])

-            data = {"text": transcript.strip()}
+def transcribe(file_path):
+    print("transcribe", file_path)
+    filename = os.path.basename(file_path)
+    file_dir = os.path.dirname(file_path)
+    id = filename.split(".")[0]
+
+    if app.state.config.STT_ENGINE == "":
+        if app.state.faster_whisper_model is None:
+            set_faster_whisper_model(app.state.config.WHISPER_MODEL)
+
+        model = app.state.faster_whisper_model
+        segments, info = model.transcribe(file_path, beam_size=5)
+        log.info(
+            "Detected language '%s' with probability %f"
+            % (info.language, info.language_probability)
+        )
+
+        transcript = "".join([segment.text for segment in list(segments)])
+        data = {"text": transcript.strip()}
+
+        # save the transcript to a json file
+        transcript_file = f"{file_dir}/{id}.json"
+        with open(transcript_file, "w") as f:
+            json.dump(data, f)
+
+        log.debug(data)
+        return data
+    elif app.state.config.STT_ENGINE == "openai":
+        if is_mp4_audio(file_path):
+            print("is_mp4_audio")
+            os.rename(file_path, file_path.replace(".wav", ".mp4"))
+            # Convert MP4 audio file to WAV format
+            convert_mp4_to_wav(file_path.replace(".wav", ".mp4"), file_path)
+
+        headers = {"Authorization": f"Bearer {app.state.config.STT_OPENAI_API_KEY}"}
+
+        files = {"file": (filename, open(file_path, "rb"))}
+        data = {"model": app.state.config.STT_MODEL}
+
+        log.debug(files, data)
+
+        r = None
+        try:
+            r = requests.post(
+                url=f"{app.state.config.STT_OPENAI_API_BASE_URL}/audio/transcriptions",
+                headers=headers,
+                files=files,
+                data=data,
+            )
+
+            r.raise_for_status()
+
+            data = r.json()

            # save the transcript to a json file
            transcript_file = f"{file_dir}/{id}.json"
@@ -370,58 +453,83 @@ def transcribe(
                json.dump(data, f)

            print(data)
-
            return data
+        except Exception as e:
+            log.exception(e)
+            error_detail = "Open WebUI: Server Connection Error"
+            if r is not None:
+                try:
+                    res = r.json()
+                    if "error" in res:
+                        error_detail = f"External: {res['error']['message']}"
+                except Exception:
+                    error_detail = f"External: {e}"

-        elif app.state.config.STT_ENGINE == "openai":
-            if is_mp4_audio(file_path):
-                print("is_mp4_audio")
-                os.rename(file_path, file_path.replace(".wav", ".mp4"))
-                # Convert MP4 audio file to WAV format
-                convert_mp4_to_wav(file_path.replace(".wav", ".mp4"), file_path)
+            raise Exception(error_detail)

-            headers = {"Authorization": f"Bearer {app.state.config.STT_OPENAI_API_KEY}"}

-            files = {"file": (filename, open(file_path, "rb"))}
-            data = {"model": app.state.config.STT_MODEL}
+@app.post("/transcriptions")
+def transcription(
+    file: UploadFile = File(...),
+    user=Depends(get_verified_user),
+):
+    log.info(f"file.content_type: {file.content_type}")

-            print(files, data)
+    if file.content_type not in ["audio/mpeg", "audio/wav", "audio/ogg", "audio/x-m4a"]:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=ERROR_MESSAGES.FILE_NOT_SUPPORTED,
+        )

-            r = None
-            try:
-                r = requests.post(
-                    url=f"{app.state.config.STT_OPENAI_API_BASE_URL}/audio/transcriptions",
-                    headers=headers,
-                    files=files,
-                    data=data,
-                )
+    try:
+        ext = file.filename.split(".")[-1]
+        id = uuid.uuid4()

-                r.raise_for_status()
+        filename = f"{id}.{ext}"
+        contents = file.file.read()

-                data = r.json()
+        file_dir = f"{CACHE_DIR}/audio/transcriptions"
+        os.makedirs(file_dir, exist_ok=True)
+        file_path = f"{file_dir}/{filename}"

-                # save the transcript to a json file
-                transcript_file = f"{file_dir}/{id}.json"
-                with open(transcript_file, "w") as f:
-                    json.dump(data, f)
+        with open(file_path, "wb") as f:
+            f.write(contents)

-                print(data)
-                return data
-            except Exception as e:
-                log.exception(e)
-                error_detail = "Open WebUI: Server Connection Error"
-                if r is not None:
-                    try:
-                        res = r.json()
-                        if "error" in res:
-                            error_detail = f"External: {res['error']['message']}"
-                    except Exception:
-                        error_detail = f"External: {e}"
+        try:
+            if os.path.getsize(file_path) > MAX_FILE_SIZE:  # file is bigger than 25MB
+                log.debug(f"File size is larger than {MAX_FILE_SIZE_MB}MB")
+                audio = AudioSegment.from_file(file_path)
+                audio = audio.set_frame_rate(16000).set_channels(1)  # Compress audio
+                compressed_path = f"{file_dir}/{id}_compressed.opus"
+                audio.export(compressed_path, format="opus", bitrate="32k")
+                log.debug(f"Compressed audio to {compressed_path}")
+                file_path = compressed_path

-                raise HTTPException(
-                    status_code=r.status_code if r != None else 500,
-                    detail=error_detail,
-                )
+                if (
+                    os.path.getsize(file_path) > MAX_FILE_SIZE
+                ):  # Still larger than 25MB after compression
+                    log.debug(
+                        f"Compressed file size is still larger than {MAX_FILE_SIZE_MB}MB: {os.path.getsize(file_path)}"
+                    )
+                    raise HTTPException(
+                        status_code=status.HTTP_400_BAD_REQUEST,
+                        detail=ERROR_MESSAGES.FILE_TOO_LARGE(
+                            size=f"{MAX_FILE_SIZE_MB}MB"
+                        ),
+                    )
+
+                data = transcribe(file_path)
+            else:
+                data = transcribe(file_path)
+
+            file_path = file_path.split("/")[-1]
+            return {**data, "filename": file_path}
+        except Exception as e:
+            log.exception(e)
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.DEFAULT(e),
+            )

    except Exception as e:
        log.exception(e)
@@ -443,7 +551,7 @@ 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 +586,21 @@ 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

@@ -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,
@@ -125,22 +125,34 @@ async def comfyui_generate_image(
                    workflow[node_id]["inputs"][node.key] = model
            elif node.type == "prompt":
                for node_id in node.node_ids:
-                    workflow[node_id]["inputs"]["text"] = payload.prompt
+                    workflow[node_id]["inputs"][
+                        node.key if node.key else "text"
+                    ] = payload.prompt
            elif node.type == "negative_prompt":
                for node_id in node.node_ids:
-                    workflow[node_id]["inputs"]["text"] = payload.negative_prompt
+                    workflow[node_id]["inputs"][
+                        node.key if node.key else "text"
+                    ] = payload.negative_prompt
            elif node.type == "width":
                for node_id in node.node_ids:
-                    workflow[node_id]["inputs"]["width"] = payload.width
+                    workflow[node_id]["inputs"][
+                        node.key if node.key else "width"
+                    ] = payload.width
            elif node.type == "height":
                for node_id in node.node_ids:
-                    workflow[node_id]["inputs"]["height"] = payload.height
+                    workflow[node_id]["inputs"][
+                        node.key if node.key else "height"
+                    ] = payload.height
            elif node.type == "n":
                for node_id in node.node_ids:
-                    workflow[node_id]["inputs"]["batch_size"] = payload.n
+                    workflow[node_id]["inputs"][
+                        node.key if node.key else "batch_size"
+                    ] = payload.n
            elif node.type == "steps":
                for node_id in node.node_ids:
-                    workflow[node_id]["inputs"]["steps"] = payload.steps
+                    workflow[node_id]["inputs"][
+                        node.key if node.key else "steps"
+                    ] = payload.steps
            elif node.type == "seed":
                seed = (
                    payload.seed
@@ -12,7 +12,6 @@ import aiohttp
 import requests
 from open_webui.apps.webui.models.models import Models
 from open_webui.config import (
-    AIOHTTP_CLIENT_TIMEOUT,
    CORS_ALLOW_ORIGIN,
    ENABLE_MODEL_FILTER,
    ENABLE_OLLAMA_API,
@@ -21,6 +20,9 @@ from open_webui.config import (
    UPLOAD_DIR,
    AppConfig,
 )
+from open_webui.env import AIOHTTP_CLIENT_TIMEOUT
+
+
 from open_webui.constants import ERROR_MESSAGES
 from open_webui.env import SRC_LOG_LEVELS
 from fastapi import Depends, FastAPI, File, HTTPException, Request, UploadFile
@@ -28,11 +30,15 @@ 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

@@ -113,7 +119,7 @@ async def update_ollama_api_url(form_data: UrlUpdateForm, user=Depends(get_admin


 async def fetch_url(url):
-    timeout = aiohttp.ClientTimeout(total=5)
+    timeout = aiohttp.ClientTimeout(total=3)
    try:
        async with aiohttp.ClientSession(timeout=timeout, trust_env=True) as session:
            async with session.get(url) as response:
@@ -539,6 +545,24 @@ class GenerateEmbeddingsForm(BaseModel):
    keep_alive: Optional[Union[int, str]] = None


+class GenerateEmbedForm(BaseModel):
+    model: str
+    input: list[str] | str
+    truncate: Optional[bool] = None
+    options: Optional[dict] = None
+    keep_alive: Optional[Union[int, str]] = None
+
+
+@app.post("/api/embed")
+@app.post("/api/embed/{url_idx}")
+async def generate_embeddings(
+    form_data: GenerateEmbedForm,
+    url_idx: Optional[int] = None,
+    user=Depends(get_verified_user),
+):
+    return generate_ollama_batch_embeddings(form_data, url_idx)
+
+
@app.post("/api/embeddings")
@app.post("/api/embeddings/{url_idx}")
 async def generate_embeddings(
@@ -546,48 +570,7 @@ async def generate_embeddings(
    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/embeddings",
-        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,
-        )
+    return generate_ollama_embeddings(form_data=form_data, url_idx=url_idx)


 def generate_ollama_embeddings(
@@ -627,7 +610,64 @@ def generate_ollama_embeddings(
        log.info(f"generate_ollama_embeddings {data}")

        if "embedding" in data:
-            return data["embedding"]
+            return data
+        else:
+            raise Exception("Something went wrong :/")
+    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,
+        )
+
+
+def generate_ollama_batch_embeddings(
+    form_data: GenerateEmbedForm,
+    url_idx: Optional[int] = None,
+):
+    log.info(f"generate_ollama_batch_embeddings {form_data}")
+
+    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()
+
+        data = r.json()
+
+        log.info(f"generate_ollama_batch_embeddings {data}")
+
+        if "embeddings" in data:
+            return data
        else:
            raise Exception("Something went wrong :/")
    except Exception as e:
@@ -699,7 +739,7 @@ class GenerateChatCompletionForm(BaseModel):
    format: Optional[str] = None
    options: Optional[dict] = None
    template: Optional[str] = None
-    stream: Optional[bool] = None
+    stream: Optional[bool] = True
    keep_alive: Optional[Union[int, str]] = None


@@ -721,15 +761,16 @@ async def generate_chat_completion(
    form_data: GenerateChatCompletionForm,
    url_idx: Optional[int] = None,
    user=Depends(get_verified_user),
+    bypass_filter: Optional[bool] = False,
 ):
    payload = {**form_data.model_dump(exclude_none=True)}
-    log.debug(f"{payload = }")
+    log.debug(f"generate_chat_completion() - 1.payload = {payload}")
    if "metadata" in payload:
        del payload["metadata"]

    model_id = form_data.model

-    if app.state.config.ENABLE_MODEL_FILTER:
+    if not bypass_filter and app.state.config.ENABLE_MODEL_FILTER:
        if user.role == "user" and model_id not in app.state.config.MODEL_FILTER_LIST:
            raise HTTPException(
                status_code=403,
@@ -758,10 +799,13 @@ async def generate_chat_completion(

    url = get_ollama_url(url_idx, payload["model"])
    log.info(f"url: {url}")
-    log.debug(payload)
+    log.debug(f"generate_chat_completion() - 2.payload = {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",
    )


@@ -9,7 +9,6 @@ import aiohttp
 import requests
 from open_webui.apps.webui.models.models import Models
 from open_webui.config import (
-    AIOHTTP_CLIENT_TIMEOUT,
    CACHE_DIR,
    CORS_ALLOW_ORIGIN,
    ENABLE_MODEL_FILTER,
@@ -19,6 +18,11 @@ from open_webui.config import (
    OPENAI_API_KEYS,
    AppConfig,
 )
+from open_webui.env import (
+    AIOHTTP_CLIENT_TIMEOUT,
+    AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST,
+)
+
 from open_webui.constants import ERROR_MESSAGES
 from open_webui.env import SRC_LOG_LEVELS
 from fastapi import Depends, FastAPI, HTTPException, Request
@@ -26,10 +30,12 @@ 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__)
@@ -44,7 +50,6 @@ app.add_middleware(
    allow_headers=["*"],
 )

-
 app.state.config = AppConfig()

 app.state.config.ENABLE_MODEL_FILTER = ENABLE_MODEL_FILTER
@@ -177,7 +182,7 @@ async def speech(request: Request, user=Depends(get_verified_user)):


 async def fetch_url(url, key):
-    timeout = aiohttp.ClientTimeout(total=5)
+    timeout = aiohttp.ClientTimeout(total=AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST)
    try:
        headers = {"Authorization": f"Bearer {key}"}
        async with aiohttp.ClientSession(timeout=timeout, trust_env=True) as session:
@@ -217,7 +222,17 @@ 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",
+                        ]
+                    )
                ]
            )

@@ -225,9 +240,7 @@ def merge_models_lists(model_lists):


 def is_openai_api_disabled():
-    api_keys = app.state.config.OPENAI_API_KEYS
-    no_keys = len(api_keys) == 1 and api_keys[0] == ""
-    return no_keys or not app.state.config.ENABLE_OPENAI_API
+    return not app.state.config.ENABLE_OPENAI_API


 async def get_all_models_raw() -> list:
@@ -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 +405,32 @@ 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]
+    is_o1 = payload["model"].lower().startswith("o1-")
+
+    # Change max_completion_tokens to max_tokens (Backward compatible)
+    if "api.openai.com" not in url and not is_o1:
+        if "max_completion_tokens" in payload:
+            # Remove "max_completion_tokens" from the payload
+            payload["max_tokens"] = payload["max_completion_tokens"]
+            del payload["max_completion_tokens"]
+    else:
+        if is_o1 and "max_tokens" in payload:
+            payload["max_completion_tokens"] = payload["max_tokens"]
+            del payload["max_tokens"]
+        if "max_tokens" in payload and "max_completion_tokens" in payload:
+            del payload["max_tokens"]
+
+    # Fix: O1 does not support the "system" parameter, Modify "system" to "user"
+    if is_o1 and payload["messages"][0]["role"] == "system":
+        payload["messages"][0]["role"] = "user"
+
    # Convert the modified body back to JSON
    payload = json.dumps(payload)

    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 +441,7 @@ async def generate_chat_completion(
    r = None
    session = None
    streaming = False
+    response = None

    try:
        session = aiohttp.ClientSession(
@@ -408,8 +454,6 @@ 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 +466,23 @@ 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:
@@ -0,0 +1,190 @@
+import requests
+import logging
+import ftfy
+
+from langchain_community.document_loaders import (
+    BSHTMLLoader,
+    CSVLoader,
+    Docx2txtLoader,
+    OutlookMessageLoader,
+    PyPDFLoader,
+    TextLoader,
+    UnstructuredEPubLoader,
+    UnstructuredExcelLoader,
+    UnstructuredMarkdownLoader,
+    UnstructuredPowerPointLoader,
+    UnstructuredRSTLoader,
+    UnstructuredXMLLoader,
+    YoutubeLoader,
+)
+from langchain_core.documents import Document
+from open_webui.env import SRC_LOG_LEVELS
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
+known_source_ext = [
+    "go",
+    "py",
+    "java",
+    "sh",
+    "bat",
+    "ps1",
+    "cmd",
+    "js",
+    "ts",
+    "css",
+    "cpp",
+    "hpp",
+    "h",
+    "c",
+    "cs",
+    "sql",
+    "log",
+    "ini",
+    "pl",
+    "pm",
+    "r",
+    "dart",
+    "dockerfile",
+    "env",
+    "php",
+    "hs",
+    "hsc",
+    "lua",
+    "nginxconf",
+    "conf",
+    "m",
+    "mm",
+    "plsql",
+    "perl",
+    "rb",
+    "rs",
+    "db2",
+    "scala",
+    "bash",
+    "swift",
+    "vue",
+    "svelte",
+    "msg",
+    "ex",
+    "exs",
+    "erl",
+    "tsx",
+    "jsx",
+    "hs",
+    "lhs",
+]
+
+
+class TikaLoader:
+    def __init__(self, url, file_path, mime_type=None):
+        self.url = url
+        self.file_path = file_path
+        self.mime_type = mime_type
+
+    def load(self) -> list[Document]:
+        with open(self.file_path, "rb") as f:
+            data = f.read()
+
+        if self.mime_type is not None:
+            headers = {"Content-Type": self.mime_type}
+        else:
+            headers = {}
+
+        endpoint = self.url
+        if not endpoint.endswith("/"):
+            endpoint += "/"
+        endpoint += "tika/text"
+
+        r = requests.put(endpoint, data=data, headers=headers)
+
+        if r.ok:
+            raw_metadata = r.json()
+            text = raw_metadata.get("X-TIKA:content", "<No text content found>")
+
+            if "Content-Type" in raw_metadata:
+                headers["Content-Type"] = raw_metadata["Content-Type"]
+
+            log.info("Tika extracted text: %s", text)
+
+            return [Document(page_content=text, metadata=headers)]
+        else:
+            raise Exception(f"Error calling Tika: {r.reason}")
+
+
+class Loader:
+    def __init__(self, engine: str = "", **kwargs):
+        self.engine = engine
+        self.kwargs = kwargs
+
+    def load(
+        self, filename: str, file_content_type: str, file_path: str
+    ) -> list[Document]:
+        loader = self._get_loader(filename, file_content_type, file_path)
+        docs = loader.load()
+
+        return [
+            Document(
+                page_content=ftfy.fix_text(doc.page_content), metadata=doc.metadata
+            )
+            for doc in docs
+        ]
+
+    def _get_loader(self, filename: str, file_content_type: str, file_path: str):
+        file_ext = filename.split(".")[-1].lower()
+
+        if self.engine == "tika" and self.kwargs.get("TIKA_SERVER_URL"):
+            if file_ext in known_source_ext or (
+                file_content_type and file_content_type.find("text/") >= 0
+            ):
+                loader = TextLoader(file_path, autodetect_encoding=True)
+            else:
+                loader = TikaLoader(
+                    url=self.kwargs.get("TIKA_SERVER_URL"),
+                    file_path=file_path,
+                    mime_type=file_content_type,
+                )
+        else:
+            if file_ext == "pdf":
+                loader = PyPDFLoader(
+                    file_path, extract_images=self.kwargs.get("PDF_EXTRACT_IMAGES")
+                )
+            elif file_ext == "csv":
+                loader = CSVLoader(file_path)
+            elif file_ext == "rst":
+                loader = UnstructuredRSTLoader(file_path, mode="elements")
+            elif file_ext == "xml":
+                loader = UnstructuredXMLLoader(file_path)
+            elif file_ext in ["htm", "html"]:
+                loader = BSHTMLLoader(file_path, open_encoding="unicode_escape")
+            elif file_ext == "md":
+                loader = UnstructuredMarkdownLoader(file_path)
+            elif file_content_type == "application/epub+zip":
+                loader = UnstructuredEPubLoader(file_path)
+            elif (
+                file_content_type
+                == "application/vnd.openxmlformats-officedocument.wordprocessingml.document"
+                or file_ext == "docx"
+            ):
+                loader = Docx2txtLoader(file_path)
+            elif file_content_type in [
+                "application/vnd.ms-excel",
+                "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
+            ] or file_ext in ["xls", "xlsx"]:
+                loader = UnstructuredExcelLoader(file_path)
+            elif file_content_type in [
+                "application/vnd.ms-powerpoint",
+                "application/vnd.openxmlformats-officedocument.presentationml.presentation",
+            ] or file_ext in ["ppt", "pptx"]:
+                loader = UnstructuredPowerPointLoader(file_path)
+            elif file_ext == "msg":
+                loader = OutlookMessageLoader(file_path)
+            elif file_ext in known_source_ext or (
+                file_content_type and file_content_type.find("text/") >= 0
+            ):
+                loader = TextLoader(file_path, autodetect_encoding=True)
+            else:
+                loader = TextLoader(file_path, autodetect_encoding=True)
+
+        return loader
@@ -0,0 +1,81 @@
+import os
+import torch
+import numpy as np
+from colbert.infra import ColBERTConfig
+from colbert.modeling.checkpoint import Checkpoint
+
+
+class ColBERT:
+    def __init__(self, name, **kwargs) -> None:
+        print("ColBERT: Loading model", name)
+        self.device = "cuda" if torch.cuda.is_available() else "cpu"
+
+        DOCKER = kwargs.get("env") == "docker"
+        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
@@ -1,42 +1,85 @@
 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 (
+    GenerateEmbedForm,
+    generate_ollama_batch_embeddings,
+)
+from open_webui.apps.retrieval.vector.connector import VECTOR_DB_CLIENT
 from open_webui.utils.misc import get_last_user_message

+from open_webui.env import SRC_LOG_LEVELS
+from open_webui.config import DEFAULT_RAG_TEMPLATE
+
+
 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,
-    embedding_function,
+    query_embedding: list[float],
    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=[query_embedding],
+            limit=k,
        )

        log.info(f"query_doc:result {result}")
        return result
    except Exception as e:
+        print(e)
        raise e


@@ -47,27 +90,25 @@ 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 +133,9 @@ 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 = []
@@ -138,20 +181,23 @@ def query_collection(
    query: str,
    embedding_function,
    k: int,
-):
+) -> dict:
+
    results = []
+    query_embedding = embedding_function(query)
+
    for collection_name in collection_names:
        if collection_name:
            try:
                result = query_doc(
                    collection_name=collection_name,
-                    query=query,
                    k=k,
-                    embedding_function=embedding_function,
+                    query_embedding=query_embedding,
                )
-                results.append(result)
-            except Exception:
-                pass
+                if result is not None:
+                    results.append(result.model_dump())
+            except Exception as e:
+                log.exception(f"Error when querying the collection: {e}")
        else:
            pass

@@ -165,8 +211,9 @@ 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 +225,55 @@ 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):
+    if template == "":
+        template = DEFAULT_RAG_TEMPLATE
+
+    if "[context]" not in template and "{{CONTEXT}}" not in template:
+        log.debug(
+            "WARNING: The RAG template does not contain the '[context]' or '{{CONTEXT}}' placeholder."
+        )
+
+    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."
+        )
+
+    query_placeholders = []
+    if "[query]" in context:
+        query_placeholder = "{{QUERY" + str(uuid.uuid4()) + "}}"
+        template = template.replace("[query]", query_placeholder)
+        query_placeholders.append(query_placeholder)
+
+    if "{{QUERY}}" in context:
+        query_placeholder = "{{QUERY" + str(uuid.uuid4()) + "}}"
+        template = template.replace("{{QUERY}}", query_placeholder)
+        query_placeholders.append(query_placeholder)
+
    template = template.replace("[context]", context)
+    template = template.replace("{{CONTEXT}}", context)
    template = template.replace("[query]", query)
+    template = template.replace("{{QUERY}}", query)
+
+    for query_placeholder in query_placeholders:
+        template = template.replace(query_placeholder, query)
+
    return template


@@ -195,39 +283,27 @@ def get_embedding_function(
    embedding_function,
    openai_key,
    openai_url,
-    batch_size,
+    embedding_batch_size,
 ):
    if embedding_engine == "":
        return lambda query: embedding_function.encode(query).tolist()
    elif embedding_engine in ["ollama", "openai"]:
-        if embedding_engine == "ollama":
-            func = lambda query: generate_ollama_embeddings(
-                GenerateEmbeddingsForm(
-                    **{
-                        "model": embedding_model,
-                        "prompt": query,
-                    }
-                )
-            )
-        elif embedding_engine == "openai":
-            func = lambda query: generate_openai_embeddings(
-                model=embedding_model,
-                text=query,
-                key=openai_key,
-                url=openai_url,
-            )
+        func = lambda query: generate_embeddings(
+            engine=embedding_engine,
+            model=embedding_model,
+            text=query,
+            key=openai_key if embedding_engine == "openai" else "",
+            url=openai_url if embedding_engine == "openai" else "",
+        )

-        def generate_multiple(query, f):
+        def generate_multiple(query, func):
            if isinstance(query, list):
-                if embedding_engine == "openai":
-                    embeddings = []
-                    for i in range(0, len(query), batch_size):
-                        embeddings.extend(f(query[i : i + batch_size]))
-                    return embeddings
-                else:
-                    return [f(q) for q in query]
+                embeddings = []
+                for i in range(0, len(query), embedding_batch_size):
+                    embeddings.extend(func(query[i : i + embedding_batch_size]))
+                return embeddings
            else:
-                return f(query)
+                return func(query)

        return lambda query: generate_multiple(query, func)

@@ -248,71 +324,107 @@ def get_rag_context(
    relevant_contexts = []

    for file in files:
-        context = None
-
-        collection_names = (
-            file["collection_names"]
-            if file["type"] == "collection"
-            else [file["collection_name"]] if file["collection_name"] else []
-        )
-
-        collection_names = set(collection_names).difference(extracted_collections)
-        if not collection_names:
-            log.debug(f"skipping {file} as it has already been extracted")
-            continue
-
-        try:
-            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:
-                    context = query_collection(
-                        collection_names=collection_names,
-                        query=query,
-                        embedding_function=embedding_function,
-                        k=k,
-                    )
-        except Exception as e:
-            log.exception(e)
+        if file.get("context") == "full":
+            context = {
+                "documents": [[file.get("file").get("data", {}).get("content")]],
+                "metadatas": [[{"file_id": file.get("id"), "name": file.get("name")}]],
+            }
+        else:
            context = None

-        if context:
-            relevant_contexts.append({**context, "source": file})
+            collection_names = []
+            if file.get("type") == "collection":
+                if file.get("legacy"):
+                    collection_names = file.get("collection_names", [])
+                else:
+                    collection_names.append(file["id"])
+            elif file.get("collection_name"):
+                collection_names.append(file["collection_name"])
+            elif file.get("id"):
+                if file.get("legacy"):
+                    collection_names.append(f"{file['id']}")
+                else:
+                    collection_names.append(f"file-{file['id']}")

-        extracted_collections.extend(collection_names)
+            collection_names = set(collection_names).difference(extracted_collections)
+            if not collection_names:
+                log.debug(f"skipping {file} as it has already been extracted")
+                continue
+
+            try:
+                context = None
+                if file.get("type") == "text":
+                    context = file["content"]
+                else:
+                    if hybrid_search:
+                        try:
+                            context = query_collection_with_hybrid_search(
+                                collection_names=collection_names,
+                                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,
+                            embedding_function=embedding_function,
+                            k=k,
+                        )
+            except Exception as e:
+                log.exception(e)
+
+            extracted_collections.extend(collection_names)
+
+        if context:
+            if "data" in file:
+                del file["data"]
+            relevant_contexts.append({**context, "file": file})

    contexts = []
    citations = []
-
    for context in relevant_contexts:
        try:
            if "documents" in context:
+                file_names = list(
+                    set(
+                        [
+                            metadata["name"]
+                            for metadata in context["metadatas"][0]
+                            if metadata is not None and "name" in metadata
+                        ]
+                    )
+                )
                contexts.append(
-                    "\n\n".join(
+                    ((", ".join(file_names) + ":\n\n") if file_names else "")
+                    + "\n\n".join(
                        [text for text in context["documents"][0] if text is not None]
                    )
                )

                if "metadatas" in context:
-                    citations.append(
-                        {
-                            "source": context["source"],
-                            "document": context["documents"][0],
-                            "metadata": context["metadatas"][0],
-                        }
-                    )
+                    citation = {
+                        "source": context["file"],
+                        "document": context["documents"][0],
+                        "metadata": context["metadatas"][0],
+                    }
+                    if "distances" in context and context["distances"]:
+                        citation["distances"] = context["distances"][0]
+                    citations.append(citation)
        except Exception as e:
            log.exception(e)

+    print("contexts", contexts)
+    print("citations", citations)
+
    return contexts, citations


@@ -354,20 +466,6 @@ def get_model_path(model: str, update_model: bool = False):
        return model


-def generate_openai_embeddings(
-    model: str,
-    text: Union[str, list[str]],
-    key: str,
-    url: str = "https://api.openai.com/v1",
-):
-    if isinstance(text, list):
-        embeddings = generate_openai_batch_embeddings(model, text, key, url)
-    else:
-        embeddings = generate_openai_batch_embeddings(model, [text], key, url)
-
-    return embeddings[0] if isinstance(text, str) else embeddings
-
-
 def generate_openai_batch_embeddings(
    model: str, texts: list[str], key: str, url: str = "https://api.openai.com/v1"
 ) -> Optional[list[list[float]]]:
@@ -391,43 +489,31 @@ 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],
-                )
+def generate_embeddings(engine: str, model: str, text: Union[str, list[str]], **kwargs):
+    if engine == "ollama":
+        if isinstance(text, list):
+            embeddings = generate_ollama_batch_embeddings(
+                GenerateEmbedForm(**{"model": model, "input": text})
            )
-        return results
+        else:
+            embeddings = generate_ollama_batch_embeddings(
+                GenerateEmbedForm(**{"model": model, "input": [text]})
+            )
+        return (
+            embeddings["embeddings"][0]
+            if isinstance(text, str)
+            else embeddings["embeddings"]
+        )
+    elif engine == "openai":
+        key = kwargs.get("key", "")
+        url = kwargs.get("url", "https://api.openai.com/v1")
+
+        if isinstance(text, list):
+            embeddings = generate_openai_batch_embeddings(model, text, key, url)
+        else:
+            embeddings = generate_openai_batch_embeddings(model, [text], key, url)
+
+        return embeddings[0] if isinstance(text, str) else embeddings


 import operator
@@ -435,7 +521,6 @@ 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 +530,7 @@ class RerankCompressor(BaseDocumentCompressor):
    r_score: float

    class Config:
-        extra = Extra.forbid
+        extra = "forbid"
        arbitrary_types_allowed = True

    def compress_documents(
@@ -0,0 +1,14 @@
+from open_webui.config import VECTOR_DB
+
+if VECTOR_DB == "milvus":
+    from open_webui.apps.retrieval.vector.dbs.milvus import MilvusClient
+
+    VECTOR_DB_CLIENT = MilvusClient()
+elif VECTOR_DB == "qdrant":
+    from open_webui.apps.retrieval.vector.dbs.qdrant import QdrantClient
+
+    VECTOR_DB_CLIENT = QdrantClient()
+else:
+    from open_webui.apps.retrieval.vector.dbs.chroma import ChromaClient
+
+    VECTOR_DB_CLIENT = ChromaClient()
@@ -0,0 +1,161 @@
+import chromadb
+from chromadb import Settings
+from chromadb.utils.batch_utils import create_batches
+
+from typing import Optional
+
+from open_webui.apps.retrieval.vector.main import VectorItem, SearchResult, GetResult
+from open_webui.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.
+        try:
+            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
+        except Exception as e:
+            return None
+
+    def query(
+        self, collection_name: str, filter: dict, limit: Optional[int] = None
+    ) -> Optional[GetResult]:
+        # Query the items from the collection based on the filter.
+        try:
+            collection = self.client.get_collection(name=collection_name)
+            if collection:
+                result = collection.get(
+                    where=filter,
+                    limit=limit,
+                )
+
+                return GetResult(
+                    **{
+                        "ids": [result["ids"]],
+                        "documents": [result["documents"]],
+                        "metadatas": [result["metadatas"]],
+                    }
+                )
+            return None
+        except Exception as e:
+            print(e)
+            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, metadata={"hnsw:space": "cosine"}
+        )
+
+        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, metadata={"hnsw:space": "cosine"}
+        )
+
+        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: Optional[list[str]] = None,
+        filter: Optional[dict] = None,
+    ):
+        # Delete the items from the collection based on the ids.
+        collection = self.client.get_collection(name=collection_name)
+        if collection:
+            if ids:
+                collection.delete(ids=ids)
+            elif filter:
+                collection.delete(where=filter)
+
+    def reset(self):
+        # Resets the database. This will delete all collections and item entries.
+        return self.client.reset()
@@ -0,0 +1,286 @@
+from pymilvus import MilvusClient as Client
+from pymilvus import FieldSchema, DataType
+import json
+
+from typing import Optional
+
+from open_webui.apps.retrieval.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:
+        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:
+        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={"M": 16, "efConstruction": 100},
+        )
+
+        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.
+        collection_name = collection_name.replace("-", "_")
+        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.
+        collection_name = collection_name.replace("-", "_")
+        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.
+        collection_name = collection_name.replace("-", "_")
+        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 query(self, collection_name: str, filter: dict, limit: Optional[int] = None):
+        # Construct the filter string for querying
+        collection_name = collection_name.replace("-", "_")
+        if not self.has_collection(collection_name):
+            return None
+
+        filter_string = " && ".join(
+            [
+                f'metadata["{key}"] == {json.dumps(value)}'
+                for key, value in filter.items()
+            ]
+        )
+
+        max_limit = 16383  # The maximum number of records per request
+        all_results = []
+
+        if limit is None:
+            limit = float("inf")  # Use infinity as a placeholder for no limit
+
+        # Initialize offset and remaining to handle pagination
+        offset = 0
+        remaining = limit
+
+        try:
+            # Loop until there are no more items to fetch or the desired limit is reached
+            while remaining > 0:
+                print("remaining", remaining)
+                current_fetch = min(
+                    max_limit, remaining
+                )  # Determine how many items to fetch in this iteration
+
+                results = self.client.query(
+                    collection_name=f"{self.collection_prefix}_{collection_name}",
+                    filter=filter_string,
+                    output_fields=["*"],
+                    limit=current_fetch,
+                    offset=offset,
+                )
+
+                if not results:
+                    break
+
+                all_results.extend(results)
+                results_count = len(results)
+                remaining -= (
+                    results_count  # Decrease remaining by the number of items fetched
+                )
+                offset += results_count
+
+                # Break the loop if the results returned are less than the requested fetch count
+                if results_count < current_fetch:
+                    break
+
+            print(all_results)
+            return self._result_to_get_result([all_results])
+        except Exception as e:
+            print(e)
+            return None
+
+    def get(self, collection_name: str) -> Optional[GetResult]:
+        # Get all the items in the collection.
+        collection_name = collection_name.replace("-", "_")
+        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.
+        collection_name = collection_name.replace("-", "_")
+        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.
+        collection_name = collection_name.replace("-", "_")
+        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: Optional[list[str]] = None,
+        filter: Optional[dict] = None,
+    ):
+        # Delete the items from the collection based on the ids.
+        collection_name = collection_name.replace("-", "_")
+        if ids:
+            return self.client.delete(
+                collection_name=f"{self.collection_prefix}_{collection_name}",
+                ids=ids,
+            )
+        elif filter:
+            # Convert the filter dictionary to a string using JSON_CONTAINS.
+            filter_string = " && ".join(
+                [
+                    f'metadata["{key}"] == {json.dumps(value)}'
+                    for key, value in filter.items()
+                ]
+            )
+
+            return self.client.delete(
+                collection_name=f"{self.collection_prefix}_{collection_name}",
+                filter=filter_string,
+            )
+
+    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)
@@ -0,0 +1,179 @@
+from typing import Optional
+
+from qdrant_client import QdrantClient as Qclient
+from qdrant_client.http.models import PointStruct
+from qdrant_client.models import models
+
+from open_webui.apps.retrieval.vector.main import VectorItem, SearchResult, GetResult
+from open_webui.config import QDRANT_URI
+
+NO_LIMIT = 999999999
+
+
+class QdrantClient:
+    def __init__(self):
+        self.collection_prefix = "open-webui"
+        self.QDRANT_URI = QDRANT_URI
+        self.client = Qclient(url=self.QDRANT_URI) if self.QDRANT_URI else None
+
+    def _result_to_get_result(self, points) -> GetResult:
+        ids = []
+        documents = []
+        metadatas = []
+
+        for point in points:
+            payload = point.payload
+            ids.append(point.id)
+            documents.append(payload["text"])
+            metadatas.append(payload["metadata"])
+
+        return GetResult(
+            **{
+                "ids": [ids],
+                "documents": [documents],
+                "metadatas": [metadatas],
+            }
+        )
+
+    def _create_collection(self, collection_name: str, dimension: int):
+        collection_name_with_prefix = f"{self.collection_prefix}_{collection_name}"
+        self.client.create_collection(
+            collection_name=collection_name_with_prefix,
+            vectors_config=models.VectorParams(
+                size=dimension, distance=models.Distance.COSINE
+            ),
+        )
+
+        print(f"collection {collection_name_with_prefix} successfully created!")
+
+    def _create_collection_if_not_exists(self, collection_name, dimension):
+        if not self.has_collection(collection_name=collection_name):
+            self._create_collection(
+                collection_name=collection_name, dimension=dimension
+            )
+
+    def _create_points(self, items: list[VectorItem]):
+        return [
+            PointStruct(
+                id=item["id"],
+                vector=item["vector"],
+                payload={"text": item["text"], "metadata": item["metadata"]},
+            )
+            for item in items
+        ]
+
+    def has_collection(self, collection_name: str) -> bool:
+        return self.client.collection_exists(
+            f"{self.collection_prefix}_{collection_name}"
+        )
+
+    def delete_collection(self, collection_name: str):
+        return self.client.delete_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.
+        if limit is None:
+            limit = NO_LIMIT  # otherwise qdrant would set limit to 10!
+
+        query_response = self.client.query_points(
+            collection_name=f"{self.collection_prefix}_{collection_name}",
+            query=vectors[0],
+            limit=limit,
+        )
+        get_result = self._result_to_get_result(query_response.points)
+        return SearchResult(
+            ids=get_result.ids,
+            documents=get_result.documents,
+            metadatas=get_result.metadatas,
+            distances=[[point.score for point in query_response.points]],
+        )
+
+    def query(self, collection_name: str, filter: dict, limit: Optional[int] = None):
+        # Construct the filter string for querying
+        if not self.has_collection(collection_name):
+            return None
+        try:
+            if limit is None:
+                limit = NO_LIMIT  # otherwise qdrant would set limit to 10!
+
+            field_conditions = []
+            for key, value in filter.items():
+                field_conditions.append(
+                    models.FieldCondition(
+                        key=f"metadata.{key}", match=models.MatchValue(value=value)
+                    )
+                )
+
+            points = self.client.query_points(
+                collection_name=f"{self.collection_prefix}_{collection_name}",
+                query_filter=models.Filter(should=field_conditions),
+                limit=limit,
+            )
+            return self._result_to_get_result(points.points)
+        except Exception as e:
+            print(e)
+            return None
+
+    def get(self, collection_name: str) -> Optional[GetResult]:
+        # Get all the items in the collection.
+        points = self.client.query_points(
+            collection_name=f"{self.collection_prefix}_{collection_name}",
+            limit=NO_LIMIT,  # otherwise qdrant would set limit to 10!
+        )
+        return self._result_to_get_result(points.points)
+
+    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.
+        self._create_collection_if_not_exists(collection_name, len(items[0]["vector"]))
+        points = self._create_points(items)
+        self.client.upload_points(f"{self.collection_prefix}_{collection_name}", points)
+
+    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.
+        self._create_collection_if_not_exists(collection_name, len(items[0]["vector"]))
+        points = self._create_points(items)
+        return self.client.upsert(f"{self.collection_prefix}_{collection_name}", points)
+
+    def delete(
+        self,
+        collection_name: str,
+        ids: Optional[list[str]] = None,
+        filter: Optional[dict] = None,
+    ):
+        # Delete the items from the collection based on the ids.
+        field_conditions = []
+
+        if ids:
+            for id_value in ids:
+                field_conditions.append(
+                    models.FieldCondition(
+                        key="metadata.id",
+                        match=models.MatchValue(value=id_value),
+                    ),
+                ),
+        elif filter:
+            for key, value in filter.items():
+                field_conditions.append(
+                    models.FieldCondition(
+                        key=f"metadata.{key}",
+                        match=models.MatchValue(value=value),
+                    ),
+                ),
+
+        return self.client.delete(
+            collection_name=f"{self.collection_prefix}_{collection_name}",
+            points_selector=models.FilterSelector(
+                filter=models.Filter(must=field_conditions)
+            ),
+        )
+
+    def reset(self):
+        # Resets the database. This will delete all collections and item entries.
+        collection_names = self.client.get_collections().collections
+        for collection_name in collection_names:
+            if collection_name.name.startswith(self.collection_prefix):
+                self.client.delete_collection(collection_name=collection_name.name)
@@ -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]]]
@@ -2,7 +2,7 @@ import logging
 from typing import Optional

 import requests
-from open_webui.apps.rag.search.main import SearchResult, get_filtered_results
+from open_webui.apps.retrieval.web.main import SearchResult, get_filtered_results
 from open_webui.env import SRC_LOG_LEVELS

 log = logging.getLogger(__name__)
@@ -1,7 +1,7 @@
 import logging
 from typing import Optional

-from open_webui.apps.rag.search.main import SearchResult, get_filtered_results
+from open_webui.apps.retrieval.web.main import SearchResult, get_filtered_results
 from duckduckgo_search import DDGS
 from open_webui.env import SRC_LOG_LEVELS

@@ -2,7 +2,7 @@ import logging
 from typing import Optional

 import requests
-from open_webui.apps.rag.search.main import SearchResult, get_filtered_results
+from open_webui.apps.retrieval.web.main import SearchResult, get_filtered_results
 from open_webui.env import SRC_LOG_LEVELS

 log = logging.getLogger(__name__)
@@ -1,7 +1,7 @@
 import logging

 import requests
-from open_webui.apps.rag.search.main import SearchResult
+from open_webui.apps.retrieval.web.main import SearchResult
 from open_webui.env import SRC_LOG_LEVELS
 from yarl import URL

@@ -3,7 +3,7 @@ from typing import Optional
 from urllib.parse import urlencode

 import requests
-from open_webui.apps.rag.search.main import SearchResult, get_filtered_results
+from open_webui.apps.retrieval.web.main import SearchResult, get_filtered_results
 from open_webui.env import SRC_LOG_LEVELS

 log = logging.getLogger(__name__)
@@ -2,7 +2,7 @@ import logging
 from typing import Optional

 import requests
-from open_webui.apps.rag.search.main import SearchResult, get_filtered_results
+from open_webui.apps.retrieval.web.main import SearchResult, get_filtered_results
 from open_webui.env import SRC_LOG_LEVELS

 log = logging.getLogger(__name__)
@@ -3,7 +3,7 @@ import logging
 from typing import Optional

 import requests
-from open_webui.apps.rag.search.main import SearchResult, get_filtered_results
+from open_webui.apps.retrieval.web.main import SearchResult, get_filtered_results
 from open_webui.env import SRC_LOG_LEVELS

 log = logging.getLogger(__name__)
@@ -3,7 +3,7 @@ from typing import Optional
 from urllib.parse import urlencode

 import requests
-from open_webui.apps.rag.search.main import SearchResult, get_filtered_results
+from open_webui.apps.retrieval.web.main import SearchResult, get_filtered_results
 from open_webui.env import SRC_LOG_LEVELS

 log = logging.getLogger(__name__)
@@ -2,7 +2,7 @@ import logging
 from typing import Optional

 import requests
-from open_webui.apps.rag.search.main import SearchResult, get_filtered_results
+from open_webui.apps.retrieval.web.main import SearchResult, get_filtered_results
 from open_webui.env import SRC_LOG_LEVELS

 log = logging.getLogger(__name__)
@@ -1,7 +1,7 @@
 import logging

 import requests
-from open_webui.apps.rag.search.main import SearchResult
+from open_webui.apps.retrieval.web.main import SearchResult
 from open_webui.env import SRC_LOG_LEVELS

 log = logging.getLogger(__name__)
@@ -0,0 +1,97 @@
+import socket
+import urllib.parse
+import validators
+from typing import Union, Sequence, Iterator
+
+from langchain_community.document_loaders import (
+    WebBaseLoader,
+)
+from langchain_core.documents import Document
+
+
+from open_webui.constants import ERROR_MESSAGES
+from open_webui.config import ENABLE_RAG_LOCAL_WEB_FETCH
+from open_webui.env import SRC_LOG_LEVELS
+
+import logging
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["RAG"])
+
+
+def validate_url(url: Union[str, Sequence[str]]):
+    if isinstance(url, str):
+        if isinstance(validators.url(url), validators.ValidationError):
+            raise ValueError(ERROR_MESSAGES.INVALID_URL)
+        if not ENABLE_RAG_LOCAL_WEB_FETCH:
+            # Local web fetch is disabled, filter out any URLs that resolve to private IP addresses
+            parsed_url = urllib.parse.urlparse(url)
+            # Get IPv4 and IPv6 addresses
+            ipv4_addresses, ipv6_addresses = resolve_hostname(parsed_url.hostname)
+            # Check if any of the resolved addresses are private
+            # This is technically still vulnerable to DNS rebinding attacks, as we don't control WebBaseLoader
+            for ip in ipv4_addresses:
+                if validators.ipv4(ip, private=True):
+                    raise ValueError(ERROR_MESSAGES.INVALID_URL)
+            for ip in ipv6_addresses:
+                if validators.ipv6(ip, private=True):
+                    raise ValueError(ERROR_MESSAGES.INVALID_URL)
+        return True
+    elif isinstance(url, Sequence):
+        return all(validate_url(u) for u in url)
+    else:
+        return False
+
+
+def resolve_hostname(hostname):
+    # Get address information
+    addr_info = socket.getaddrinfo(hostname, None)
+
+    # Extract IP addresses from address information
+    ipv4_addresses = [info[4][0] for info in addr_info if info[0] == socket.AF_INET]
+    ipv6_addresses = [info[4][0] for info in addr_info if info[0] == socket.AF_INET6]
+
+    return ipv4_addresses, ipv6_addresses
+
+
+class SafeWebBaseLoader(WebBaseLoader):
+    """WebBaseLoader with enhanced error handling for URLs."""
+
+    def lazy_load(self) -> Iterator[Document]:
+        """Lazy load text from the url(s) in web_path with error handling."""
+        for path in self.web_paths:
+            try:
+                soup = self._scrape(path, bs_kwargs=self.bs_kwargs)
+                text = soup.get_text(**self.bs_get_text_kwargs)
+
+                # Build metadata
+                metadata = {"source": path}
+                if title := soup.find("title"):
+                    metadata["title"] = title.get_text()
+                if description := soup.find("meta", attrs={"name": "description"}):
+                    metadata["description"] = description.get(
+                        "content", "No description found."
+                    )
+                if html := soup.find("html"):
+                    metadata["language"] = html.get("lang", "No language found.")
+
+                yield Document(page_content=text, metadata=metadata)
+            except Exception as e:
+                # Log the error and continue with the next URL
+                log.error(f"Error loading {path}: {e}")
+
+
+def get_web_loader(
+    url: Union[str, Sequence[str]],
+    verify_ssl: bool = True,
+    requests_per_second: int = 2,
+):
+    # Check if the URL is valid
+    if not validate_url(url):
+        raise ValueError(ERROR_MESSAGES.INVALID_URL)
+    return SafeWebBaseLoader(
+        url,
+        verify_ssl=verify_ssl,
+        requests_per_second=requests_per_second,
+        continue_on_failure=True,
+    )
@@ -1,21 +1,123 @@
 import asyncio
-
 import socketio
-from open_webui.apps.webui.models.users import Users
-from open_webui.utils.utils import decode_token
+import logging
+import sys
+import time
+
+from open_webui.apps.webui.models.users import Users
+from open_webui.env import (
+    ENABLE_WEBSOCKET_SUPPORT,
+    WEBSOCKET_MANAGER,
+    WEBSOCKET_REDIS_URL,
+)
+from open_webui.utils.utils import decode_token
+from open_webui.apps.socket.utils import RedisDict
+
+from open_webui.env import (
+    GLOBAL_LOG_LEVEL,
+    SRC_LOG_LEVELS,
+)
+
+
+logging.basicConfig(stream=sys.stdout, level=GLOBAL_LOG_LEVEL)
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["SOCKET"])
+
+
+if WEBSOCKET_MANAGER == "redis":
+    mgr = socketio.AsyncRedisManager(WEBSOCKET_REDIS_URL)
+    sio = socketio.AsyncServer(
+        cors_allowed_origins=[],
+        async_mode="asgi",
+        transports=(
+            ["polling", "websocket"] if ENABLE_WEBSOCKET_SUPPORT else ["polling"]
+        ),
+        allow_upgrades=ENABLE_WEBSOCKET_SUPPORT,
+        always_connect=True,
+        client_manager=mgr,
+    )
+else:
+    sio = socketio.AsyncServer(
+        cors_allowed_origins=[],
+        async_mode="asgi",
+        transports=(
+            ["polling", "websocket"] if ENABLE_WEBSOCKET_SUPPORT else ["polling"]
+        ),
+        allow_upgrades=ENABLE_WEBSOCKET_SUPPORT,
+        always_connect=True,
+    )

-sio = socketio.AsyncServer(cors_allowed_origins=[], async_mode="asgi")
-app = socketio.ASGIApp(sio, socketio_path="/ws/socket.io")

 # Dictionary to maintain the user pool

-SESSION_POOL = {}
-USER_POOL = {}
-USAGE_POOL = {}
+if WEBSOCKET_MANAGER == "redis":
+    SESSION_POOL = RedisDict("open-webui:session_pool", redis_url=WEBSOCKET_REDIS_URL)
+    USER_POOL = RedisDict("open-webui:user_pool", redis_url=WEBSOCKET_REDIS_URL)
+    USAGE_POOL = RedisDict("open-webui:usage_pool", redis_url=WEBSOCKET_REDIS_URL)
+else:
+    SESSION_POOL = {}
+    USER_POOL = {}
+    USAGE_POOL = {}
+
+
 # Timeout duration in seconds
 TIMEOUT_DURATION = 3


+async def periodic_usage_pool_cleanup():
+    while True:
+        now = int(time.time())
+        for model_id, connections in list(USAGE_POOL.items()):
+            # Creating a list of sids to remove if they have timed out
+            expired_sids = [
+                sid
+                for sid, details in connections.items()
+                if now - details["updated_at"] > TIMEOUT_DURATION
+            ]
+
+            for sid in expired_sids:
+                del connections[sid]
+
+            if not connections:
+                log.debug(f"Cleaning up model {model_id} from usage pool")
+                del USAGE_POOL[model_id]
+            else:
+                USAGE_POOL[model_id] = connections
+
+            # Emit updated usage information after cleaning
+            await sio.emit("usage", {"models": get_models_in_use()})
+
+        await asyncio.sleep(TIMEOUT_DURATION)
+
+
+app = socketio.ASGIApp(
+    sio,
+    socketio_path="/ws/socket.io",
+)
+
+
+def get_models_in_use():
+    # List models that are currently in use
+    models_in_use = list(USAGE_POOL.keys())
+    return models_in_use
+
+
+@sio.on("usage")
+async def usage(sid, data):
+    model_id = data["model"]
+    # Record the timestamp for the last update
+    current_time = int(time.time())
+
+    # Store the new usage data and task
+    USAGE_POOL[model_id] = {
+        **(USAGE_POOL[model_id] if model_id in USAGE_POOL else {}),
+        sid: {"updated_at": current_time},
+    }
+
+    # Broadcast the usage data to all clients
+    await sio.emit("usage", {"models": get_models_in_use()})
+
+
@sio.event
 async def connect(sid, environ, auth):
    user = None
@@ -32,15 +134,14 @@ async def connect(sid, environ, auth):
            else:
                USER_POOL[user.id] = [sid]

-            print(f"user {user.name}({user.id}) connected with session ID {sid}")
-
-            await sio.emit("user-count", {"count": len(set(USER_POOL))})
+            # print(f"user {user.name}({user.id}) connected with session ID {sid}")
+            await sio.emit("user-count", {"count": len(USER_POOL.items())})
            await sio.emit("usage", {"models": get_models_in_use()})


@sio.on("user-join")
 async def user_join(sid, data):
-    print("user-join", sid, data)
+    # print("user-join", sid, data)

    auth = data["auth"] if "auth" in data else None
    if not auth or "token" not in auth:
@@ -60,67 +161,14 @@ 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))})
+    await sio.emit("user-count", {"count": len(USER_POOL.items())})


@sio.on("user-count")
 async def user_count(sid):
-    await sio.emit("user-count", {"count": len(set(USER_POOL))})
-
-
-def get_models_in_use():
-    # Aggregate all models in use
-    models_in_use = []
-    for model_id, data in USAGE_POOL.items():
-        models_in_use.append(model_id)
-
-    return models_in_use
-
-
-@sio.on("usage")
-async def usage(sid, data):
-    model_id = data["model"]
-
-    # Cancel previous callback if there is one
-    if model_id in USAGE_POOL:
-        USAGE_POOL[model_id]["callback"].cancel()
-
-    # Store the new usage data and task
-
-    if model_id in USAGE_POOL:
-        USAGE_POOL[model_id]["sids"].append(sid)
-        USAGE_POOL[model_id]["sids"] = list(set(USAGE_POOL[model_id]["sids"]))
-
-    else:
-        USAGE_POOL[model_id] = {"sids": [sid]}
-
-    # Schedule a task to remove the usage data after TIMEOUT_DURATION
-    USAGE_POOL[model_id]["callback"] = asyncio.create_task(
-        remove_after_timeout(sid, model_id)
-    )
-
-    # Broadcast the usage data to all clients
-    await sio.emit("usage", {"models": get_models_in_use()})
-
-
-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"])
-            USAGE_POOL[model_id]["sids"].remove(sid)
-            USAGE_POOL[model_id]["sids"] = list(set(USAGE_POOL[model_id]["sids"]))
-
-            if len(USAGE_POOL[model_id]["sids"]) == 0:
-                del USAGE_POOL[model_id]
-
-            # Broadcast the usage data to all clients
-            await sio.emit("usage", {"models": get_models_in_use()})
-    except asyncio.CancelledError:
-        # Task was cancelled due to new 'usage' event
-        pass
+    await sio.emit("user-count", {"count": len(USER_POOL.items())})


@sio.event
@@ -129,14 +177,15 @@ async def disconnect(sid):
        user_id = SESSION_POOL[sid]
        del SESSION_POOL[sid]

-        USER_POOL[user_id].remove(sid)
+        USER_POOL[user_id] = [_sid for _sid in USER_POOL[user_id] if _sid != sid]

        if len(USER_POOL[user_id]) == 0:
            del USER_POOL[user_id]

        await sio.emit("user-count", {"count": len(USER_POOL)})
    else:
-        print(f"Unknown session ID {sid} disconnected")
+        pass
+        # print(f"Unknown session ID {sid} disconnected")


 def get_event_emitter(request_info):
@@ -0,0 +1,59 @@
+import json
+import redis
+
+
+class RedisDict:
+    def __init__(self, name, redis_url):
+        self.name = name
+        self.redis = redis.Redis.from_url(redis_url, decode_responses=True)
+
+    def __setitem__(self, key, value):
+        serialized_value = json.dumps(value)
+        self.redis.hset(self.name, key, serialized_value)
+
+    def __getitem__(self, key):
+        value = self.redis.hget(self.name, key)
+        if value is None:
+            raise KeyError(key)
+        return json.loads(value)
+
+    def __delitem__(self, key):
+        result = self.redis.hdel(self.name, key)
+        if result == 0:
+            raise KeyError(key)
+
+    def __contains__(self, key):
+        return self.redis.hexists(self.name, key)
+
+    def __len__(self):
+        return self.redis.hlen(self.name)
+
+    def keys(self):
+        return self.redis.hkeys(self.name)
+
+    def values(self):
+        return [json.loads(v) for v in self.redis.hvals(self.name)]
+
+    def items(self):
+        return [(k, json.loads(v)) for k, v in self.redis.hgetall(self.name).items()]
+
+    def get(self, key, default=None):
+        try:
+            return self[key]
+        except KeyError:
+            return default
+
+    def clear(self):
+        self.redis.delete(self.name)
+
+    def update(self, other=None, **kwargs):
+        if other is not None:
+            for k, v in other.items() if hasattr(other, "items") else other:
+                self[k] = v
+        for k, v in kwargs.items():
+            self[k] = v
+
+    def setdefault(self, key, default=None):
+        if key not in self:
+            self[key] = default
+        return self[key]
@@ -4,11 +4,20 @@ from contextlib import contextmanager
 from typing import Any, Optional

 from open_webui.apps.webui.internal.wrappers import register_connection
-from open_webui.env import OPEN_WEBUI_DIR, DATABASE_URL, SRC_LOG_LEVELS
+from open_webui.env import (
+    OPEN_WEBUI_DIR,
+    DATABASE_URL,
+    SRC_LOG_LEVELS,
+    DATABASE_POOL_MAX_OVERFLOW,
+    DATABASE_POOL_RECYCLE,
+    DATABASE_POOL_SIZE,
+    DATABASE_POOL_TIMEOUT,
+)
 from peewee_migrate import Router
 from sqlalchemy import Dialect, create_engine, types
 from sqlalchemy.ext.declarative import declarative_base
 from sqlalchemy.orm import scoped_session, sessionmaker
+from sqlalchemy.pool import QueuePool, NullPool
 from sqlalchemy.sql.type_api import _T
 from typing_extensions import Self

@@ -71,7 +80,20 @@ if "sqlite" in SQLALCHEMY_DATABASE_URL:
        SQLALCHEMY_DATABASE_URL, connect_args={"check_same_thread": False}
    )
 else:
-    engine = create_engine(SQLALCHEMY_DATABASE_URL, pool_pre_ping=True)
+    if DATABASE_POOL_SIZE > 0:
+        engine = create_engine(
+            SQLALCHEMY_DATABASE_URL,
+            pool_size=DATABASE_POOL_SIZE,
+            max_overflow=DATABASE_POOL_MAX_OVERFLOW,
+            pool_timeout=DATABASE_POOL_TIMEOUT,
+            pool_recycle=DATABASE_POOL_RECYCLE,
+            pool_pre_ping=True,
+            poolclass=QueuePool,
+        )
+    else:
+        engine = create_engine(
+            SQLALCHEMY_DATABASE_URL, pool_pre_ping=True, poolclass=NullPool
+        )


 SessionLocal = sessionmaker(
@@ -1,6 +1,7 @@
 import inspect
 import json
 import logging
+import time
 from typing import AsyncGenerator, Generator, Iterator

 from open_webui.apps.socket.main import get_event_call, get_event_emitter
@@ -9,13 +10,15 @@ from open_webui.apps.webui.models.models import Models
 from open_webui.apps.webui.routers import (
    auths,
    chats,
+    folders,
    configs,
-    documents,
    files,
    functions,
    memories,
    models,
+    knowledge,
    prompts,
+    evaluations,
    tools,
    users,
    utils,
@@ -31,10 +34,17 @@ from open_webui.config import (
    ENABLE_LOGIN_FORM,
    ENABLE_MESSAGE_RATING,
    ENABLE_SIGNUP,
+    ENABLE_EVALUATION_ARENA_MODELS,
+    EVALUATION_ARENA_MODELS,
+    DEFAULT_ARENA_MODEL,
    JWT_EXPIRES_IN,
+    ENABLE_OAUTH_ROLE_MANAGEMENT,
+    OAUTH_ROLES_CLAIM,
    OAUTH_EMAIL_CLAIM,
    OAUTH_PICTURE_CLAIM,
    OAUTH_USERNAME_CLAIM,
+    OAUTH_ALLOWED_ROLES,
+    OAUTH_ADMIN_ROLES,
    SHOW_ADMIN_DETAILS,
    USER_PERMISSIONS,
    WEBHOOK_URL,
@@ -51,11 +61,15 @@ 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()
@@ -85,10 +99,18 @@ app.state.config.BANNERS = WEBUI_BANNERS
 app.state.config.ENABLE_COMMUNITY_SHARING = ENABLE_COMMUNITY_SHARING
 app.state.config.ENABLE_MESSAGE_RATING = ENABLE_MESSAGE_RATING

+app.state.config.ENABLE_EVALUATION_ARENA_MODELS = ENABLE_EVALUATION_ARENA_MODELS
+app.state.config.EVALUATION_ARENA_MODELS = EVALUATION_ARENA_MODELS
+
 app.state.config.OAUTH_USERNAME_CLAIM = OAUTH_USERNAME_CLAIM
 app.state.config.OAUTH_PICTURE_CLAIM = OAUTH_PICTURE_CLAIM
 app.state.config.OAUTH_EMAIL_CLAIM = OAUTH_EMAIL_CLAIM

+app.state.config.ENABLE_OAUTH_ROLE_MANAGEMENT = ENABLE_OAUTH_ROLE_MANAGEMENT
+app.state.config.OAUTH_ROLES_CLAIM = OAUTH_ROLES_CLAIM
+app.state.config.OAUTH_ALLOWED_ROLES = OAUTH_ALLOWED_ROLES
+app.state.config.OAUTH_ADMIN_ROLES = OAUTH_ADMIN_ROLES
+
 app.state.MODELS = {}
 app.state.TOOLS = {}
 app.state.FUNCTIONS = {}
@@ -103,19 +125,24 @@ app.add_middleware(


 app.include_router(configs.router, prefix="/configs", tags=["configs"])
+
 app.include_router(auths.router, prefix="/auths", tags=["auths"])
 app.include_router(users.router, prefix="/users", tags=["users"])
+
 app.include_router(chats.router, prefix="/chats", tags=["chats"])

-app.include_router(documents.router, prefix="/documents", tags=["documents"])
 app.include_router(models.router, prefix="/models", tags=["models"])
+app.include_router(knowledge.router, prefix="/knowledge", tags=["knowledge"])
 app.include_router(prompts.router, prefix="/prompts", tags=["prompts"])
-
-app.include_router(memories.router, prefix="/memories", tags=["memories"])
-app.include_router(files.router, prefix="/files", tags=["files"])
 app.include_router(tools.router, prefix="/tools", tags=["tools"])
 app.include_router(functions.router, prefix="/functions", tags=["functions"])

+app.include_router(memories.router, prefix="/memories", tags=["memories"])
+app.include_router(evaluations.router, prefix="/evaluations", tags=["evaluations"])
+
+app.include_router(folders.router, prefix="/folders", tags=["folders"])
+app.include_router(files.router, prefix="/files", tags=["files"])
+
 app.include_router(utils.router, prefix="/utils", tags=["utils"])


@@ -129,6 +156,47 @@ async def get_status():
    }


+async def get_all_models():
+    models = []
+    pipe_models = await get_pipe_models()
+    models = models + pipe_models
+
+    if app.state.config.ENABLE_EVALUATION_ARENA_MODELS:
+        arena_models = []
+        if len(app.state.config.EVALUATION_ARENA_MODELS) > 0:
+            arena_models = [
+                {
+                    "id": model["id"],
+                    "name": model["name"],
+                    "info": {
+                        "meta": model["meta"],
+                    },
+                    "object": "model",
+                    "created": int(time.time()),
+                    "owned_by": "arena",
+                    "arena": True,
+                }
+                for model in app.state.config.EVALUATION_ARENA_MODELS
+            ]
+        else:
+            # Add default arena model
+            arena_models = [
+                {
+                    "id": DEFAULT_ARENA_MODEL["id"],
+                    "name": DEFAULT_ARENA_MODEL["name"],
+                    "info": {
+                        "meta": DEFAULT_ARENA_MODEL["meta"],
+                    },
+                    "object": "model",
+                    "created": int(time.time()),
+                    "owned_by": "arena",
+                    "arena": True,
+                }
+            ]
+        models = models + arena_models
+    return models
+
+
 def get_function_module(pipe_id: str):
    # Check if function is already loaded
    if pipe_id not in app.state.FUNCTIONS:
@@ -152,29 +220,33 @@ 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 +255,6 @@ 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(
                {
@@ -281,17 +351,20 @@ async def generate_function_chat_completion(form_data, user):
    __event_emitter__ = None
    __event_call__ = None
    __task__ = None
+    __task_body__ = None

    if metadata:
        if all(k in metadata for k in ("session_id", "chat_id", "message_id")):
            __event_emitter__ = get_event_emitter(metadata)
            __event_call__ = get_event_call(metadata)
        __task__ = metadata.get("task", None)
+        __task_body__ = metadata.get("task_body", None)

    extra_params = {
        "__event_emitter__": __event_emitter__,
        "__event_call__": __event_call__,
        "__task__": __task__,
+        "__task_body__": __task_body__,
        "__files__": files,
        "__user__": {
            "id": user.id,
@@ -326,7 +399,7 @@ async def generate_function_chat_completion(form_data, user):
    pipe = function_module.pipe
    params = get_function_params(function_module, form_data, user, extra_params)

-    if form_data["stream"]:
+    if form_data.get("stream", False):

        async def stream_content():
            try:
@@ -4,8 +4,13 @@ import uuid
 from typing import Optional

 from open_webui.apps.webui.internal.db import Base, get_db
+from open_webui.apps.webui.models.tags import TagModel, Tag, Tags
+
+
 from pydantic import BaseModel, ConfigDict
-from sqlalchemy import BigInteger, Boolean, Column, String, Text
+from sqlalchemy import BigInteger, Boolean, Column, String, Text, JSON
+from sqlalchemy import or_, func, select, and_, text
+from sqlalchemy.sql import exists

 ####################
 # Chat DB Schema
@@ -18,13 +23,17 @@ class Chat(Base):
    id = Column(String, primary_key=True)
    user_id = Column(String)
    title = Column(Text)
-    chat = Column(Text)  # Save Chat JSON as Text
+    chat = Column(JSON)

    created_at = Column(BigInteger)
    updated_at = Column(BigInteger)

    share_id = Column(Text, unique=True, nullable=True)
    archived = Column(Boolean, default=False)
+    pinned = Column(Boolean, default=False, nullable=True)
+
+    meta = Column(JSON, server_default="{}")
+    folder_id = Column(Text, nullable=True)


 class ChatModel(BaseModel):
@@ -33,13 +42,17 @@ class ChatModel(BaseModel):
    id: str
    user_id: str
    title: str
-    chat: str
+    chat: dict

    created_at: int  # timestamp in epoch
    updated_at: int  # timestamp in epoch

    share_id: Optional[str] = None
    archived: bool = False
+    pinned: Optional[bool] = False
+
+    meta: dict = {}
+    folder_id: Optional[str] = None


 ####################
@@ -51,6 +64,17 @@ class ChatForm(BaseModel):
    chat: dict


+class ChatImportForm(ChatForm):
+    meta: Optional[dict] = {}
+    pinned: Optional[bool] = False
+    folder_id: Optional[str] = None
+
+
+class ChatTitleMessagesForm(BaseModel):
+    title: str
+    messages: list[dict]
+
+
 class ChatTitleForm(BaseModel):
    title: str

@@ -64,6 +88,9 @@ class ChatResponse(BaseModel):
    created_at: int  # timestamp in epoch
    share_id: Optional[str] = None  # id of the chat to be shared
    archived: bool
+    pinned: Optional[bool] = False
+    meta: dict = {}
+    folder_id: Optional[str] = None


 class ChatTitleIdResponse(BaseModel):
@@ -86,7 +113,36 @@ class ChatTable:
                        if "title" in form_data.chat
                        else "New Chat"
                    ),
-                    "chat": json.dumps(form_data.chat),
+                    "chat": form_data.chat,
+                    "created_at": int(time.time()),
+                    "updated_at": int(time.time()),
+                }
+            )
+
+            result = Chat(**chat.model_dump())
+            db.add(result)
+            db.commit()
+            db.refresh(result)
+            return ChatModel.model_validate(result) if result else None
+
+    def import_chat(
+        self, user_id: str, form_data: ChatImportForm
+    ) -> Optional[ChatModel]:
+        with get_db() as db:
+            id = str(uuid.uuid4())
+            chat = ChatModel(
+                **{
+                    "id": id,
+                    "user_id": user_id,
+                    "title": (
+                        form_data.chat["title"]
+                        if "title" in form_data.chat
+                        else "New Chat"
+                    ),
+                    "chat": form_data.chat,
+                    "meta": form_data.meta,
+                    "pinned": form_data.pinned,
+                    "folder_id": form_data.folder_id,
                    "created_at": int(time.time()),
                    "updated_at": int(time.time()),
                }
@@ -101,14 +157,14 @@ class ChatTable:
    def update_chat_by_id(self, id: str, chat: dict) -> Optional[ChatModel]:
        try:
            with get_db() as db:
-                chat_obj = db.get(Chat, id)
-                chat_obj.chat = json.dumps(chat)
-                chat_obj.title = chat["title"] if "title" in chat else "New Chat"
-                chat_obj.updated_at = int(time.time())
+                chat_item = db.get(Chat, id)
+                chat_item.chat = chat
+                chat_item.title = chat["title"] if "title" in chat else "New Chat"
+                chat_item.updated_at = int(time.time())
                db.commit()
-                db.refresh(chat_obj)
+                db.refresh(chat_item)

-                return ChatModel.model_validate(chat_obj)
+                return ChatModel.model_validate(chat_item)
        except Exception:
            return None

@@ -182,11 +238,24 @@ class ChatTable:
        except Exception:
            return None

+    def toggle_chat_pinned_by_id(self, id: str) -> Optional[ChatModel]:
+        try:
+            with get_db() as db:
+                chat = db.get(Chat, id)
+                chat.pinned = not chat.pinned
+                chat.updated_at = int(time.time())
+                db.commit()
+                db.refresh(chat)
+                return ChatModel.model_validate(chat)
+        except Exception:
+            return None
+
    def toggle_chat_archive_by_id(self, id: str) -> Optional[ChatModel]:
        try:
            with get_db() as db:
                chat = db.get(Chat, id)
                chat.archived = not chat.archived
+                chat.updated_at = int(time.time())
                db.commit()
                db.refresh(chat)
                return ChatModel.model_validate(chat)
@@ -223,14 +292,18 @@ class ChatTable:
        limit: int = 50,
    ) -> list[ChatModel]:
        with get_db() as db:
-            query = db.query(Chat).filter_by(user_id=user_id)
+            query = db.query(Chat).filter_by(user_id=user_id).filter_by(folder_id=None)
            if not include_archived:
                query = query.filter_by(archived=False)
-            all_chats = (
-                query.order_by(Chat.updated_at.desc())
-                # .limit(limit).offset(skip)
-                .all()
-            )
+
+            query = query.order_by(Chat.updated_at.desc())
+
+            if skip:
+                query = query.offset(skip)
+            if limit:
+                query = query.limit(limit)
+
+            all_chats = query.all()
            return [ChatModel.model_validate(chat) for chat in all_chats]

    def get_chat_title_id_list_by_user_id(
@@ -241,7 +314,9 @@ class ChatTable:
        limit: Optional[int] = None,
    ) -> list[ChatTitleIdResponse]:
        with get_db() as db:
-            query = db.query(Chat).filter_by(user_id=user_id)
+            query = db.query(Chat).filter_by(user_id=user_id).filter_by(folder_id=None)
+            query = query.filter(or_(Chat.pinned == False, Chat.pinned == None))
+
            if not include_archived:
                query = query.filter_by(archived=False)

@@ -249,10 +324,10 @@ class ChatTable:
                Chat.id, Chat.title, Chat.updated_at, Chat.created_at
            )

-            if limit:
-                query = query.limit(limit)
            if skip:
                query = query.offset(skip)
+            if limit:
+                query = query.limit(limit)

            all_chats = query.all()

@@ -328,6 +403,15 @@ class ChatTable:
            )
            return [ChatModel.model_validate(chat) for chat in all_chats]

+    def get_pinned_chats_by_user_id(self, user_id: str) -> list[ChatModel]:
+        with get_db() as db:
+            all_chats = (
+                db.query(Chat)
+                .filter_by(user_id=user_id, pinned=True, archived=False)
+                .order_by(Chat.updated_at.desc())
+            )
+            return [ChatModel.model_validate(chat) for chat in all_chats]
+
    def get_archived_chats_by_user_id(self, user_id: str) -> list[ChatModel]:
        with get_db() as db:
            all_chats = (
@@ -337,6 +421,329 @@ class ChatTable:
            )
            return [ChatModel.model_validate(chat) for chat in all_chats]

+    def get_chats_by_user_id_and_search_text(
+        self,
+        user_id: str,
+        search_text: str,
+        include_archived: bool = False,
+        skip: int = 0,
+        limit: int = 60,
+    ) -> list[ChatModel]:
+        """
+        Filters chats based on a search query using Python, allowing pagination using skip and limit.
+        """
+        search_text = search_text.lower().strip()
+
+        if not search_text:
+            return self.get_chat_list_by_user_id(user_id, include_archived, skip, limit)
+
+        search_text_words = search_text.split(" ")
+
+        # search_text might contain 'tag:tag_name' format so we need to extract the tag_name, split the search_text and remove the tags
+        tag_ids = [
+            word.replace("tag:", "").replace(" ", "_").lower()
+            for word in search_text_words
+            if word.startswith("tag:")
+        ]
+
+        search_text_words = [
+            word for word in search_text_words if not word.startswith("tag:")
+        ]
+
+        search_text = " ".join(search_text_words)
+
+        with get_db() as db:
+            query = db.query(Chat).filter(Chat.user_id == user_id)
+
+            if not include_archived:
+                query = query.filter(Chat.archived == False)
+
+            query = query.order_by(Chat.updated_at.desc())
+
+            # Check if the database dialect is either 'sqlite' or 'postgresql'
+            dialect_name = db.bind.dialect.name
+            if dialect_name == "sqlite":
+                # SQLite case: using JSON1 extension for JSON searching
+                query = query.filter(
+                    (
+                        Chat.title.ilike(
+                            f"%{search_text}%"
+                        )  # Case-insensitive search in title
+                        | text(
+                            """
+                            EXISTS (
+                                SELECT 1 
+                                FROM json_each(Chat.chat, '$.messages') AS message 
+                                WHERE LOWER(message.value->>'content') LIKE '%' || :search_text || '%'
+                            )
+                            """
+                        )
+                    ).params(search_text=search_text)
+                )
+
+                # Check if there are any tags to filter, it should have all the tags
+                if "none" in tag_ids:
+                    query = query.filter(
+                        text(
+                            """
+                            NOT EXISTS (
+                                SELECT 1
+                                FROM json_each(Chat.meta, '$.tags') AS tag
+                            )
+                            """
+                        )
+                    )
+                elif tag_ids:
+                    query = query.filter(
+                        and_(
+                            *[
+                                text(
+                                    f"""
+                                    EXISTS (
+                                        SELECT 1
+                                        FROM json_each(Chat.meta, '$.tags') AS tag
+                                        WHERE tag.value = :tag_id_{tag_idx}
+                                    )
+                                    """
+                                ).params(**{f"tag_id_{tag_idx}": tag_id})
+                                for tag_idx, tag_id in enumerate(tag_ids)
+                            ]
+                        )
+                    )
+
+            elif dialect_name == "postgresql":
+                # PostgreSQL relies on proper JSON query for search
+                query = query.filter(
+                    (
+                        Chat.title.ilike(
+                            f"%{search_text}%"
+                        )  # Case-insensitive search in title
+                        | text(
+                            """
+                            EXISTS (
+                                SELECT 1
+                                FROM json_array_elements(Chat.chat->'messages') AS message
+                                WHERE LOWER(message->>'content') LIKE '%' || :search_text || '%'
+                            )
+                            """
+                        )
+                    ).params(search_text=search_text)
+                )
+
+                # Check if there are any tags to filter, it should have all the tags
+                if "none" in tag_ids:
+                    query = query.filter(
+                        text(
+                            """
+                            NOT EXISTS (
+                                SELECT 1
+                                FROM json_array_elements_text(Chat.meta->'tags') AS tag
+                            )
+                            """
+                        )
+                    )
+                elif tag_ids:
+                    query = query.filter(
+                        and_(
+                            *[
+                                text(
+                                    f"""
+                                    EXISTS (
+                                        SELECT 1
+                                        FROM json_array_elements_text(Chat.meta->'tags') AS tag
+                                        WHERE tag = :tag_id_{tag_idx}
+                                    )
+                                    """
+                                ).params(**{f"tag_id_{tag_idx}": tag_id})
+                                for tag_idx, tag_id in enumerate(tag_ids)
+                            ]
+                        )
+                    )
+            else:
+                raise NotImplementedError(
+                    f"Unsupported dialect: {db.bind.dialect.name}"
+                )
+
+            # Perform pagination at the SQL level
+            all_chats = query.offset(skip).limit(limit).all()
+
+            print(len(all_chats))
+
+            # Validate and return chats
+            return [ChatModel.model_validate(chat) for chat in all_chats]
+
+    def get_chats_by_folder_id_and_user_id(
+        self, folder_id: str, user_id: str
+    ) -> list[ChatModel]:
+        with get_db() as db:
+            query = db.query(Chat).filter_by(folder_id=folder_id, user_id=user_id)
+            query = query.filter(or_(Chat.pinned == False, Chat.pinned == None))
+            query = query.filter_by(archived=False)
+
+            query = query.order_by(Chat.updated_at.desc())
+
+            all_chats = query.all()
+            return [ChatModel.model_validate(chat) for chat in all_chats]
+
+    def get_chats_by_folder_ids_and_user_id(
+        self, folder_ids: list[str], user_id: str
+    ) -> list[ChatModel]:
+        with get_db() as db:
+            query = db.query(Chat).filter(
+                Chat.folder_id.in_(folder_ids), Chat.user_id == user_id
+            )
+            query = query.filter(or_(Chat.pinned == False, Chat.pinned == None))
+            query = query.filter_by(archived=False)
+
+            query = query.order_by(Chat.updated_at.desc())
+
+            all_chats = query.all()
+            return [ChatModel.model_validate(chat) for chat in all_chats]
+
+    def update_chat_folder_id_by_id_and_user_id(
+        self, id: str, user_id: str, folder_id: str
+    ) -> Optional[ChatModel]:
+        try:
+            with get_db() as db:
+                chat = db.get(Chat, id)
+                chat.folder_id = folder_id
+                chat.updated_at = int(time.time())
+                chat.pinned = False
+                db.commit()
+                db.refresh(chat)
+                return ChatModel.model_validate(chat)
+        except Exception:
+            return None
+
+    def get_chat_tags_by_id_and_user_id(self, id: str, user_id: str) -> list[TagModel]:
+        with get_db() as db:
+            chat = db.get(Chat, id)
+            tags = chat.meta.get("tags", [])
+            return [Tags.get_tag_by_name_and_user_id(tag, user_id) for tag in tags]
+
+    def get_chat_list_by_user_id_and_tag_name(
+        self, user_id: str, tag_name: str, skip: int = 0, limit: int = 50
+    ) -> list[ChatModel]:
+        with get_db() as db:
+            query = db.query(Chat).filter_by(user_id=user_id)
+            tag_id = tag_name.replace(" ", "_").lower()
+
+            print(db.bind.dialect.name)
+            if db.bind.dialect.name == "sqlite":
+                # SQLite JSON1 querying for tags within the meta JSON field
+                query = query.filter(
+                    text(
+                        f"EXISTS (SELECT 1 FROM json_each(Chat.meta, '$.tags') WHERE json_each.value = :tag_id)"
+                    )
+                ).params(tag_id=tag_id)
+            elif db.bind.dialect.name == "postgresql":
+                # PostgreSQL JSON query for tags within the meta JSON field (for `json` type)
+                query = query.filter(
+                    text(
+                        "EXISTS (SELECT 1 FROM json_array_elements_text(Chat.meta->'tags') elem WHERE elem = :tag_id)"
+                    )
+                ).params(tag_id=tag_id)
+            else:
+                raise NotImplementedError(
+                    f"Unsupported dialect: {db.bind.dialect.name}"
+                )
+
+            all_chats = query.all()
+            print("all_chats", all_chats)
+            return [ChatModel.model_validate(chat) for chat in all_chats]
+
+    def add_chat_tag_by_id_and_user_id_and_tag_name(
+        self, id: str, user_id: str, tag_name: str
+    ) -> Optional[ChatModel]:
+        tag = Tags.get_tag_by_name_and_user_id(tag_name, user_id)
+        if tag is None:
+            tag = Tags.insert_new_tag(tag_name, user_id)
+        try:
+            with get_db() as db:
+                chat = db.get(Chat, id)
+
+                tag_id = tag.id
+                if tag_id not in chat.meta.get("tags", []):
+                    chat.meta = {
+                        **chat.meta,
+                        "tags": list(set(chat.meta.get("tags", []) + [tag_id])),
+                    }
+
+                db.commit()
+                db.refresh(chat)
+                return ChatModel.model_validate(chat)
+        except Exception:
+            return None
+
+    def count_chats_by_tag_name_and_user_id(self, tag_name: str, user_id: str) -> int:
+        with get_db() as db:  # Assuming `get_db()` returns a session object
+            query = db.query(Chat).filter_by(user_id=user_id, archived=False)
+
+            # Normalize the tag_name for consistency
+            tag_id = tag_name.replace(" ", "_").lower()
+
+            if db.bind.dialect.name == "sqlite":
+                # SQLite JSON1 support for querying the tags inside the `meta` JSON field
+                query = query.filter(
+                    text(
+                        f"EXISTS (SELECT 1 FROM json_each(Chat.meta, '$.tags') WHERE json_each.value = :tag_id)"
+                    )
+                ).params(tag_id=tag_id)
+
+            elif db.bind.dialect.name == "postgresql":
+                # PostgreSQL JSONB support for querying the tags inside the `meta` JSON field
+                query = query.filter(
+                    text(
+                        "EXISTS (SELECT 1 FROM json_array_elements_text(Chat.meta->'tags') elem WHERE elem = :tag_id)"
+                    )
+                ).params(tag_id=tag_id)
+
+            else:
+                raise NotImplementedError(
+                    f"Unsupported dialect: {db.bind.dialect.name}"
+                )
+
+            # Get the count of matching records
+            count = query.count()
+
+            # Debugging output for inspection
+            print(f"Count of chats for tag '{tag_name}':", count)
+
+            return count
+
+    def delete_tag_by_id_and_user_id_and_tag_name(
+        self, id: str, user_id: str, tag_name: str
+    ) -> bool:
+        try:
+            with get_db() as db:
+                chat = db.get(Chat, id)
+                tags = chat.meta.get("tags", [])
+                tag_id = tag_name.replace(" ", "_").lower()
+
+                tags = [tag for tag in tags if tag != tag_id]
+                chat.meta = {
+                    **chat.meta,
+                    "tags": list(set(tags)),
+                }
+                db.commit()
+                return True
+        except Exception:
+            return False
+
+    def delete_all_tags_by_id_and_user_id(self, id: str, user_id: str) -> bool:
+        try:
+            with get_db() as db:
+                chat = db.get(Chat, id)
+                chat.meta = {
+                    **chat.meta,
+                    "tags": [],
+                }
+                db.commit()
+
+                return True
+        except Exception:
+            return False
+
    def delete_chat_by_id(self, id: str) -> bool:
        try:
            with get_db() as db:
@@ -369,6 +776,18 @@ class ChatTable:
        except Exception:
            return False

+    def delete_chats_by_user_id_and_folder_id(
+        self, user_id: str, folder_id: str
+    ) -> bool:
+        try:
+            with get_db() as db:
+                db.query(Chat).filter_by(user_id=user_id, folder_id=folder_id).delete()
+                db.commit()
+
+                return True
+        except Exception:
+            return False
+
    def delete_shared_chats_by_user_id(self, user_id: str) -> bool:
        try:
            with get_db() as db:
@@ -0,0 +1,254 @@
+import logging
+import time
+import uuid
+from typing import Optional
+
+from open_webui.apps.webui.internal.db import Base, get_db
+from open_webui.apps.webui.models.chats import Chats
+
+from open_webui.env import SRC_LOG_LEVELS
+from pydantic import BaseModel, ConfigDict
+from sqlalchemy import BigInteger, Column, Text, JSON, Boolean
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["MODELS"])
+
+
+####################
+# Feedback DB Schema
+####################
+
+
+class Feedback(Base):
+    __tablename__ = "feedback"
+    id = Column(Text, primary_key=True)
+    user_id = Column(Text)
+    version = Column(BigInteger, default=0)
+    type = Column(Text)
+    data = Column(JSON, nullable=True)
+    meta = Column(JSON, nullable=True)
+    snapshot = Column(JSON, nullable=True)
+    created_at = Column(BigInteger)
+    updated_at = Column(BigInteger)
+
+
+class FeedbackModel(BaseModel):
+    id: str
+    user_id: str
+    version: int
+    type: str
+    data: Optional[dict] = None
+    meta: Optional[dict] = None
+    snapshot: Optional[dict] = None
+    created_at: int
+    updated_at: int
+
+    model_config = ConfigDict(from_attributes=True)
+
+
+####################
+# Forms
+####################
+
+
+class FeedbackResponse(BaseModel):
+    id: str
+    user_id: str
+    version: int
+    type: str
+    data: Optional[dict] = None
+    meta: Optional[dict] = None
+    created_at: int
+    updated_at: int
+
+
+class RatingData(BaseModel):
+    rating: Optional[str | int] = None
+    model_id: Optional[str] = None
+    sibling_model_ids: Optional[list[str]] = None
+    reason: Optional[str] = None
+    comment: Optional[str] = None
+    model_config = ConfigDict(extra="allow", protected_namespaces=())
+
+
+class MetaData(BaseModel):
+    arena: Optional[bool] = None
+    chat_id: Optional[str] = None
+    message_id: Optional[str] = None
+    tags: Optional[list[str]] = None
+    model_config = ConfigDict(extra="allow")
+
+
+class SnapshotData(BaseModel):
+    chat: Optional[dict] = None
+    model_config = ConfigDict(extra="allow")
+
+
+class FeedbackForm(BaseModel):
+    type: str
+    data: Optional[RatingData] = None
+    meta: Optional[dict] = None
+    snapshot: Optional[SnapshotData] = None
+    model_config = ConfigDict(extra="allow")
+
+
+class FeedbackTable:
+    def insert_new_feedback(
+        self, user_id: str, form_data: FeedbackForm
+    ) -> Optional[FeedbackModel]:
+        with get_db() as db:
+            id = str(uuid.uuid4())
+            feedback = FeedbackModel(
+                **{
+                    "id": id,
+                    "user_id": user_id,
+                    "version": 0,
+                    **form_data.model_dump(),
+                    "created_at": int(time.time()),
+                    "updated_at": int(time.time()),
+                }
+            )
+            try:
+                result = Feedback(**feedback.model_dump())
+                db.add(result)
+                db.commit()
+                db.refresh(result)
+                if result:
+                    return FeedbackModel.model_validate(result)
+                else:
+                    return None
+            except Exception as e:
+                print(e)
+                return None
+
+    def get_feedback_by_id(self, id: str) -> Optional[FeedbackModel]:
+        try:
+            with get_db() as db:
+                feedback = db.query(Feedback).filter_by(id=id).first()
+                if not feedback:
+                    return None
+                return FeedbackModel.model_validate(feedback)
+        except Exception:
+            return None
+
+    def get_feedback_by_id_and_user_id(
+        self, id: str, user_id: str
+    ) -> Optional[FeedbackModel]:
+        try:
+            with get_db() as db:
+                feedback = db.query(Feedback).filter_by(id=id, user_id=user_id).first()
+                if not feedback:
+                    return None
+                return FeedbackModel.model_validate(feedback)
+        except Exception:
+            return None
+
+    def get_all_feedbacks(self) -> list[FeedbackModel]:
+        with get_db() as db:
+            return [
+                FeedbackModel.model_validate(feedback)
+                for feedback in db.query(Feedback)
+                .order_by(Feedback.updated_at.desc())
+                .all()
+            ]
+
+    def get_feedbacks_by_type(self, type: str) -> list[FeedbackModel]:
+        with get_db() as db:
+            return [
+                FeedbackModel.model_validate(feedback)
+                for feedback in db.query(Feedback)
+                .filter_by(type=type)
+                .order_by(Feedback.updated_at.desc())
+                .all()
+            ]
+
+    def get_feedbacks_by_user_id(self, user_id: str) -> list[FeedbackModel]:
+        with get_db() as db:
+            return [
+                FeedbackModel.model_validate(feedback)
+                for feedback in db.query(Feedback)
+                .filter_by(user_id=user_id)
+                .order_by(Feedback.updated_at.desc())
+                .all()
+            ]
+
+    def update_feedback_by_id(
+        self, id: str, form_data: FeedbackForm
+    ) -> Optional[FeedbackModel]:
+        with get_db() as db:
+            feedback = db.query(Feedback).filter_by(id=id).first()
+            if not feedback:
+                return None
+
+            if form_data.data:
+                feedback.data = form_data.data.model_dump()
+            if form_data.meta:
+                feedback.meta = form_data.meta
+            if form_data.snapshot:
+                feedback.snapshot = form_data.snapshot.model_dump()
+
+            feedback.updated_at = int(time.time())
+
+            db.commit()
+            return FeedbackModel.model_validate(feedback)
+
+    def update_feedback_by_id_and_user_id(
+        self, id: str, user_id: str, form_data: FeedbackForm
+    ) -> Optional[FeedbackModel]:
+        with get_db() as db:
+            feedback = db.query(Feedback).filter_by(id=id, user_id=user_id).first()
+            if not feedback:
+                return None
+
+            if form_data.data:
+                feedback.data = form_data.data.model_dump()
+            if form_data.meta:
+                feedback.meta = form_data.meta
+            if form_data.snapshot:
+                feedback.snapshot = form_data.snapshot.model_dump()
+
+            feedback.updated_at = int(time.time())
+
+            db.commit()
+            return FeedbackModel.model_validate(feedback)
+
+    def delete_feedback_by_id(self, id: str) -> bool:
+        with get_db() as db:
+            feedback = db.query(Feedback).filter_by(id=id).first()
+            if not feedback:
+                return False
+            db.delete(feedback)
+            db.commit()
+            return True
+
+    def delete_feedback_by_id_and_user_id(self, id: str, user_id: str) -> bool:
+        with get_db() as db:
+            feedback = db.query(Feedback).filter_by(id=id, user_id=user_id).first()
+            if not feedback:
+                return False
+            db.delete(feedback)
+            db.commit()
+            return True
+
+    def delete_feedbacks_by_user_id(self, user_id: str) -> bool:
+        with get_db() as db:
+            feedbacks = db.query(Feedback).filter_by(user_id=user_id).all()
+            if not feedbacks:
+                return False
+            for feedback in feedbacks:
+                db.delete(feedback)
+            db.commit()
+            return True
+
+    def delete_all_feedbacks(self) -> bool:
+        with get_db() as db:
+            feedbacks = db.query(Feedback).all()
+            if not feedbacks:
+                return False
+            for feedback in feedbacks:
+                db.delete(feedback)
+            db.commit()
+            return True
+
+
+Feedbacks = FeedbackTable()
@@ -5,7 +5,7 @@ from typing import Optional
 from open_webui.apps.webui.internal.db import Base, JSONField, get_db
 from open_webui.env import SRC_LOG_LEVELS
 from pydantic import BaseModel, ConfigDict
-from sqlalchemy import BigInteger, Column, String, Text
+from sqlalchemy import BigInteger, Column, String, Text, JSON

 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["MODELS"])
@@ -17,22 +17,35 @@ log.setLevel(SRC_LOG_LEVELS["MODELS"])

 class File(Base):
    __tablename__ = "file"
-
    id = Column(String, primary_key=True)
    user_id = Column(String)
+    hash = Column(Text, nullable=True)
+
    filename = Column(Text)
-    meta = Column(JSONField)
+    path = Column(Text, nullable=True)
+
+    data = Column(JSON, nullable=True)
+    meta = Column(JSON, nullable=True)
+
    created_at = Column(BigInteger)
+    updated_at = Column(BigInteger)


 class FileModel(BaseModel):
+    model_config = ConfigDict(from_attributes=True)
+
    id: str
    user_id: str
-    filename: str
-    meta: dict
-    created_at: int  # timestamp in epoch
+    hash: Optional[str] = None

-    model_config = ConfigDict(from_attributes=True)
+    filename: str
+    path: Optional[str] = None
+
+    data: Optional[dict] = None
+    meta: Optional[dict] = None
+
+    created_at: Optional[int]  # timestamp in epoch
+    updated_at: Optional[int]  # timestamp in epoch


 ####################
@@ -40,17 +53,42 @@ class FileModel(BaseModel):
 ####################


+class FileMeta(BaseModel):
+    name: Optional[str] = None
+    content_type: Optional[str] = None
+    size: Optional[int] = None
+
+    model_config = ConfigDict(extra="allow")
+
+
 class FileModelResponse(BaseModel):
    id: str
    user_id: str
+    hash: Optional[str] = None
+
    filename: str
+    data: Optional[dict] = None
+    meta: FileMeta
+
+    created_at: int  # timestamp in epoch
+    updated_at: int  # timestamp in epoch
+
+    model_config = ConfigDict(extra="allow")
+
+
+class FileMetadataResponse(BaseModel):
+    id: str
    meta: dict
    created_at: int  # timestamp in epoch
+    updated_at: int  # timestamp in epoch


 class FileForm(BaseModel):
    id: str
+    hash: Optional[str] = None
    filename: str
+    path: str
+    data: dict = {}
    meta: dict = {}


@@ -62,6 +100,7 @@ class FilesTable:
                    **form_data.model_dump(),
                    "user_id": user_id,
                    "created_at": int(time.time()),
+                    "updated_at": int(time.time()),
                }
            )

@@ -86,10 +125,48 @@ class FilesTable:
            except Exception:
                return None

+    def get_file_metadata_by_id(self, id: str) -> Optional[FileMetadataResponse]:
+        with get_db() as db:
+            try:
+                file = db.get(File, id)
+                return FileMetadataResponse(
+                    id=file.id,
+                    meta=file.meta,
+                    created_at=file.created_at,
+                    updated_at=file.updated_at,
+                )
+            except Exception:
+                return None
+
    def get_files(self) -> list[FileModel]:
        with get_db() as db:
            return [FileModel.model_validate(file) for file in db.query(File).all()]

+    def get_files_by_ids(self, ids: list[str]) -> list[FileModel]:
+        with get_db() as db:
+            return [
+                FileModel.model_validate(file)
+                for file in db.query(File)
+                .filter(File.id.in_(ids))
+                .order_by(File.updated_at.desc())
+                .all()
+            ]
+
+    def get_file_metadatas_by_ids(self, ids: list[str]) -> list[FileMetadataResponse]:
+        with get_db() as db:
+            return [
+                FileMetadataResponse(
+                    id=file.id,
+                    meta=file.meta,
+                    created_at=file.created_at,
+                    updated_at=file.updated_at,
+                )
+                for file in db.query(File)
+                .filter(File.id.in_(ids))
+                .order_by(File.updated_at.desc())
+                .all()
+            ]
+
    def get_files_by_user_id(self, user_id: str) -> list[FileModel]:
        with get_db() as db:
            return [
@@ -97,6 +174,38 @@ class FilesTable:
                for file in db.query(File).filter_by(user_id=user_id).all()
            ]

+    def update_file_hash_by_id(self, id: str, hash: str) -> Optional[FileModel]:
+        with get_db() as db:
+            try:
+                file = db.query(File).filter_by(id=id).first()
+                file.hash = hash
+                db.commit()
+
+                return FileModel.model_validate(file)
+            except Exception:
+                return None
+
+    def update_file_data_by_id(self, id: str, data: dict) -> Optional[FileModel]:
+        with get_db() as db:
+            try:
+                file = db.query(File).filter_by(id=id).first()
+                file.data = {**(file.data if file.data else {}), **data}
+                db.commit()
+                return FileModel.model_validate(file)
+            except Exception as e:
+
+                return None
+
+    def update_file_metadata_by_id(self, id: str, meta: dict) -> Optional[FileModel]:
+        with get_db() as db:
+            try:
+                file = db.query(File).filter_by(id=id).first()
+                file.meta = {**(file.meta if file.meta else {}), **meta}
+                db.commit()
+                return FileModel.model_validate(file)
+            except Exception:
+                return None
+
    def delete_file_by_id(self, id: str) -> bool:
        with get_db() as db:
            try:
@@ -0,0 +1,271 @@
+import logging
+import time
+import uuid
+from typing import Optional
+
+from open_webui.apps.webui.internal.db import Base, get_db
+from open_webui.apps.webui.models.chats import Chats
+
+from open_webui.env import SRC_LOG_LEVELS
+from pydantic import BaseModel, ConfigDict
+from sqlalchemy import BigInteger, Column, Text, JSON, Boolean
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["MODELS"])
+
+
+####################
+# Folder DB Schema
+####################
+
+
+class Folder(Base):
+    __tablename__ = "folder"
+    id = Column(Text, primary_key=True)
+    parent_id = Column(Text, nullable=True)
+    user_id = Column(Text)
+    name = Column(Text)
+    items = Column(JSON, nullable=True)
+    meta = Column(JSON, nullable=True)
+    is_expanded = Column(Boolean, default=False)
+    created_at = Column(BigInteger)
+    updated_at = Column(BigInteger)
+
+
+class FolderModel(BaseModel):
+    id: str
+    parent_id: Optional[str] = None
+    user_id: str
+    name: str
+    items: Optional[dict] = None
+    meta: Optional[dict] = None
+    is_expanded: bool = False
+    created_at: int
+    updated_at: int
+
+    model_config = ConfigDict(from_attributes=True)
+
+
+####################
+# Forms
+####################
+
+
+class FolderForm(BaseModel):
+    name: str
+    model_config = ConfigDict(extra="allow")
+
+
+class FolderTable:
+    def insert_new_folder(
+        self, user_id: str, name: str, parent_id: Optional[str] = None
+    ) -> Optional[FolderModel]:
+        with get_db() as db:
+            id = str(uuid.uuid4())
+            folder = FolderModel(
+                **{
+                    "id": id,
+                    "user_id": user_id,
+                    "name": name,
+                    "parent_id": parent_id,
+                    "created_at": int(time.time()),
+                    "updated_at": int(time.time()),
+                }
+            )
+            try:
+                result = Folder(**folder.model_dump())
+                db.add(result)
+                db.commit()
+                db.refresh(result)
+                if result:
+                    return FolderModel.model_validate(result)
+                else:
+                    return None
+            except Exception as e:
+                print(e)
+                return None
+
+    def get_folder_by_id_and_user_id(
+        self, id: str, user_id: str
+    ) -> Optional[FolderModel]:
+        try:
+            with get_db() as db:
+                folder = db.query(Folder).filter_by(id=id, user_id=user_id).first()
+
+                if not folder:
+                    return None
+
+                return FolderModel.model_validate(folder)
+        except Exception:
+            return None
+
+    def get_children_folders_by_id_and_user_id(
+        self, id: str, user_id: str
+    ) -> Optional[FolderModel]:
+        try:
+            with get_db() as db:
+                folders = []
+
+                def get_children(folder):
+                    children = self.get_folders_by_parent_id_and_user_id(
+                        folder.id, user_id
+                    )
+                    for child in children:
+                        get_children(child)
+                        folders.append(child)
+
+                folder = db.query(Folder).filter_by(id=id, user_id=user_id).first()
+                if not folder:
+                    return None
+
+                get_children(folder)
+                return folders
+        except Exception:
+            return None
+
+    def get_folders_by_user_id(self, user_id: str) -> list[FolderModel]:
+        with get_db() as db:
+            return [
+                FolderModel.model_validate(folder)
+                for folder in db.query(Folder).filter_by(user_id=user_id).all()
+            ]
+
+    def get_folder_by_parent_id_and_user_id_and_name(
+        self, parent_id: Optional[str], user_id: str, name: str
+    ) -> Optional[FolderModel]:
+        try:
+            with get_db() as db:
+                # Check if folder exists
+                folder = (
+                    db.query(Folder)
+                    .filter_by(parent_id=parent_id, user_id=user_id)
+                    .filter(Folder.name.ilike(name))
+                    .first()
+                )
+
+                if not folder:
+                    return None
+
+                return FolderModel.model_validate(folder)
+        except Exception as e:
+            log.error(f"get_folder_by_parent_id_and_user_id_and_name: {e}")
+            return None
+
+    def get_folders_by_parent_id_and_user_id(
+        self, parent_id: Optional[str], user_id: str
+    ) -> list[FolderModel]:
+        with get_db() as db:
+            return [
+                FolderModel.model_validate(folder)
+                for folder in db.query(Folder)
+                .filter_by(parent_id=parent_id, user_id=user_id)
+                .all()
+            ]
+
+    def update_folder_parent_id_by_id_and_user_id(
+        self,
+        id: str,
+        user_id: str,
+        parent_id: str,
+    ) -> Optional[FolderModel]:
+        try:
+            with get_db() as db:
+                folder = db.query(Folder).filter_by(id=id, user_id=user_id).first()
+
+                if not folder:
+                    return None
+
+                folder.parent_id = parent_id
+                folder.updated_at = int(time.time())
+
+                db.commit()
+
+                return FolderModel.model_validate(folder)
+        except Exception as e:
+            log.error(f"update_folder: {e}")
+            return
+
+    def update_folder_name_by_id_and_user_id(
+        self, id: str, user_id: str, name: str
+    ) -> Optional[FolderModel]:
+        try:
+            with get_db() as db:
+                folder = db.query(Folder).filter_by(id=id, user_id=user_id).first()
+
+                if not folder:
+                    return None
+
+                existing_folder = (
+                    db.query(Folder)
+                    .filter_by(name=name, parent_id=folder.parent_id, user_id=user_id)
+                    .first()
+                )
+
+                if existing_folder:
+                    return None
+
+                folder.name = name
+                folder.updated_at = int(time.time())
+
+                db.commit()
+
+                return FolderModel.model_validate(folder)
+        except Exception as e:
+            log.error(f"update_folder: {e}")
+            return
+
+    def update_folder_is_expanded_by_id_and_user_id(
+        self, id: str, user_id: str, is_expanded: bool
+    ) -> Optional[FolderModel]:
+        try:
+            with get_db() as db:
+                folder = db.query(Folder).filter_by(id=id, user_id=user_id).first()
+
+                if not folder:
+                    return None
+
+                folder.is_expanded = is_expanded
+                folder.updated_at = int(time.time())
+
+                db.commit()
+
+                return FolderModel.model_validate(folder)
+        except Exception as e:
+            log.error(f"update_folder: {e}")
+            return
+
+    def delete_folder_by_id_and_user_id(self, id: str, user_id: str) -> bool:
+        try:
+            with get_db() as db:
+                folder = db.query(Folder).filter_by(id=id, user_id=user_id).first()
+                if not folder:
+                    return False
+
+                # Delete all chats in the folder
+                Chats.delete_chats_by_user_id_and_folder_id(user_id, folder.id)
+
+                # Delete all children folders
+                def delete_children(folder):
+                    folder_children = self.get_folders_by_parent_id_and_user_id(
+                        folder.id, user_id
+                    )
+                    for folder_child in folder_children:
+                        Chats.delete_chats_by_user_id_and_folder_id(
+                            user_id, folder_child.id
+                        )
+                        delete_children(folder_child)
+
+                        folder = db.query(Folder).filter_by(id=folder_child.id).first()
+                        db.delete(folder)
+                        db.commit()
+
+                delete_children(folder)
+                db.delete(folder)
+                db.commit()
+                return True
+        except Exception as e:
+            log.error(f"delete_folder: {e}")
+            return False
+
+
+Folders = FolderTable()
@@ -0,0 +1,168 @@
+import json
+import logging
+import time
+from typing import Optional
+import uuid
+
+from open_webui.apps.webui.internal.db import Base, get_db
+from open_webui.env import SRC_LOG_LEVELS
+
+from open_webui.apps.webui.models.files import FileMetadataResponse
+
+
+from pydantic import BaseModel, ConfigDict
+from sqlalchemy import BigInteger, Column, String, Text, JSON
+
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["MODELS"])
+
+####################
+# Knowledge DB Schema
+####################
+
+
+class Knowledge(Base):
+    __tablename__ = "knowledge"
+
+    id = Column(Text, unique=True, primary_key=True)
+    user_id = Column(Text)
+
+    name = Column(Text)
+    description = Column(Text)
+
+    data = Column(JSON, nullable=True)
+    meta = Column(JSON, nullable=True)
+
+    created_at = Column(BigInteger)
+    updated_at = Column(BigInteger)
+
+
+class KnowledgeModel(BaseModel):
+    model_config = ConfigDict(from_attributes=True)
+
+    id: str
+    user_id: str
+
+    name: str
+    description: str
+
+    data: Optional[dict] = None
+    meta: Optional[dict] = None
+
+    created_at: int  # timestamp in epoch
+    updated_at: int  # timestamp in epoch
+
+
+####################
+# Forms
+####################
+
+
+class KnowledgeResponse(BaseModel):
+    id: str
+    name: str
+    description: str
+    data: Optional[dict] = None
+    meta: Optional[dict] = None
+    created_at: int  # timestamp in epoch
+    updated_at: int  # timestamp in epoch
+
+    files: Optional[list[FileMetadataResponse | dict]] = None
+
+
+class KnowledgeForm(BaseModel):
+    name: str
+    description: str
+    data: Optional[dict] = None
+
+
+class KnowledgeUpdateForm(BaseModel):
+    name: Optional[str] = None
+    description: Optional[str] = None
+    data: Optional[dict] = None
+
+
+class KnowledgeTable:
+    def insert_new_knowledge(
+        self, user_id: str, form_data: KnowledgeForm
+    ) -> Optional[KnowledgeModel]:
+        with get_db() as db:
+            knowledge = KnowledgeModel(
+                **{
+                    **form_data.model_dump(),
+                    "id": str(uuid.uuid4()),
+                    "user_id": user_id,
+                    "created_at": int(time.time()),
+                    "updated_at": int(time.time()),
+                }
+            )
+
+            try:
+                result = Knowledge(**knowledge.model_dump())
+                db.add(result)
+                db.commit()
+                db.refresh(result)
+                if result:
+                    return KnowledgeModel.model_validate(result)
+                else:
+                    return None
+            except Exception:
+                return None
+
+    def get_knowledge_items(self) -> list[KnowledgeModel]:
+        with get_db() as db:
+            return [
+                KnowledgeModel.model_validate(knowledge)
+                for knowledge in db.query(Knowledge)
+                .order_by(Knowledge.updated_at.desc())
+                .all()
+            ]
+
+    def get_knowledge_by_id(self, id: str) -> Optional[KnowledgeModel]:
+        try:
+            with get_db() as db:
+                knowledge = db.query(Knowledge).filter_by(id=id).first()
+                return KnowledgeModel.model_validate(knowledge) if knowledge else None
+        except Exception:
+            return None
+
+    def update_knowledge_by_id(
+        self, id: str, form_data: KnowledgeUpdateForm, overwrite: bool = False
+    ) -> Optional[KnowledgeModel]:
+        try:
+            with get_db() as db:
+                knowledge = self.get_knowledge_by_id(id=id)
+                db.query(Knowledge).filter_by(id=id).update(
+                    {
+                        **form_data.model_dump(exclude_none=True),
+                        "updated_at": int(time.time()),
+                    }
+                )
+                db.commit()
+                return self.get_knowledge_by_id(id=id)
+        except Exception as e:
+            log.exception(e)
+            return None
+
+    def delete_knowledge_by_id(self, id: str) -> bool:
+        try:
+            with get_db() as db:
+                db.query(Knowledge).filter_by(id=id).delete()
+                db.commit()
+                return True
+        except Exception:
+            return False
+
+    def delete_all_knowledge(self) -> bool:
+        with get_db() as db:
+            try:
+                db.query(Knowledge).delete()
+                db.commit()
+
+                return True
+            except Exception:
+                return False
+
+
+Knowledges = KnowledgeTable()
@@ -4,53 +4,35 @@ import uuid
 from typing import Optional

 from open_webui.apps.webui.internal.db import Base, get_db
+
+
 from open_webui.env import SRC_LOG_LEVELS
 from pydantic import BaseModel, ConfigDict
-from sqlalchemy import BigInteger, Column, String, Text
+from sqlalchemy import BigInteger, Column, String, JSON, PrimaryKeyConstraint

 log = logging.getLogger(__name__)
 log.setLevel(SRC_LOG_LEVELS["MODELS"])

+
 ####################
 # Tag DB Schema
 ####################
-
-
 class Tag(Base):
    __tablename__ = "tag"
-
-    id = Column(String, primary_key=True)
+    id = Column(String)
    name = Column(String)
    user_id = Column(String)
-    data = Column(Text, nullable=True)
+    meta = Column(JSON, nullable=True)

-
-class ChatIdTag(Base):
-    __tablename__ = "chatidtag"
-
-    id = Column(String, primary_key=True)
-    tag_name = Column(String)
-    chat_id = Column(String)
-    user_id = Column(String)
-    timestamp = Column(BigInteger)
+    # Unique constraint ensuring (id, user_id) is unique, not just the `id` column
+    __table_args__ = (PrimaryKeyConstraint("id", "user_id", name="pk_id_user_id"),)


 class TagModel(BaseModel):
    id: str
    name: str
    user_id: str
-    data: Optional[str] = None
-
-    model_config = ConfigDict(from_attributes=True)
-
-
-class ChatIdTagModel(BaseModel):
-    id: str
-    tag_name: str
-    chat_id: str
-    user_id: str
-    timestamp: int
-
+    meta: Optional[dict] = None
    model_config = ConfigDict(from_attributes=True)


@@ -59,23 +41,15 @@ class ChatIdTagModel(BaseModel):
 ####################


-class ChatIdTagForm(BaseModel):
-    tag_name: str
+class TagChatIdForm(BaseModel):
+    name: str
    chat_id: str


-class TagChatIdsResponse(BaseModel):
-    chat_ids: list[str]
-
-
-class ChatTagsResponse(BaseModel):
-    tags: list[str]
-
-
 class TagTable:
    def insert_new_tag(self, name: str, user_id: str) -> Optional[TagModel]:
        with get_db() as db:
-            id = str(uuid.uuid4())
+            id = name.replace(" ", "_").lower()
            tag = TagModel(**{"id": id, "user_id": user_id, "name": name})
            try:
                result = Tag(**tag.model_dump())
@@ -86,177 +60,50 @@ class TagTable:
                    return TagModel.model_validate(result)
                else:
                    return None
-            except Exception:
+            except Exception as e:
+                print(e)
                return None

    def get_tag_by_name_and_user_id(
        self, name: str, user_id: str
    ) -> Optional[TagModel]:
        try:
+            id = name.replace(" ", "_").lower()
            with get_db() as db:
-                tag = db.query(Tag).filter_by(name=name, user_id=user_id).first()
+                tag = db.query(Tag).filter_by(id=id, user_id=user_id).first()
                return TagModel.model_validate(tag)
        except Exception:
            return None

-    def add_tag_to_chat(
-        self, user_id: str, form_data: ChatIdTagForm
-    ) -> Optional[ChatIdTagModel]:
-        tag = self.get_tag_by_name_and_user_id(form_data.tag_name, user_id)
-        if tag is None:
-            tag = self.insert_new_tag(form_data.tag_name, user_id)
-
-        id = str(uuid.uuid4())
-        chatIdTag = ChatIdTagModel(
-            **{
-                "id": id,
-                "user_id": user_id,
-                "chat_id": form_data.chat_id,
-                "tag_name": tag.name,
-                "timestamp": int(time.time()),
-            }
-        )
-        try:
-            with get_db() as db:
-                result = ChatIdTag(**chatIdTag.model_dump())
-                db.add(result)
-                db.commit()
-                db.refresh(result)
-                if result:
-                    return ChatIdTagModel.model_validate(result)
-                else:
-                    return None
-        except Exception:
-            return None
-
    def get_tags_by_user_id(self, user_id: str) -> list[TagModel]:
        with get_db() as db:
-            tag_names = [
-                chat_id_tag.tag_name
-                for chat_id_tag in (
-                    db.query(ChatIdTag)
-                    .filter_by(user_id=user_id)
-                    .order_by(ChatIdTag.timestamp.desc())
-                    .all()
-                )
-            ]
-
            return [
                TagModel.model_validate(tag)
-                for tag in (
-                    db.query(Tag)
-                    .filter_by(user_id=user_id)
-                    .filter(Tag.name.in_(tag_names))
-                    .all()
-                )
+                for tag in (db.query(Tag).filter_by(user_id=user_id).all())
            ]

-    def get_tags_by_chat_id_and_user_id(
-        self, chat_id: str, user_id: str
+    def get_tags_by_ids_and_user_id(
+        self, ids: list[str], user_id: str
    ) -> list[TagModel]:
        with get_db() as db:
-            tag_names = [
-                chat_id_tag.tag_name
-                for chat_id_tag in (
-                    db.query(ChatIdTag)
-                    .filter_by(user_id=user_id, chat_id=chat_id)
-                    .order_by(ChatIdTag.timestamp.desc())
-                    .all()
-                )
-            ]
-
            return [
                TagModel.model_validate(tag)
                for tag in (
-                    db.query(Tag)
-                    .filter_by(user_id=user_id)
-                    .filter(Tag.name.in_(tag_names))
-                    .all()
+                    db.query(Tag).filter(Tag.id.in_(ids), Tag.user_id == user_id).all()
                )
            ]

-    def get_chat_ids_by_tag_name_and_user_id(
-        self, tag_name: str, user_id: str
-    ) -> list[ChatIdTagModel]:
-        with get_db() as db:
-            return [
-                ChatIdTagModel.model_validate(chat_id_tag)
-                for chat_id_tag in (
-                    db.query(ChatIdTag)
-                    .filter_by(user_id=user_id, tag_name=tag_name)
-                    .order_by(ChatIdTag.timestamp.desc())
-                    .all()
-                )
-            ]
-
-    def count_chat_ids_by_tag_name_and_user_id(
-        self, tag_name: str, user_id: str
-    ) -> int:
-        with get_db() as db:
-            return (
-                db.query(ChatIdTag)
-                .filter_by(tag_name=tag_name, user_id=user_id)
-                .count()
-            )
-
-    def delete_tag_by_tag_name_and_user_id(self, tag_name: str, user_id: str) -> bool:
+    def delete_tag_by_name_and_user_id(self, name: str, user_id: str) -> bool:
        try:
            with get_db() as db:
-                res = (
-                    db.query(ChatIdTag)
-                    .filter_by(tag_name=tag_name, user_id=user_id)
-                    .delete()
-                )
+                id = name.replace(" ", "_").lower()
+                res = db.query(Tag).filter_by(id=id, user_id=user_id).delete()
                log.debug(f"res: {res}")
                db.commit()
-
-                tag_count = self.count_chat_ids_by_tag_name_and_user_id(
-                    tag_name, user_id
-                )
-                if tag_count == 0:
-                    # Remove tag item from Tag col as well
-                    db.query(Tag).filter_by(name=tag_name, user_id=user_id).delete()
-                    db.commit()
                return True
        except Exception as e:
            log.error(f"delete_tag: {e}")
            return False

-    def delete_tag_by_tag_name_and_chat_id_and_user_id(
-        self, tag_name: str, chat_id: str, user_id: str
-    ) -> bool:
-        try:
-            with get_db() as db:
-                res = (
-                    db.query(ChatIdTag)
-                    .filter_by(tag_name=tag_name, chat_id=chat_id, user_id=user_id)
-                    .delete()
-                )
-                log.debug(f"res: {res}")
-                db.commit()
-
-                tag_count = self.count_chat_ids_by_tag_name_and_user_id(
-                    tag_name, user_id
-                )
-                if tag_count == 0:
-                    # Remove tag item from Tag col as well
-                    db.query(Tag).filter_by(name=tag_name, user_id=user_id).delete()
-                    db.commit()
-
-                return True
-        except Exception as e:
-            log.error(f"delete_tag: {e}")
-            return False
-
-    def delete_tags_by_chat_id_and_user_id(self, chat_id: str, user_id: str) -> bool:
-        tags = self.get_tags_by_chat_id_and_user_id(chat_id, user_id)
-
-        for tag in tags:
-            self.delete_tag_by_tag_name_and_chat_id_and_user_id(
-                tag.tag_name, chat_id, user_id
-            )
-
-        return True
-

 Tags = TagTable()
@@ -1,10 +1,13 @@
 import re
 import uuid
+import time
+import datetime

 from open_webui.apps.webui.models.auths import (
    AddUserForm,
    ApiKey,
    Auths,
+    Token,
    SigninForm,
    SigninResponse,
    SignupForm,
@@ -18,6 +21,8 @@ from open_webui.constants import ERROR_MESSAGES, WEBHOOK_MESSAGES
 from open_webui.env import (
    WEBUI_AUTH_TRUSTED_EMAIL_HEADER,
    WEBUI_AUTH_TRUSTED_NAME_HEADER,
+    WEBUI_SESSION_COOKIE_SAME_SITE,
+    WEBUI_SESSION_COOKIE_SECURE,
 )
 from fastapi import APIRouter, Depends, HTTPException, Request, status
 from fastapi.responses import Response
@@ -27,10 +32,12 @@ from open_webui.utils.utils import (
    create_api_key,
    create_token,
    get_admin_user,
+    get_verified_user,
    get_current_user,
    get_password_hash,
 )
 from open_webui.utils.webhook import post_webhook
+from typing import Optional

 router = APIRouter()

@@ -39,23 +46,44 @@ router = APIRouter()
 ############################


-@router.get("/", response_model=UserResponse)
+class SessionUserResponse(Token, UserResponse):
+    expires_at: Optional[int] = None
+
+
+@router.get("/", response_model=SessionUserResponse)
 async def get_session_user(
    request: Request, response: Response, user=Depends(get_current_user)
 ):
+    expires_delta = parse_duration(request.app.state.config.JWT_EXPIRES_IN)
+    expires_at = None
+    if expires_delta:
+        expires_at = int(time.time()) + int(expires_delta.total_seconds())
+
    token = create_token(
        data={"id": user.id},
-        expires_delta=parse_duration(request.app.state.config.JWT_EXPIRES_IN),
+        expires_delta=expires_delta,
+    )
+
+    datetime_expires_at = (
+        datetime.datetime.fromtimestamp(expires_at, datetime.timezone.utc)
+        if expires_at
+        else None
    )

    # Set the cookie token
    response.set_cookie(
        key="token",
        value=token,
+        expires=datetime_expires_at,
        httponly=True,  # Ensures the cookie is not accessible via JavaScript
+        samesite=WEBUI_SESSION_COOKIE_SAME_SITE,
+        secure=WEBUI_SESSION_COOKIE_SECURE,
    )

    return {
+        "token": token,
+        "token_type": "Bearer",
+        "expires_at": expires_at,
        "id": user.id,
        "email": user.email,
        "name": user.name,
@@ -71,7 +99,7 @@ async def get_session_user(

@router.post("/update/profile", response_model=UserResponse)
 async def update_profile(
-    form_data: UpdateProfileForm, session_user=Depends(get_current_user)
+    form_data: UpdateProfileForm, session_user=Depends(get_verified_user)
 ):
    if session_user:
        user = Users.update_user_by_id(
@@ -114,7 +142,7 @@ async def update_password(
 ############################


-@router.post("/signin", response_model=SigninResponse)
+@router.post("/signin", response_model=SessionUserResponse)
 async def signin(request: Request, response: Response, form_data: SigninForm):
    if WEBUI_AUTH_TRUSTED_EMAIL_HEADER:
        if WEBUI_AUTH_TRUSTED_EMAIL_HEADER not in request.headers:
@@ -156,21 +184,37 @@ async def signin(request: Request, response: Response, form_data: SigninForm):
        user = Auths.authenticate_user(form_data.email.lower(), form_data.password)

    if user:
+
+        expires_delta = parse_duration(request.app.state.config.JWT_EXPIRES_IN)
+        expires_at = None
+        if expires_delta:
+            expires_at = int(time.time()) + int(expires_delta.total_seconds())
+
        token = create_token(
            data={"id": user.id},
-            expires_delta=parse_duration(request.app.state.config.JWT_EXPIRES_IN),
+            expires_delta=expires_delta,
+        )
+
+        datetime_expires_at = (
+            datetime.datetime.fromtimestamp(expires_at, datetime.timezone.utc)
+            if expires_at
+            else None
        )

        # Set the cookie token
        response.set_cookie(
            key="token",
            value=token,
+            expires=datetime_expires_at,
            httponly=True,  # Ensures the cookie is not accessible via JavaScript
+            samesite=WEBUI_SESSION_COOKIE_SAME_SITE,
+            secure=WEBUI_SESSION_COOKIE_SECURE,
        )

        return {
            "token": token,
            "token_type": "Bearer",
+            "expires_at": expires_at,
            "id": user.id,
            "email": user.email,
            "name": user.name,
@@ -186,16 +230,21 @@ async def signin(request: Request, response: Response, form_data: SigninForm):
 ############################


-@router.post("/signup", response_model=SigninResponse)
+@router.post("/signup", response_model=SessionUserResponse)
 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
-    ):
-        raise HTTPException(
-            status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.ACCESS_PROHIBITED
-        )
+    if WEBUI_AUTH:
+        if (
+            not request.app.state.config.ENABLE_SIGNUP
+            or not request.app.state.config.ENABLE_LOGIN_FORM
+        ):
+            raise HTTPException(
+                status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.ACCESS_PROHIBITED
+            )
+    else:
+        if Users.get_num_users() != 0:
+            raise HTTPException(
+                status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.ACCESS_PROHIBITED
+            )

    if not validate_email_format(form_data.email.lower()):
        raise HTTPException(
@@ -221,16 +270,30 @@ async def signup(request: Request, response: Response, form_data: SignupForm):
        )

        if user:
+            expires_delta = parse_duration(request.app.state.config.JWT_EXPIRES_IN)
+            expires_at = None
+            if expires_delta:
+                expires_at = int(time.time()) + int(expires_delta.total_seconds())
+
            token = create_token(
                data={"id": user.id},
-                expires_delta=parse_duration(request.app.state.config.JWT_EXPIRES_IN),
+                expires_delta=expires_delta,
+            )
+
+            datetime_expires_at = (
+                datetime.datetime.fromtimestamp(expires_at, datetime.timezone.utc)
+                if expires_at
+                else None
            )

            # Set the cookie token
            response.set_cookie(
                key="token",
                value=token,
+                expires=datetime_expires_at,
                httponly=True,  # Ensures the cookie is not accessible via JavaScript
+                samesite=WEBUI_SESSION_COOKIE_SAME_SITE,
+                secure=WEBUI_SESSION_COOKIE_SECURE,
            )

            if request.app.state.config.WEBHOOK_URL:
@@ -247,6 +310,7 @@ async def signup(request: Request, response: Response, form_data: SignupForm):
            return {
                "token": token,
                "token_type": "Bearer",
+                "expires_at": expires_at,
                "id": user.id,
                "email": user.email,
                "name": user.name,
@@ -259,6 +323,12 @@ async def signup(request: Request, response: Response, form_data: SignupForm):
        raise HTTPException(500, detail=ERROR_MESSAGES.DEFAULT(err))


+@router.get("/signout")
+async def signout(response: Response):
+    response.delete_cookie("token")
+    return {"status": True}
+
+
 ############################
 # AddUser
 ############################
@@ -4,16 +4,14 @@ from typing import Optional

 from open_webui.apps.webui.models.chats import (
    ChatForm,
+    ChatImportForm,
    ChatResponse,
    Chats,
    ChatTitleIdResponse,
 )
-from open_webui.apps.webui.models.tags import (
-    ChatIdTagForm,
-    ChatIdTagModel,
-    TagModel,
-    Tags,
-)
+from open_webui.apps.webui.models.tags import TagModel, Tags
+from open_webui.apps.webui.models.folders import Folders
+
 from open_webui.config import ENABLE_ADMIN_CHAT_ACCESS, ENABLE_ADMIN_EXPORT
 from open_webui.constants import ERROR_MESSAGES
 from open_webui.env import SRC_LOG_LEVELS
@@ -52,10 +50,9 @@ async def get_session_user_chat_list(

@router.delete("/", response_model=bool)
 async def delete_all_user_chats(request: Request, user=Depends(get_verified_user)):
-    if (
-        user.role == "user"
-        and not request.app.state.config.USER_PERMISSIONS["chat"]["deletion"]
-    ):
+    if user.role == "user" and not request.app.state.config.USER_PERMISSIONS.get(
+        "chat", {}
+    ).get("deletion", {}):
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED,
            detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
@@ -96,7 +93,35 @@ async def get_user_chat_list_by_user_id(
 async def create_new_chat(form_data: ChatForm, user=Depends(get_verified_user)):
    try:
        chat = Chats.insert_new_chat(user.id, form_data)
-        return ChatResponse(**{**chat.model_dump(), "chat": json.loads(chat.chat)})
+        return ChatResponse(**chat.model_dump())
+    except Exception as e:
+        log.exception(e)
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST, detail=ERROR_MESSAGES.DEFAULT()
+        )
+
+
+############################
+# ImportChat
+############################
+
+
+@router.post("/import", response_model=Optional[ChatResponse])
+async def import_chat(form_data: ChatImportForm, user=Depends(get_verified_user)):
+    try:
+        chat = Chats.import_chat(user.id, form_data)
+        if chat:
+            tags = chat.meta.get("tags", [])
+            for tag_id in tags:
+                tag_id = tag_id.replace(" ", "_").lower()
+                tag_name = " ".join([word.capitalize() for word in tag_id.split("_")])
+                if (
+                    tag_id != "none"
+                    and Tags.get_tag_by_name_and_user_id(tag_name, user.id) is None
+                ):
+                    Tags.insert_new_tag(tag_name, user.id)
+
+        return ChatResponse(**chat.model_dump())
    except Exception as e:
        log.exception(e)
        raise HTTPException(
@@ -109,10 +134,77 @@ async def create_new_chat(form_data: ChatForm, user=Depends(get_verified_user)):
 ############################


+@router.get("/search", response_model=list[ChatTitleIdResponse])
+async def search_user_chats(
+    text: str, page: Optional[int] = None, user=Depends(get_verified_user)
+):
+    if page is None:
+        page = 1
+
+    limit = 60
+    skip = (page - 1) * limit
+
+    chat_list = [
+        ChatTitleIdResponse(**chat.model_dump())
+        for chat in Chats.get_chats_by_user_id_and_search_text(
+            user.id, text, skip=skip, limit=limit
+        )
+    ]
+
+    # Delete tag if no chat is found
+    words = text.strip().split(" ")
+    if page == 1 and len(words) == 1 and words[0].startswith("tag:"):
+        tag_id = words[0].replace("tag:", "")
+        if len(chat_list) == 0:
+            if Tags.get_tag_by_name_and_user_id(tag_id, user.id):
+                log.debug(f"deleting tag: {tag_id}")
+                Tags.delete_tag_by_name_and_user_id(tag_id, user.id)
+
+    return chat_list
+
+
+############################
+# GetChatsByFolderId
+############################
+
+
+@router.get("/folder/{folder_id}", response_model=list[ChatResponse])
+async def get_chats_by_folder_id(folder_id: str, user=Depends(get_verified_user)):
+    folder_ids = [folder_id]
+    children_folders = Folders.get_children_folders_by_id_and_user_id(
+        folder_id, user.id
+    )
+    if children_folders:
+        folder_ids.extend([folder.id for folder in children_folders])
+
+    return [
+        ChatResponse(**chat.model_dump())
+        for chat in Chats.get_chats_by_folder_ids_and_user_id(folder_ids, user.id)
+    ]
+
+
+############################
+# GetPinnedChats
+############################
+
+
+@router.get("/pinned", response_model=list[ChatResponse])
+async def get_user_pinned_chats(user=Depends(get_verified_user)):
+    return [
+        ChatResponse(**chat.model_dump())
+        for chat in Chats.get_pinned_chats_by_user_id(user.id)
+    ]
+
+
+############################
+# GetChats
+############################
+
+
@router.get("/all", response_model=list[ChatResponse])
 async def get_user_chats(user=Depends(get_verified_user)):
    return [
-        ChatResponse(**{**chat.model_dump(), "chat": json.loads(chat.chat)})
+        ChatResponse(**chat.model_dump())
        for chat in Chats.get_chats_by_user_id(user.id)
    ]

@@ -125,11 +217,28 @@ async def get_user_chats(user=Depends(get_verified_user)):
@router.get("/all/archived", response_model=list[ChatResponse])
 async def get_user_archived_chats(user=Depends(get_verified_user)):
    return [
-        ChatResponse(**{**chat.model_dump(), "chat": json.loads(chat.chat)})
+        ChatResponse(**chat.model_dump())
        for chat in Chats.get_archived_chats_by_user_id(user.id)
    ]


+############################
+# GetAllTags
+############################
+
+
+@router.get("/all/tags", response_model=list[TagModel])
+async def get_all_user_tags(user=Depends(get_verified_user)):
+    try:
+        tags = Tags.get_tags_by_user_id(user.id)
+        return tags
+    except Exception as e:
+        log.exception(e)
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST, detail=ERROR_MESSAGES.DEFAULT()
+        )
+
+
 ############################
 # GetAllChatsInDB
 ############################
@@ -142,10 +251,7 @@ async def get_all_user_chats_in_db(user=Depends(get_admin_user)):
            status_code=status.HTTP_401_UNAUTHORIZED,
            detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
        )
-    return [
-        ChatResponse(**{**chat.model_dump(), "chat": json.loads(chat.chat)})
-        for chat in Chats.get_chats()
-    ]
+    return [ChatResponse(**chat.model_dump()) for chat in Chats.get_chats()]


 ############################
@@ -188,7 +294,8 @@ async def get_shared_chat_by_id(share_id: str, user=Depends(get_verified_user)):
        chat = Chats.get_chat_by_id(share_id)

    if chat:
-        return ChatResponse(**{**chat.model_dump(), "chat": json.loads(chat.chat)})
+        return ChatResponse(**chat.model_dump())
+
    else:
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED, detail=ERROR_MESSAGES.NOT_FOUND
@@ -200,48 +307,28 @@ async def get_shared_chat_by_id(share_id: str, user=Depends(get_verified_user)):
 ############################


-class TagNameForm(BaseModel):
+class TagForm(BaseModel):
    name: str
+
+
+class TagFilterForm(TagForm):
    skip: Optional[int] = 0
    limit: Optional[int] = 50


@router.post("/tags", response_model=list[ChatTitleIdResponse])
 async def get_user_chat_list_by_tag_name(
-    form_data: TagNameForm, user=Depends(get_verified_user)
+    form_data: TagFilterForm, user=Depends(get_verified_user)
 ):
-    chat_ids = [
-        chat_id_tag.chat_id
-        for chat_id_tag in Tags.get_chat_ids_by_tag_name_and_user_id(
-            form_data.name, user.id
-        )
-    ]
-
-    chats = Chats.get_chat_list_by_chat_ids(chat_ids, form_data.skip, form_data.limit)
-
+    chats = Chats.get_chat_list_by_user_id_and_tag_name(
+        user.id, form_data.name, form_data.skip, form_data.limit
+    )
    if len(chats) == 0:
-        Tags.delete_tag_by_tag_name_and_user_id(form_data.name, user.id)
+        Tags.delete_tag_by_name_and_user_id(form_data.name, user.id)

    return chats


-############################
-# GetAllTags
-############################
-
-
-@router.get("/tags/all", response_model=list[TagModel])
-async def get_all_tags(user=Depends(get_verified_user)):
-    try:
-        tags = Tags.get_tags_by_user_id(user.id)
-        return tags
-    except Exception as e:
-        log.exception(e)
-        raise HTTPException(
-            status_code=status.HTTP_400_BAD_REQUEST, detail=ERROR_MESSAGES.DEFAULT()
-        )
-
-
 ############################
 # GetChatById
 ############################
@@ -252,7 +339,8 @@ async def get_chat_by_id(id: str, user=Depends(get_verified_user)):
    chat = Chats.get_chat_by_id_and_user_id(id, user.id)

    if chat:
-        return ChatResponse(**{**chat.model_dump(), "chat": json.loads(chat.chat)})
+        return ChatResponse(**chat.model_dump())
+
    else:
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED, detail=ERROR_MESSAGES.NOT_FOUND
@@ -270,10 +358,9 @@ async def update_chat_by_id(
 ):
    chat = Chats.get_chat_by_id_and_user_id(id, user.id)
    if chat:
-        updated_chat = {**json.loads(chat.chat), **form_data.chat}
-
+        updated_chat = {**chat.chat, **form_data.chat}
        chat = Chats.update_chat_by_id(id, updated_chat)
-        return ChatResponse(**{**chat.model_dump(), "chat": json.loads(chat.chat)})
+        return ChatResponse(**chat.model_dump())
    else:
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED,
@@ -289,38 +376,83 @@ async def update_chat_by_id(
@router.delete("/{id}", response_model=bool)
 async def delete_chat_by_id(request: Request, id: str, user=Depends(get_verified_user)):
    if user.role == "admin":
+        chat = Chats.get_chat_by_id(id)
+        for tag in chat.meta.get("tags", []):
+            if Chats.count_chats_by_tag_name_and_user_id(tag, user.id) == 1:
+                Tags.delete_tag_by_name_and_user_id(tag, user.id)
+
        result = Chats.delete_chat_by_id(id)
+
        return result
    else:
-        if not request.app.state.config.USER_PERMISSIONS["chat"]["deletion"]:
+        if not request.app.state.config.USER_PERMISSIONS.get("chat", {}).get(
+            "deletion", {}
+        ):
            raise HTTPException(
                status_code=status.HTTP_401_UNAUTHORIZED,
                detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
            )

+        chat = Chats.get_chat_by_id(id)
+        for tag in chat.meta.get("tags", []):
+            if Chats.count_chats_by_tag_name_and_user_id(tag, user.id) == 1:
+                Tags.delete_tag_by_name_and_user_id(tag, user.id)
+
        result = Chats.delete_chat_by_id_and_user_id(id, user.id)
        return result


+############################
+# GetPinnedStatusById
+############################
+
+
+@router.get("/{id}/pinned", response_model=Optional[bool])
+async def get_pinned_status_by_id(id: str, user=Depends(get_verified_user)):
+    chat = Chats.get_chat_by_id_and_user_id(id, user.id)
+    if chat:
+        return chat.pinned
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_401_UNAUTHORIZED, detail=ERROR_MESSAGES.DEFAULT()
+        )
+
+
+############################
+# PinChatById
+############################
+
+
+@router.post("/{id}/pin", response_model=Optional[ChatResponse])
+async def pin_chat_by_id(id: str, user=Depends(get_verified_user)):
+    chat = Chats.get_chat_by_id_and_user_id(id, user.id)
+    if chat:
+        chat = Chats.toggle_chat_pinned_by_id(id)
+        return chat
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_401_UNAUTHORIZED, detail=ERROR_MESSAGES.DEFAULT()
+        )
+
+
 ############################
 # CloneChat
 ############################


-@router.get("/{id}/clone", response_model=Optional[ChatResponse])
+@router.post("/{id}/clone", response_model=Optional[ChatResponse])
 async def clone_chat_by_id(id: str, user=Depends(get_verified_user)):
    chat = Chats.get_chat_by_id_and_user_id(id, user.id)
    if chat:
-        chat_body = json.loads(chat.chat)
        updated_chat = {
-            **chat_body,
+            **chat.chat,
            "originalChatId": chat.id,
-            "branchPointMessageId": chat_body["history"]["currentId"],
+            "branchPointMessageId": chat.chat["history"]["currentId"],
            "title": f"Clone of {chat.title}",
        }

        chat = Chats.insert_new_chat(user.id, ChatForm(**{"chat": updated_chat}))
-        return ChatResponse(**{**chat.model_dump(), "chat": json.loads(chat.chat)})
+        return ChatResponse(**chat.model_dump())
    else:
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED, detail=ERROR_MESSAGES.DEFAULT()
@@ -332,12 +464,26 @@ async def clone_chat_by_id(id: str, user=Depends(get_verified_user)):
 ############################


-@router.get("/{id}/archive", response_model=Optional[ChatResponse])
+@router.post("/{id}/archive", response_model=Optional[ChatResponse])
 async def archive_chat_by_id(id: str, user=Depends(get_verified_user)):
    chat = Chats.get_chat_by_id_and_user_id(id, user.id)
    if chat:
        chat = Chats.toggle_chat_archive_by_id(id)
-        return ChatResponse(**{**chat.model_dump(), "chat": json.loads(chat.chat)})
+
+        # Delete tags if chat is archived
+        if chat.archived:
+            for tag_id in chat.meta.get("tags", []):
+                if Chats.count_chats_by_tag_name_and_user_id(tag_id, user.id) == 0:
+                    log.debug(f"deleting tag: {tag_id}")
+                    Tags.delete_tag_by_name_and_user_id(tag_id, user.id)
+        else:
+            for tag_id in chat.meta.get("tags", []):
+                tag = Tags.get_tag_by_name_and_user_id(tag_id, user.id)
+                if tag is None:
+                    log.debug(f"inserting tag: {tag_id}")
+                    tag = Tags.insert_new_tag(tag_id, user.id)
+
+        return ChatResponse(**chat.model_dump())
    else:
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED, detail=ERROR_MESSAGES.DEFAULT()
@@ -355,9 +501,7 @@ async def share_chat_by_id(id: str, user=Depends(get_verified_user)):
    if chat:
        if chat.share_id:
            shared_chat = Chats.update_shared_chat_by_chat_id(chat.id)
-            return ChatResponse(
-                **{**shared_chat.model_dump(), "chat": json.loads(shared_chat.chat)}
-            )
+            return ChatResponse(**shared_chat.model_dump())

        shared_chat = Chats.insert_shared_chat_by_chat_id(chat.id)
        if not shared_chat:
@@ -365,10 +509,8 @@ async def share_chat_by_id(id: str, user=Depends(get_verified_user)):
                status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
                detail=ERROR_MESSAGES.DEFAULT(),
            )
+        return ChatResponse(**shared_chat.model_dump())

-        return ChatResponse(
-            **{**shared_chat.model_dump(), "chat": json.loads(shared_chat.chat)}
-        )
    else:
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED,
@@ -399,6 +541,31 @@ async def delete_shared_chat_by_id(id: str, user=Depends(get_verified_user)):
        )


+############################
+# UpdateChatFolderIdById
+############################
+
+
+class ChatFolderIdForm(BaseModel):
+    folder_id: Optional[str] = None
+
+
+@router.post("/{id}/folder", response_model=Optional[ChatResponse])
+async def update_chat_folder_id_by_id(
+    id: str, form_data: ChatFolderIdForm, user=Depends(get_verified_user)
+):
+    chat = Chats.get_chat_by_id_and_user_id(id, user.id)
+    if chat:
+        chat = Chats.update_chat_folder_id_by_id_and_user_id(
+            id, user.id, form_data.folder_id
+        )
+        return ChatResponse(**chat.model_dump())
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_401_UNAUTHORIZED, detail=ERROR_MESSAGES.DEFAULT()
+        )
+
+
 ############################
 # GetChatTagsById
 ############################
@@ -406,10 +573,10 @@ async def delete_shared_chat_by_id(id: str, user=Depends(get_verified_user)):

@router.get("/{id}/tags", response_model=list[TagModel])
 async def get_chat_tags_by_id(id: str, user=Depends(get_verified_user)):
-    tags = Tags.get_tags_by_chat_id_and_user_id(id, user.id)
-
-    if tags != None:
-        return tags
+    chat = Chats.get_chat_by_id_and_user_id(id, user.id)
+    if chat:
+        tags = chat.meta.get("tags", [])
+        return Tags.get_tags_by_ids_and_user_id(tags, user.id)
    else:
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED, detail=ERROR_MESSAGES.NOT_FOUND
@@ -421,22 +588,30 @@ async def get_chat_tags_by_id(id: str, user=Depends(get_verified_user)):
 ############################


-@router.post("/{id}/tags", response_model=Optional[ChatIdTagModel])
-async def add_chat_tag_by_id(
-    id: str, form_data: ChatIdTagForm, user=Depends(get_verified_user)
+@router.post("/{id}/tags", response_model=list[TagModel])
+async def add_tag_by_id_and_tag_name(
+    id: str, form_data: TagForm, user=Depends(get_verified_user)
 ):
-    tags = Tags.get_tags_by_chat_id_and_user_id(id, user.id)
+    chat = Chats.get_chat_by_id_and_user_id(id, user.id)
+    if chat:
+        tags = chat.meta.get("tags", [])
+        tag_id = form_data.name.replace(" ", "_").lower()

-    if form_data.tag_name not in tags:
-        tag = Tags.add_tag_to_chat(user.id, form_data)
-
-        if tag:
-            return tag
-        else:
+        if tag_id == "none":
            raise HTTPException(
-                status_code=status.HTTP_401_UNAUTHORIZED,
-                detail=ERROR_MESSAGES.NOT_FOUND,
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.DEFAULT("Tag name cannot be 'None'"),
            )
+
+        print(tags, tag_id)
+        if tag_id not in tags:
+            Chats.add_chat_tag_by_id_and_user_id_and_tag_name(
+                id, user.id, form_data.name
+            )
+
+        chat = Chats.get_chat_by_id_and_user_id(id, user.id)
+        tags = chat.meta.get("tags", [])
+        return Tags.get_tags_by_ids_and_user_id(tags, user.id)
    else:
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED, detail=ERROR_MESSAGES.DEFAULT()
@@ -448,16 +623,20 @@ async def add_chat_tag_by_id(
 ############################


-@router.delete("/{id}/tags", response_model=Optional[bool])
-async def delete_chat_tag_by_id(
-    id: str, form_data: ChatIdTagForm, user=Depends(get_verified_user)
+@router.delete("/{id}/tags", response_model=list[TagModel])
+async def delete_tag_by_id_and_tag_name(
+    id: str, form_data: TagForm, user=Depends(get_verified_user)
 ):
-    result = Tags.delete_tag_by_tag_name_and_chat_id_and_user_id(
-        form_data.tag_name, id, user.id
-    )
+    chat = Chats.get_chat_by_id_and_user_id(id, user.id)
+    if chat:
+        Chats.delete_tag_by_id_and_user_id_and_tag_name(id, user.id, form_data.name)

-    if result:
-        return result
+        if Chats.count_chats_by_tag_name_and_user_id(form_data.name, user.id) == 0:
+            Tags.delete_tag_by_name_and_user_id(form_data.name, user.id)
+
+        chat = Chats.get_chat_by_id_and_user_id(id, user.id)
+        tags = chat.meta.get("tags", [])
+        return Tags.get_tags_by_ids_and_user_id(tags, user.id)
    else:
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED, detail=ERROR_MESSAGES.NOT_FOUND
@@ -465,16 +644,21 @@ async def delete_chat_tag_by_id(


 ############################
-# DeleteAllChatTagsById
+# DeleteAllTagsById
 ############################


@router.delete("/{id}/tags/all", response_model=Optional[bool])
-async def delete_all_chat_tags_by_id(id: str, user=Depends(get_verified_user)):
-    result = Tags.delete_tags_by_chat_id_and_user_id(id, user.id)
+async def delete_all_tags_by_id(id: str, user=Depends(get_verified_user)):
+    chat = Chats.get_chat_by_id_and_user_id(id, user.id)
+    if chat:
+        Chats.delete_all_tags_by_id_and_user_id(id, user.id)

-    if result:
-        return result
+        for tag in chat.meta.get("tags", []):
+            if Chats.count_chats_by_tag_name_and_user_id(tag, user.id) == 0:
+                Tags.delete_tag_by_name_and_user_id(tag, user.id)
+
+        return True
    else:
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED, detail=ERROR_MESSAGES.NOT_FOUND
@@ -0,0 +1,159 @@
+from typing import Optional
+from fastapi import APIRouter, Depends, HTTPException, status, Request
+from pydantic import BaseModel
+
+from open_webui.apps.webui.models.users import Users, UserModel
+from open_webui.apps.webui.models.feedbacks import (
+    FeedbackModel,
+    FeedbackResponse,
+    FeedbackForm,
+    Feedbacks,
+)
+
+from open_webui.constants import ERROR_MESSAGES
+from open_webui.utils.utils import get_admin_user, get_verified_user
+
+router = APIRouter()
+
+
+############################
+# GetConfig
+############################
+
+
+@router.get("/config")
+async def get_config(request: Request, user=Depends(get_admin_user)):
+    return {
+        "ENABLE_EVALUATION_ARENA_MODELS": request.app.state.config.ENABLE_EVALUATION_ARENA_MODELS,
+        "EVALUATION_ARENA_MODELS": request.app.state.config.EVALUATION_ARENA_MODELS,
+    }
+
+
+############################
+# UpdateConfig
+############################
+
+
+class UpdateConfigForm(BaseModel):
+    ENABLE_EVALUATION_ARENA_MODELS: Optional[bool] = None
+    EVALUATION_ARENA_MODELS: Optional[list[dict]] = None
+
+
+@router.post("/config")
+async def update_config(
+    request: Request,
+    form_data: UpdateConfigForm,
+    user=Depends(get_admin_user),
+):
+    config = request.app.state.config
+    if form_data.ENABLE_EVALUATION_ARENA_MODELS is not None:
+        config.ENABLE_EVALUATION_ARENA_MODELS = form_data.ENABLE_EVALUATION_ARENA_MODELS
+    if form_data.EVALUATION_ARENA_MODELS is not None:
+        config.EVALUATION_ARENA_MODELS = form_data.EVALUATION_ARENA_MODELS
+    return {
+        "ENABLE_EVALUATION_ARENA_MODELS": config.ENABLE_EVALUATION_ARENA_MODELS,
+        "EVALUATION_ARENA_MODELS": config.EVALUATION_ARENA_MODELS,
+    }
+
+
+class FeedbackUserResponse(FeedbackResponse):
+    user: Optional[UserModel] = None
+
+
+@router.get("/feedbacks/all", response_model=list[FeedbackUserResponse])
+async def get_all_feedbacks(user=Depends(get_admin_user)):
+    feedbacks = Feedbacks.get_all_feedbacks()
+    return [
+        FeedbackUserResponse(
+            **feedback.model_dump(), user=Users.get_user_by_id(feedback.user_id)
+        )
+        for feedback in feedbacks
+    ]
+
+
+@router.delete("/feedbacks/all")
+async def delete_all_feedbacks(user=Depends(get_admin_user)):
+    success = Feedbacks.delete_all_feedbacks()
+    return success
+
+
+@router.get("/feedbacks/all/export", response_model=list[FeedbackModel])
+async def get_all_feedbacks(user=Depends(get_admin_user)):
+    feedbacks = Feedbacks.get_all_feedbacks()
+    return [
+        FeedbackModel(
+            **feedback.model_dump(), user=Users.get_user_by_id(feedback.user_id)
+        )
+        for feedback in feedbacks
+    ]
+
+
+@router.get("/feedbacks/user", response_model=list[FeedbackUserResponse])
+async def get_feedbacks(user=Depends(get_verified_user)):
+    feedbacks = Feedbacks.get_feedbacks_by_user_id(user.id)
+    return feedbacks
+
+
+@router.delete("/feedbacks", response_model=bool)
+async def delete_feedbacks(user=Depends(get_verified_user)):
+    success = Feedbacks.delete_feedbacks_by_user_id(user.id)
+    return success
+
+
+@router.post("/feedback", response_model=FeedbackModel)
+async def create_feedback(
+    request: Request,
+    form_data: FeedbackForm,
+    user=Depends(get_verified_user),
+):
+    feedback = Feedbacks.insert_new_feedback(user_id=user.id, form_data=form_data)
+    if not feedback:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=ERROR_MESSAGES.DEFAULT(),
+        )
+
+    return feedback
+
+
+@router.get("/feedback/{id}", response_model=FeedbackModel)
+async def get_feedback_by_id(id: str, user=Depends(get_verified_user)):
+    feedback = Feedbacks.get_feedback_by_id_and_user_id(id=id, user_id=user.id)
+
+    if not feedback:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail=ERROR_MESSAGES.NOT_FOUND
+        )
+
+    return feedback
+
+
+@router.post("/feedback/{id}", response_model=FeedbackModel)
+async def update_feedback_by_id(
+    id: str, form_data: FeedbackForm, user=Depends(get_verified_user)
+):
+    feedback = Feedbacks.update_feedback_by_id_and_user_id(
+        id=id, user_id=user.id, form_data=form_data
+    )
+
+    if not feedback:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail=ERROR_MESSAGES.NOT_FOUND
+        )
+
+    return feedback
+
+
+@router.delete("/feedback/{id}")
+async def delete_feedback_by_id(id: str, user=Depends(get_verified_user)):
+    if user.role == "admin":
+        success = Feedbacks.delete_feedback_by_id(id=id)
+    else:
+        success = Feedbacks.delete_feedback_by_id_and_user_id(id=id, user_id=user.id)
+
+    if not success:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail=ERROR_MESSAGES.NOT_FOUND
+        )
+
+    return success
@@ -1,16 +1,30 @@
 import logging
 import os
-import shutil
 import uuid
 from pathlib import Path
 from typing import Optional
+from pydantic import BaseModel
+import mimetypes
+
+from open_webui.storage.provider import Storage
+
+from open_webui.apps.webui.models.files import (
+    FileForm,
+    FileModel,
+    FileModelResponse,
+    Files,
+)
+from open_webui.apps.retrieval.main import process_file, ProcessFileForm

-from open_webui.apps.webui.models.files import FileForm, FileModel, Files
 from open_webui.config import UPLOAD_DIR
-from open_webui.constants import ERROR_MESSAGES
 from open_webui.env import SRC_LOG_LEVELS
+from open_webui.constants import ERROR_MESSAGES
+
+
 from fastapi import APIRouter, Depends, File, HTTPException, UploadFile, status
-from fastapi.responses import FileResponse
+from fastapi.responses import FileResponse, StreamingResponse
+
+
 from open_webui.utils.utils import get_admin_user, get_verified_user

 log = logging.getLogger(__name__)
@@ -24,7 +38,7 @@ router = APIRouter()
 ############################


-@router.post("/")
+@router.post("/", response_model=FileModelResponse)
 def upload_file(file: UploadFile = File(...), user=Depends(get_verified_user)):
    log.info(f"file.content_type: {file.content_type}")
    try:
@@ -35,31 +49,39 @@ def upload_file(file: UploadFile = File(...), user=Depends(get_verified_user)):
        id = str(uuid.uuid4())
        name = filename
        filename = f"{id}_{filename}"
-        file_path = f"{UPLOAD_DIR}/{filename}"
+        contents, file_path = Storage.upload_file(file.file, filename)

-        contents = file.file.read()
-        with open(file_path, "wb") as f:
-            f.write(contents)
-            f.close()
-
-        file = Files.insert_new_file(
+        file_item = Files.insert_new_file(
            user.id,
            FileForm(
                **{
                    "id": id,
                    "filename": filename,
+                    "path": file_path,
                    "meta": {
                        "name": name,
                        "content_type": file.content_type,
                        "size": len(contents),
-                        "path": file_path,
                    },
                }
            ),
        )

-        if file:
-            return file
+        try:
+            process_file(ProcessFileForm(file_id=id))
+            file_item = Files.get_file_by_id(id=id)
+        except Exception as e:
+            log.exception(e)
+            log.error(f"Error processing file: {file_item.id}")
+            file_item = FileModelResponse(
+                **{
+                    **file_item.model_dump(),
+                    "error": str(e.detail) if hasattr(e, "detail") else str(e),
+                }
+            )
+
+        if file_item:
+            return file_item
        else:
            raise HTTPException(
                status_code=status.HTTP_400_BAD_REQUEST,
@@ -79,7 +101,7 @@ def upload_file(file: UploadFile = File(...), user=Depends(get_verified_user)):
 ############################


-@router.get("/", response_model=list[FileModel])
+@router.get("/", response_model=list[FileModelResponse])
 async def list_files(user=Depends(get_verified_user)):
    if user.role == "admin":
        files = Files.get_files()
@@ -96,27 +118,16 @@ async def list_files(user=Depends(get_verified_user)):
@router.delete("/all")
 async def delete_all_files(user=Depends(get_admin_user)):
    result = Files.delete_all_files()
-
    if result:
-        folder = f"{UPLOAD_DIR}"
        try:
-            # Check if the directory exists
-            if os.path.exists(folder):
-                # Iterate over all the files and directories in the specified directory
-                for filename in os.listdir(folder):
-                    file_path = os.path.join(folder, filename)
-                    try:
-                        if os.path.isfile(file_path) or os.path.islink(file_path):
-                            os.unlink(file_path)  # Remove the file or link
-                        elif os.path.isdir(file_path):
-                            shutil.rmtree(file_path)  # Remove the directory
-                    except Exception as e:
-                        print(f"Failed to delete {file_path}. Reason: {e}")
-            else:
-                print(f"The directory {folder} does not exist")
+            Storage.delete_all_files()
        except Exception as e:
-            print(f"Failed to process the directory {folder}. Reason: {e}")
-
+            log.exception(e)
+            log.error(f"Error deleting files")
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.DEFAULT("Error deleting files"),
+            )
        return {"message": "All files deleted successfully"}
    else:
        raise HTTPException(
@@ -143,26 +154,86 @@ async def get_file_by_id(id: str, user=Depends(get_verified_user)):
        )


+############################
+# Get File Data Content By Id
+############################
+
+
+@router.get("/{id}/data/content")
+async def get_file_data_content_by_id(id: str, user=Depends(get_verified_user)):
+    file = Files.get_file_by_id(id)
+
+    if file and (file.user_id == user.id or user.role == "admin"):
+        return {"content": file.data.get("content", "")}
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+
+############################
+# Update File Data Content By Id
+############################
+
+
+class ContentForm(BaseModel):
+    content: str
+
+
+@router.post("/{id}/data/content/update")
+async def update_file_data_content_by_id(
+    id: str, form_data: ContentForm, user=Depends(get_verified_user)
+):
+    file = Files.get_file_by_id(id)
+
+    if file and (file.user_id == user.id or user.role == "admin"):
+        try:
+            process_file(ProcessFileForm(file_id=id, content=form_data.content))
+            file = Files.get_file_by_id(id=id)
+        except Exception as e:
+            log.exception(e)
+            log.error(f"Error processing file: {file.id}")
+
+        return {"content": file.data.get("content", "")}
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+
 ############################
 # Get File Content By Id
 ############################


-@router.get("/{id}/content", response_model=Optional[FileModel])
+@router.get("/{id}/content")
 async def get_file_content_by_id(id: str, user=Depends(get_verified_user)):
    file = Files.get_file_by_id(id)
-
    if file and (file.user_id == user.id or user.role == "admin"):
-        file_path = Path(file.meta["path"])
+        try:
+            file_path = Storage.get_file(file.path)
+            file_path = Path(file_path)

-        # Check if the file already exists in the cache
-        if file_path.is_file():
-            print(f"file_path: {file_path}")
-            return FileResponse(file_path)
-        else:
+            # Check if the file already exists in the cache
+            if file_path.is_file():
+                print(f"file_path: {file_path}")
+                headers = {
+                    "Content-Disposition": f'attachment; filename="{file.meta.get("name", file.filename)}"'
+                }
+                return FileResponse(file_path, headers=headers)
+            else:
+                raise HTTPException(
+                    status_code=status.HTTP_404_NOT_FOUND,
+                    detail=ERROR_MESSAGES.NOT_FOUND,
+                )
+        except Exception as e:
+            log.exception(e)
+            log.error(f"Error getting file content")
            raise HTTPException(
-                status_code=status.HTTP_404_NOT_FOUND,
-                detail=ERROR_MESSAGES.NOT_FOUND,
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.DEFAULT("Error getting file content"),
            )
    else:
        raise HTTPException(
@@ -171,21 +242,72 @@ async def get_file_content_by_id(id: str, user=Depends(get_verified_user)):
        )


-@router.get("/{id}/content/{file_name}", response_model=Optional[FileModel])
+@router.get("/{id}/content/html")
+async def get_html_file_content_by_id(id: str, user=Depends(get_verified_user)):
+    file = Files.get_file_by_id(id)
+    if file and (file.user_id == user.id or user.role == "admin"):
+        try:
+            file_path = Storage.get_file(file.path)
+            file_path = Path(file_path)
+
+            # Check if the file already exists in the cache
+            if file_path.is_file():
+                print(f"file_path: {file_path}")
+                return FileResponse(file_path)
+            else:
+                raise HTTPException(
+                    status_code=status.HTTP_404_NOT_FOUND,
+                    detail=ERROR_MESSAGES.NOT_FOUND,
+                )
+        except Exception as e:
+            log.exception(e)
+            log.error(f"Error getting file content")
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.DEFAULT("Error getting file content"),
+            )
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+
+@router.get("/{id}/content/{file_name}")
 async def get_file_content_by_id(id: str, user=Depends(get_verified_user)):
    file = Files.get_file_by_id(id)

    if file and (file.user_id == user.id or user.role == "admin"):
-        file_path = Path(file.meta["path"])
+        file_path = file.path
+        if file_path:
+            file_path = Storage.get_file(file_path)
+            file_path = Path(file_path)

-        # Check if the file already exists in the cache
-        if file_path.is_file():
-            print(f"file_path: {file_path}")
-            return FileResponse(file_path)
+            # Check if the file already exists in the cache
+            if file_path.is_file():
+                print(f"file_path: {file_path}")
+                headers = {
+                    "Content-Disposition": f'attachment; filename="{file.meta.get("name", file.filename)}"'
+                }
+                return FileResponse(file_path, headers=headers)
+            else:
+                raise HTTPException(
+                    status_code=status.HTTP_404_NOT_FOUND,
+                    detail=ERROR_MESSAGES.NOT_FOUND,
+                )
        else:
-            raise HTTPException(
-                status_code=status.HTTP_404_NOT_FOUND,
-                detail=ERROR_MESSAGES.NOT_FOUND,
+            # File path doesn’t exist, return the content as .txt if possible
+            file_content = file.content.get("content", "")
+            file_name = file.filename
+
+            # Create a generator that encodes the file content
+            def generator():
+                yield file_content.encode("utf-8")
+
+            return StreamingResponse(
+                generator(),
+                media_type="text/plain",
+                headers={"Content-Disposition": f"attachment; filename={file_name}"},
            )
    else:
        raise HTTPException(
@@ -205,6 +327,15 @@ async def delete_file_by_id(id: str, user=Depends(get_verified_user)):
    if file and (file.user_id == user.id or user.role == "admin"):
        result = Files.delete_file_by_id(id)
        if result:
+            try:
+                Storage.delete_file(file.filename)
+            except Exception as e:
+                log.exception(e)
+                log.error(f"Error deleting files")
+                raise HTTPException(
+                    status_code=status.HTTP_400_BAD_REQUEST,
+                    detail=ERROR_MESSAGES.DEFAULT("Error deleting files"),
+                )
            return {"message": "File deleted successfully"}
        else:
            raise HTTPException(
@@ -0,0 +1,251 @@
+import logging
+import os
+import shutil
+import uuid
+from pathlib import Path
+from typing import Optional
+from pydantic import BaseModel
+import mimetypes
+
+
+from open_webui.apps.webui.models.folders import (
+    FolderForm,
+    FolderModel,
+    Folders,
+)
+from open_webui.apps.webui.models.chats import Chats
+
+from open_webui.config import UPLOAD_DIR
+from open_webui.env import SRC_LOG_LEVELS
+from open_webui.constants import ERROR_MESSAGES
+
+
+from fastapi import APIRouter, Depends, File, HTTPException, UploadFile, status
+from fastapi.responses import FileResponse, StreamingResponse
+
+
+from open_webui.utils.utils import get_admin_user, get_verified_user
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["MODELS"])
+
+
+router = APIRouter()
+
+
+############################
+# Get Folders
+############################
+
+
+@router.get("/", response_model=list[FolderModel])
+async def get_folders(user=Depends(get_verified_user)):
+    folders = Folders.get_folders_by_user_id(user.id)
+
+    return [
+        {
+            **folder.model_dump(),
+            "items": {
+                "chats": [
+                    {"title": chat.title, "id": chat.id}
+                    for chat in Chats.get_chats_by_folder_id_and_user_id(
+                        folder.id, user.id
+                    )
+                ]
+            },
+        }
+        for folder in folders
+    ]
+
+
+############################
+# Create Folder
+############################
+
+
+@router.post("/")
+def create_folder(form_data: FolderForm, user=Depends(get_verified_user)):
+    folder = Folders.get_folder_by_parent_id_and_user_id_and_name(
+        None, user.id, form_data.name
+    )
+
+    if folder:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=ERROR_MESSAGES.DEFAULT("Folder already exists"),
+        )
+
+    try:
+        folder = Folders.insert_new_folder(user.id, form_data.name)
+        return folder
+    except Exception as e:
+        log.exception(e)
+        log.error("Error creating folder")
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=ERROR_MESSAGES.DEFAULT("Error creating folder"),
+        )
+
+
+############################
+# Get Folders By Id
+############################
+
+
+@router.get("/{id}", response_model=Optional[FolderModel])
+async def get_folder_by_id(id: str, user=Depends(get_verified_user)):
+    folder = Folders.get_folder_by_id_and_user_id(id, user.id)
+    if folder:
+        return folder
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+
+############################
+# Update Folder Name By Id
+############################
+
+
+@router.post("/{id}/update")
+async def update_folder_name_by_id(
+    id: str, form_data: FolderForm, user=Depends(get_verified_user)
+):
+    folder = Folders.get_folder_by_id_and_user_id(id, user.id)
+    if folder:
+        existing_folder = Folders.get_folder_by_parent_id_and_user_id_and_name(
+            folder.parent_id, user.id, form_data.name
+        )
+        if existing_folder:
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.DEFAULT("Folder already exists"),
+            )
+
+        try:
+            folder = Folders.update_folder_name_by_id_and_user_id(
+                id, user.id, form_data.name
+            )
+
+            return folder
+        except Exception as e:
+            log.exception(e)
+            log.error(f"Error updating folder: {id}")
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.DEFAULT("Error updating folder"),
+            )
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+
+############################
+# Update Folder Parent Id By Id
+############################
+
+
+class FolderParentIdForm(BaseModel):
+    parent_id: Optional[str] = None
+
+
+@router.post("/{id}/update/parent")
+async def update_folder_parent_id_by_id(
+    id: str, form_data: FolderParentIdForm, user=Depends(get_verified_user)
+):
+    folder = Folders.get_folder_by_id_and_user_id(id, user.id)
+    if folder:
+        existing_folder = Folders.get_folder_by_parent_id_and_user_id_and_name(
+            form_data.parent_id, user.id, folder.name
+        )
+
+        if existing_folder:
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.DEFAULT("Folder already exists"),
+            )
+
+        try:
+            folder = Folders.update_folder_parent_id_by_id_and_user_id(
+                id, user.id, form_data.parent_id
+            )
+            return folder
+        except Exception as e:
+            log.exception(e)
+            log.error(f"Error updating folder: {id}")
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.DEFAULT("Error updating folder"),
+            )
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+
+############################
+# Update Folder Is Expanded By Id
+############################
+
+
+class FolderIsExpandedForm(BaseModel):
+    is_expanded: bool
+
+
+@router.post("/{id}/update/expanded")
+async def update_folder_is_expanded_by_id(
+    id: str, form_data: FolderIsExpandedForm, user=Depends(get_verified_user)
+):
+    folder = Folders.get_folder_by_id_and_user_id(id, user.id)
+    if folder:
+        try:
+            folder = Folders.update_folder_is_expanded_by_id_and_user_id(
+                id, user.id, form_data.is_expanded
+            )
+            return folder
+        except Exception as e:
+            log.exception(e)
+            log.error(f"Error updating folder: {id}")
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.DEFAULT("Error updating folder"),
+            )
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+
+############################
+# Delete Folder By Id
+############################
+
+
+@router.delete("/{id}")
+async def delete_folder_by_id(id: str, user=Depends(get_verified_user)):
+    folder = Folders.get_folder_by_id_and_user_id(id, user.id)
+    if folder:
+        try:
+            result = Folders.delete_folder_by_id_and_user_id(id, user.id)
+            if result:
+                return result
+            else:
+                raise Exception("Error deleting folder")
+        except Exception as e:
+            log.exception(e)
+            log.error(f"Error deleting folder: {id}")
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.DEFAULT("Error deleting folder"),
+            )
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
@@ -8,8 +8,8 @@ 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.config import CACHE_DIR, FUNCTIONS_DIR
+from open_webui.apps.webui.utils import load_function_module_by_id, replace_imports
+from open_webui.config import CACHE_DIR
 from open_webui.constants import ERROR_MESSAGES
 from fastapi import APIRouter, Depends, HTTPException, Request, status
 from open_webui.utils.utils import get_admin_user, get_verified_user
@@ -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


@@ -0,0 +1,381 @@
+import json
+from typing import Optional, Union
+from pydantic import BaseModel
+from fastapi import APIRouter, Depends, HTTPException, status
+import logging
+
+from open_webui.apps.webui.models.knowledge import (
+    Knowledges,
+    KnowledgeUpdateForm,
+    KnowledgeForm,
+    KnowledgeResponse,
+)
+from open_webui.apps.webui.models.files import Files, FileModel
+from open_webui.apps.retrieval.vector.connector import VECTOR_DB_CLIENT
+from open_webui.apps.retrieval.main import process_file, ProcessFileForm
+
+
+from open_webui.constants import ERROR_MESSAGES
+from open_webui.utils.utils import get_admin_user, get_verified_user
+from open_webui.env import SRC_LOG_LEVELS
+
+
+log = logging.getLogger(__name__)
+log.setLevel(SRC_LOG_LEVELS["MODELS"])
+
+router = APIRouter()
+
+############################
+# GetKnowledgeItems
+############################
+
+
+@router.get(
+    "/", response_model=Optional[Union[list[KnowledgeResponse], KnowledgeResponse]]
+)
+async def get_knowledge_items(
+    id: Optional[str] = None, user=Depends(get_verified_user)
+):
+    if id:
+        knowledge = Knowledges.get_knowledge_by_id(id=id)
+
+        if knowledge:
+            return knowledge
+        else:
+            raise HTTPException(
+                status_code=status.HTTP_401_UNAUTHORIZED,
+                detail=ERROR_MESSAGES.NOT_FOUND,
+            )
+    else:
+        knowledge_bases = []
+
+        for knowledge in Knowledges.get_knowledge_items():
+
+            files = []
+            if knowledge.data:
+                files = Files.get_file_metadatas_by_ids(
+                    knowledge.data.get("file_ids", [])
+                )
+
+                # Check if all files exist
+                if len(files) != len(knowledge.data.get("file_ids", [])):
+                    missing_files = list(
+                        set(knowledge.data.get("file_ids", []))
+                        - set([file.id for file in files])
+                    )
+                    if missing_files:
+                        data = knowledge.data or {}
+                        file_ids = data.get("file_ids", [])
+
+                        for missing_file in missing_files:
+                            file_ids.remove(missing_file)
+
+                        data["file_ids"] = file_ids
+                        Knowledges.update_knowledge_by_id(
+                            id=knowledge.id, form_data=KnowledgeUpdateForm(data=data)
+                        )
+
+                        files = Files.get_file_metadatas_by_ids(file_ids)
+
+            knowledge_bases.append(
+                KnowledgeResponse(
+                    **knowledge.model_dump(),
+                    files=files,
+                )
+            )
+        return knowledge_bases
+
+
+############################
+# CreateNewKnowledge
+############################
+
+
+@router.post("/create", response_model=Optional[KnowledgeResponse])
+async def create_new_knowledge(form_data: KnowledgeForm, user=Depends(get_admin_user)):
+    knowledge = Knowledges.insert_new_knowledge(user.id, form_data)
+
+    if knowledge:
+        return knowledge
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=ERROR_MESSAGES.FILE_EXISTS,
+        )
+
+
+############################
+# GetKnowledgeById
+############################
+
+
+class KnowledgeFilesResponse(KnowledgeResponse):
+    files: list[FileModel]
+
+
+@router.get("/{id}", response_model=Optional[KnowledgeFilesResponse])
+async def get_knowledge_by_id(id: str, user=Depends(get_verified_user)):
+    knowledge = Knowledges.get_knowledge_by_id(id=id)
+
+    if knowledge:
+        file_ids = knowledge.data.get("file_ids", []) if knowledge.data else []
+        files = Files.get_files_by_ids(file_ids)
+
+        return KnowledgeFilesResponse(
+            **knowledge.model_dump(),
+            files=files,
+        )
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_401_UNAUTHORIZED,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+
+############################
+# UpdateKnowledgeById
+############################
+
+
+@router.post("/{id}/update", response_model=Optional[KnowledgeFilesResponse])
+async def update_knowledge_by_id(
+    id: str,
+    form_data: KnowledgeUpdateForm,
+    user=Depends(get_admin_user),
+):
+    knowledge = Knowledges.update_knowledge_by_id(id=id, form_data=form_data)
+
+    if knowledge:
+        file_ids = knowledge.data.get("file_ids", []) if knowledge.data else []
+        files = Files.get_files_by_ids(file_ids)
+
+        return KnowledgeFilesResponse(
+            **knowledge.model_dump(),
+            files=files,
+        )
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=ERROR_MESSAGES.ID_TAKEN,
+        )
+
+
+############################
+# AddFileToKnowledge
+############################
+
+
+class KnowledgeFileIdForm(BaseModel):
+    file_id: str
+
+
+@router.post("/{id}/file/add", response_model=Optional[KnowledgeFilesResponse])
+def add_file_to_knowledge_by_id(
+    id: str,
+    form_data: KnowledgeFileIdForm,
+    user=Depends(get_admin_user),
+):
+    knowledge = Knowledges.get_knowledge_by_id(id=id)
+    file = Files.get_file_by_id(form_data.file_id)
+    if not file:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+    if not file.data:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=ERROR_MESSAGES.FILE_NOT_PROCESSED,
+        )
+
+    # Add content to the vector database
+    try:
+        process_file(ProcessFileForm(file_id=form_data.file_id, collection_name=id))
+    except Exception as e:
+        log.debug(e)
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=str(e),
+        )
+
+    if knowledge:
+        data = knowledge.data or {}
+        file_ids = data.get("file_ids", [])
+
+        if form_data.file_id not in file_ids:
+            file_ids.append(form_data.file_id)
+            data["file_ids"] = file_ids
+
+            knowledge = Knowledges.update_knowledge_by_id(
+                id=id, form_data=KnowledgeUpdateForm(data=data)
+            )
+
+            if knowledge:
+                files = Files.get_files_by_ids(file_ids)
+
+                return KnowledgeFilesResponse(
+                    **knowledge.model_dump(),
+                    files=files,
+                )
+            else:
+                raise HTTPException(
+                    status_code=status.HTTP_400_BAD_REQUEST,
+                    detail=ERROR_MESSAGES.DEFAULT("knowledge"),
+                )
+        else:
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.DEFAULT("file_id"),
+            )
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+
+@router.post("/{id}/file/update", response_model=Optional[KnowledgeFilesResponse])
+def update_file_from_knowledge_by_id(
+    id: str,
+    form_data: KnowledgeFileIdForm,
+    user=Depends(get_admin_user),
+):
+    knowledge = Knowledges.get_knowledge_by_id(id=id)
+    file = Files.get_file_by_id(form_data.file_id)
+    if not file:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+    # Remove content from the vector database
+    VECTOR_DB_CLIENT.delete(
+        collection_name=knowledge.id, filter={"file_id": form_data.file_id}
+    )
+
+    # Add content to the vector database
+    try:
+        process_file(ProcessFileForm(file_id=form_data.file_id, collection_name=id))
+    except Exception as e:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=str(e),
+        )
+
+    if knowledge:
+        data = knowledge.data or {}
+        file_ids = data.get("file_ids", [])
+
+        files = Files.get_files_by_ids(file_ids)
+
+        return KnowledgeFilesResponse(
+            **knowledge.model_dump(),
+            files=files,
+        )
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+
+############################
+# RemoveFileFromKnowledge
+############################
+
+
+@router.post("/{id}/file/remove", response_model=Optional[KnowledgeFilesResponse])
+def remove_file_from_knowledge_by_id(
+    id: str,
+    form_data: KnowledgeFileIdForm,
+    user=Depends(get_admin_user),
+):
+    knowledge = Knowledges.get_knowledge_by_id(id=id)
+    file = Files.get_file_by_id(form_data.file_id)
+    if not file:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+    # Remove content from the vector database
+    VECTOR_DB_CLIENT.delete(
+        collection_name=knowledge.id, filter={"file_id": form_data.file_id}
+    )
+
+    result = VECTOR_DB_CLIENT.query(
+        collection_name=knowledge.id,
+        filter={"file_id": form_data.file_id},
+    )
+
+    Files.delete_file_by_id(form_data.file_id)
+
+    if knowledge:
+        data = knowledge.data or {}
+        file_ids = data.get("file_ids", [])
+
+        if form_data.file_id in file_ids:
+            file_ids.remove(form_data.file_id)
+            data["file_ids"] = file_ids
+
+            knowledge = Knowledges.update_knowledge_by_id(
+                id=id, form_data=KnowledgeUpdateForm(data=data)
+            )
+
+            if knowledge:
+                files = Files.get_files_by_ids(file_ids)
+
+                return KnowledgeFilesResponse(
+                    **knowledge.model_dump(),
+                    files=files,
+                )
+            else:
+                raise HTTPException(
+                    status_code=status.HTTP_400_BAD_REQUEST,
+                    detail=ERROR_MESSAGES.DEFAULT("knowledge"),
+                )
+        else:
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail=ERROR_MESSAGES.DEFAULT("file_id"),
+            )
+    else:
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail=ERROR_MESSAGES.NOT_FOUND,
+        )
+
+
+############################
+# ResetKnowledgeById
+############################
+
+
+@router.post("/{id}/reset", response_model=Optional[KnowledgeResponse])
+async def reset_knowledge_by_id(id: str, user=Depends(get_admin_user)):
+    try:
+        VECTOR_DB_CLIENT.delete_collection(collection_name=id)
+    except Exception as e:
+        log.debug(e)
+        pass
+
+    knowledge = Knowledges.update_knowledge_by_id(
+        id=id, form_data=KnowledgeUpdateForm(data={"file_ids": []})
+    )
+    return knowledge
+
+
+############################
+# DeleteKnowledgeById
+############################
+
+
+@router.delete("/{id}/delete", response_model=bool)
+async def delete_knowledge_by_id(id: str, user=Depends(get_admin_user)):
+    try:
+        VECTOR_DB_CLIENT.delete_collection(collection_name=id)
+    except Exception as e:
+        log.debug(e)
+        pass
+    result = Knowledges.delete_knowledge_by_id(id=id)
+    return result
@@ -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.retrieval.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
@@ -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
 ############################
@@ -3,16 +3,13 @@ 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
 from open_webui.utils.tools import get_tools_specs
 from open_webui.utils.utils import get_admin_user, get_verified_user

-TOOLS_DIR = f"{DATA_DIR}/tools"
-os.makedirs(TOOLS_DIR, exist_ok=True)
-

 router = APIRouter()

@@ -59,12 +56,11 @@ 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 +122,11 @@ 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 +171,6 @@ 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


@@ -1,15 +1,14 @@
-import site
-from pathlib import Path
-
 import black
 import markdown
+
+from open_webui.apps.webui.models.chats import ChatTitleMessagesForm
 from open_webui.config import DATA_DIR, ENABLE_ADMIN_EXPORT
 from open_webui.constants import ERROR_MESSAGES
 from fastapi import APIRouter, Depends, HTTPException, Response, status
-from fpdf import FPDF
 from pydantic import BaseModel
 from starlette.responses import FileResponse
 from open_webui.utils.misc import get_gravatar_url
+from open_webui.utils.pdf_generator import PDFGenerator
 from open_webui.utils.utils import get_admin_user

 router = APIRouter()
@@ -55,61 +54,19 @@ class ChatForm(BaseModel):

@router.post("/pdf")
 async def download_chat_as_pdf(
-    form_data: ChatForm,
+    form_data: ChatTitleMessagesForm,
 ):
-    pdf = FPDF()
-    pdf.add_page()
+    try:
+        pdf_bytes = PDFGenerator(form_data).generate_chat_pdf()

-    # 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"
-    # When running using `pip install -e .` the static directory is in the site packages.
-    # This path only works if `open-webui serve` is run from the root of this project.
-    if not FONTS_DIR.exists():
-        FONTS_DIR = Path("./backend/static/fonts")
-
-    pdf.add_font("NotoSans", "", f"{FONTS_DIR}/NotoSans-Regular.ttf")
-    pdf.add_font("NotoSans", "b", f"{FONTS_DIR}/NotoSans-Bold.ttf")
-    pdf.add_font("NotoSans", "i", f"{FONTS_DIR}/NotoSans-Italic.ttf")
-    pdf.add_font("NotoSansKR", "", f"{FONTS_DIR}/NotoSansKR-Regular.ttf")
-    pdf.add_font("NotoSansJP", "", f"{FONTS_DIR}/NotoSansJP-Regular.ttf")
-    pdf.add_font("NotoSansSC", "", f"{FONTS_DIR}/NotoSansSC-Regular.ttf")
-
-    pdf.set_font("NotoSans", size=12)
-    pdf.set_fallback_fonts(["NotoSansKR", "NotoSansJP", "NotoSansSC"])
-
-    pdf.set_auto_page_break(auto=True, margin=15)
-
-    # Adjust the effective page width for multi_cell
-    effective_page_width = (
-        pdf.w - 2 * pdf.l_margin - 10
-    )  # Subtracted an additional 10 for extra padding
-
-    # Add chat messages
-    for message in form_data.messages:
-        role = message["role"]
-        content = message["content"]
-        pdf.set_font("NotoSans", "B", size=14)  # Bold for the role
-        pdf.multi_cell(effective_page_width, 10, f"{role.upper()}", 0, "L")
-        pdf.ln(1)  # Extra space between messages
-
-        pdf.set_font("NotoSans", size=10)  # Regular for content
-        pdf.multi_cell(effective_page_width, 6, content, 0, "L")
-        pdf.ln(1.5)  # Extra space between messages
-
-    # Save the pdf with name .pdf
-    pdf_bytes = pdf.output()
-
-    return Response(
-        content=bytes(pdf_bytes),
-        media_type="application/pdf",
-        headers={"Content-Disposition": "attachment;filename=chat.pdf"},
-    )
+        return Response(
+            content=pdf_bytes,
+            media_type="application/pdf",
+            headers={"Content-Disposition": "attachment;filename=chat.pdf"},
+        )
+    except Exception as e:
+        print(e)
+        raise HTTPException(status_code=400, detail=str(e))


@router.get("/db/download")
@@ -3,15 +3,16 @@ 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 +20,25 @@ 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 +46,102 @@ 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

-    try:
+        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", ""))
-        spec.loader.exec_module(module)
+
+    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)
+    temp_file.close()
+    try:
+        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)
+    temp_file.close()
+    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):
@@ -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,
@@ -88,6 +87,12 @@ def save_to_db(data):
        db.commit()


+def reset_config():
+    with get_db() as db:
+        db.query(Config).delete()
+        db.commit()
+
+
 # When initializing, check if config.json exists and migrate it to the database
 if os.path.exists(f"{DATA_DIR}/config.json"):
    data = load_json_config()
@@ -378,7 +383,7 @@ OAUTH_USERNAME_CLAIM = PersistentConfig(
 )

 OAUTH_PICTURE_CLAIM = PersistentConfig(
-    "OAUTH_USERNAME_CLAIM",
+    "OAUTH_PICTURE_CLAIM",
    "oauth.oidc.avatar_claim",
    os.environ.get("OAUTH_PICTURE_CLAIM", "picture"),
 )
@@ -389,6 +394,33 @@ OAUTH_EMAIL_CLAIM = PersistentConfig(
    os.environ.get("OAUTH_EMAIL_CLAIM", "email"),
 )

+ENABLE_OAUTH_ROLE_MANAGEMENT = PersistentConfig(
+    "ENABLE_OAUTH_ROLE_MANAGEMENT",
+    "oauth.enable_role_mapping",
+    os.environ.get("ENABLE_OAUTH_ROLE_MANAGEMENT", "False").lower() == "true",
+)
+
+OAUTH_ROLES_CLAIM = PersistentConfig(
+    "OAUTH_ROLES_CLAIM",
+    "oauth.roles_claim",
+    os.environ.get("OAUTH_ROLES_CLAIM", "roles"),
+)
+
+OAUTH_ALLOWED_ROLES = PersistentConfig(
+    "OAUTH_ALLOWED_ROLES",
+    "oauth.allowed_roles",
+    [
+        role.strip()
+        for role in os.environ.get("OAUTH_ALLOWED_ROLES", "user,admin").split(",")
+    ],
+)
+
+OAUTH_ADMIN_ROLES = PersistentConfig(
+    "OAUTH_ADMIN_ROLES",
+    "oauth.admin_roles",
+    [role.strip() for role in os.environ.get("OAUTH_ADMIN_ROLES", "admin").split(",")],
+)
+

 def load_oauth_providers():
    OAUTH_PROVIDERS.clear()
@@ -501,6 +533,18 @@ if CUSTOM_NAME:
        pass


+####################################
+# STORAGE PROVIDER
+####################################
+
+STORAGE_PROVIDER = os.environ.get("STORAGE_PROVIDER", "")  # defaults to local, s3
+
+S3_ACCESS_KEY_ID = os.environ.get("S3_ACCESS_KEY_ID", None)
+S3_SECRET_ACCESS_KEY = os.environ.get("S3_SECRET_ACCESS_KEY", None)
+S3_REGION_NAME = os.environ.get("S3_REGION_NAME", None)
+S3_BUCKET_NAME = os.environ.get("S3_BUCKET_NAME", None)
+S3_ENDPOINT_URL = os.environ.get("S3_ENDPOINT_URL", None)
+
 ####################################
 # File Upload DIR
 ####################################
@@ -516,69 +560,10 @@ Path(UPLOAD_DIR).mkdir(parents=True, exist_ok=True)
 CACHE_DIR = f"{DATA_DIR}/cache"
 Path(CACHE_DIR).mkdir(parents=True, exist_ok=True)

-
-####################################
-# Docs DIR
-####################################
-
-DOCS_DIR = os.getenv("DOCS_DIR", f"{DATA_DIR}/docs")
-Path(DOCS_DIR).mkdir(parents=True, exist_ok=True)
-
-
-####################################
-# Tools DIR
-####################################
-
-TOOLS_DIR = os.getenv("TOOLS_DIR", f"{DATA_DIR}/tools")
-Path(TOOLS_DIR).mkdir(parents=True, exist_ok=True)
-
-
-####################################
-# Functions DIR
-####################################
-
-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
 ####################################

-
 ENABLE_OLLAMA_API = PersistentConfig(
    "ENABLE_OLLAMA_API",
    "ollama.enable",
@@ -590,16 +575,6 @@ OLLAMA_API_BASE_URL = os.environ.get(
 )

 OLLAMA_BASE_URL = os.environ.get("OLLAMA_BASE_URL", "")
-AIOHTTP_CLIENT_TIMEOUT = os.environ.get("AIOHTTP_CLIENT_TIMEOUT", "")
-
-if AIOHTTP_CLIENT_TIMEOUT == "":
-    AIOHTTP_CLIENT_TIMEOUT = None
-else:
-    try:
-        AIOHTTP_CLIENT_TIMEOUT = int(AIOHTTP_CLIENT_TIMEOUT)
-    except Exception:
-        AIOHTTP_CLIENT_TIMEOUT = 300
-

 K8S_FLAG = os.environ.get("K8S_FLAG", "")
 USE_OLLAMA_DOCKER = os.environ.get("USE_OLLAMA_DOCKER", "false")
@@ -776,6 +751,28 @@ USER_PERMISSIONS = PersistentConfig(
    },
 )

+
+ENABLE_EVALUATION_ARENA_MODELS = PersistentConfig(
+    "ENABLE_EVALUATION_ARENA_MODELS",
+    "evaluation.arena.enable",
+    os.environ.get("ENABLE_EVALUATION_ARENA_MODELS", "True").lower() == "true",
+)
+EVALUATION_ARENA_MODELS = PersistentConfig(
+    "EVALUATION_ARENA_MODELS",
+    "evaluation.arena.models",
+    [],
+)
+
+DEFAULT_ARENA_MODEL = {
+    "id": "arena-model",
+    "name": "Arena Model",
+    "meta": {
+        "profile_image_url": "/favicon.png",
+        "description": "Submit your questions to anonymous AI chatbots and vote on the best response.",
+        "model_ids": None,
+    },
+}
+
 ENABLE_MODEL_FILTER = PersistentConfig(
    "ENABLE_MODEL_FILTER",
    "model_filter.enable",
@@ -898,76 +895,43 @@ 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
+TAGS_GENERATION_PROMPT_TEMPLATE = PersistentConfig(
+    "TAGS_GENERATION_PROMPT_TEMPLATE",
+    "task.tags.prompt_template",
+    os.environ.get("TAGS_GENERATION_PROMPT_TEMPLATE", ""),
+)

-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 +948,32 @@ 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")
+
+# Qdrant
+QDRANT_URI = os.environ.get("QDRANT_URI", None)
+
+####################################
+# Information Retrieval (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",
@@ -1052,10 +1040,13 @@ RAG_EMBEDDING_MODEL_TRUST_REMOTE_CODE = (
    os.environ.get("RAG_EMBEDDING_MODEL_TRUST_REMOTE_CODE", "").lower() == "true"
 )

-RAG_EMBEDDING_OPENAI_BATCH_SIZE = PersistentConfig(
-    "RAG_EMBEDDING_OPENAI_BATCH_SIZE",
-    "rag.embedding_openai_batch_size",
-    int(os.environ.get("RAG_EMBEDDING_OPENAI_BATCH_SIZE", "1")),
+RAG_EMBEDDING_BATCH_SIZE = PersistentConfig(
+    "RAG_EMBEDDING_BATCH_SIZE",
+    "rag.embedding_batch_size",
+    int(
+        os.environ.get("RAG_EMBEDDING_BATCH_SIZE")
+        or os.environ.get("RAG_EMBEDDING_OPENAI_BATCH_SIZE", "1")
+    ),
 )

 RAG_RERANKING_MODEL = PersistentConfig(
@@ -1075,35 +1066,23 @@ RAG_RERANKING_MODEL_TRUST_REMOTE_CODE = (
 )


-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,
-    )
+RAG_TEXT_SPLITTER = PersistentConfig(
+    "RAG_TEXT_SPLITTER",
+    "rag.text_splitter",
+    os.environ.get("RAG_TEXT_SPLITTER", ""),
+)


-# 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")
+TIKTOKEN_CACHE_DIR = os.environ.get("TIKTOKEN_CACHE_DIR", f"{CACHE_DIR}/tiktoken")
+TIKTOKEN_ENCODING_NAME = PersistentConfig(
+    "TIKTOKEN_ENCODING_NAME",
+    "rag.tiktoken_encoding_name",
+    os.environ.get("TIKTOKEN_ENCODING_NAME", "cl100k_base"),
+)

-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 +1090,25 @@ 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",
@@ -1256,17 +1241,6 @@ RAG_WEB_SEARCH_CONCURRENT_REQUESTS = PersistentConfig(
 )


-####################################
-# Transcribe
-####################################
-
-WHISPER_MODEL = os.getenv("WHISPER_MODEL", "base")
-WHISPER_MODEL_DIR = os.getenv("WHISPER_MODEL_DIR", f"{CACHE_DIR}/whisper/models")
-WHISPER_MODEL_AUTO_UPDATE = (
-    os.environ.get("WHISPER_MODEL_AUTO_UPDATE", "").lower() == "true"
-)
-
-
 ####################################
 # Images
 ####################################
@@ -1293,6 +1267,37 @@ 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",
@@ -1451,6 +1456,19 @@ IMAGE_GENERATION_MODEL = PersistentConfig(
 # Audio
 ####################################

+# Transcription
+WHISPER_MODEL = PersistentConfig(
+    "WHISPER_MODEL",
+    "audio.stt.whisper_model",
+    os.getenv("WHISPER_MODEL", "base"),
+)
+
+WHISPER_MODEL_DIR = os.getenv("WHISPER_MODEL_DIR", f"{CACHE_DIR}/whisper/models")
+WHISPER_MODEL_AUTO_UPDATE = (
+    os.environ.get("WHISPER_MODEL_AUTO_UPDATE", "").lower() == "true"
+)
+
+
 AUDIO_STT_OPENAI_API_BASE_URL = PersistentConfig(
    "AUDIO_STT_OPENAI_API_BASE_URL",
    "audio.stt.openai.api_base_url",
@@ -1472,7 +1490,7 @@ AUDIO_STT_ENGINE = PersistentConfig(
 AUDIO_STT_MODEL = PersistentConfig(
    "AUDIO_STT_MODEL",
    "audio.stt.model",
-    os.getenv("AUDIO_STT_MODEL", "whisper-1"),
+    os.getenv("AUDIO_STT_MODEL", ""),
 )

 AUDIO_TTS_OPENAI_API_BASE_URL = PersistentConfig(
@@ -1516,3 +1534,17 @@ 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"
+    ),
+)
@@ -20,7 +20,9 @@ class ERROR_MESSAGES(str, Enum):
    def __str__(self) -> str:
        return super().__str__()

-    DEFAULT = lambda err="": f"Something went wrong :/\n{err if err else ''}"
+    DEFAULT = (
+        lambda err="": f'{"Something went wrong :/" if err == "" else "[ERROR: " + str(err) + "]"}'
+    )
    ENV_VAR_NOT_FOUND = "Required environment variable not found. Terminating now."
    CREATE_USER_ERROR = "Oops! Something went wrong while creating your account. Please try again later. If the issue persists, contact support for assistance."
    DELETE_USER_ERROR = "Oops! Something went wrong. We encountered an issue while trying to delete the user. Please give it another shot."
@@ -34,8 +36,8 @@ class ERROR_MESSAGES(str, Enum):

    ID_TAKEN = "Uh-oh! This id is already registered. Please choose another id string."
    MODEL_ID_TAKEN = "Uh-oh! This model id is already registered. Please choose another model id string."
-
    NAME_TAG_TAKEN = "Uh-oh! This name tag is already registered. Please choose another name tag string."
+
    INVALID_TOKEN = (
        "Your session has expired or the token is invalid. Please sign in again."
    )
@@ -90,6 +92,15 @@ class ERROR_MESSAGES(str, Enum):
        "The Ollama API is disabled. Please enable it to use this feature."
    )

+    FILE_TOO_LARGE = (
+        lambda size="": f"Oops! The file you're trying to upload is too large. Please upload a file that is less than {size}."
+    )
+
+    DUPLICATE_CONTENT = (
+        "Duplicate content detected. Please provide unique content to proceed."
+    )
+    FILE_NOT_PROCESSED = "Extracted content is not available for this file. Please ensure that the file is processed before proceeding."
+

 class TASKS(str, Enum):
    def __str__(self) -> str:
@@ -97,6 +108,7 @@ class TASKS(str, Enum):

    DEFAULT = lambda task="": f"{task if task else 'generation'}"
    TITLE_GENERATION = "title_generation"
+    TAGS_GENERATION = "tags_generation"
    EMOJI_GENERATION = "emoji_generation"
    QUERY_GENERATION = "query_generation"
    FUNCTION_CALLING = "function_calling"
@@ -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",
@@ -59,6 +84,7 @@ log_sources = [
    "OPENAI",
    "RAG",
    "WEBHOOK",
+    "SOCKET",
 ]

 SRC_LOG_LEVELS = {}
@@ -88,15 +114,9 @@ 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 +207,12 @@ 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
@@ -201,28 +221,27 @@ if PIP_INSTALL:
            else:
                shutil.copy2(item, dest)

-    DATA_DIR = OPEN_WEBUI_DIR / "data"
+        # Zip the data directory
+        shutil.make_archive(DATA_DIR.parent / "open_webui_data", "zip", DATA_DIR)

+        # Remove the old data directory
+        shutil.rmtree(DATA_DIR)
+
+    DATA_DIR = Path(os.getenv("DATA_DIR", OPEN_WEBUI_DIR / "data"))
+
+
+STATIC_DIR = Path(os.getenv("STATIC_DIR", OPEN_WEBUI_DIR / "static"))
+
+FONTS_DIR = Path(os.getenv("FONTS_DIR", OPEN_WEBUI_DIR / "static" / "fonts"))

 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()


-RESET_CONFIG_ON_START = (
-    os.environ.get("RESET_CONFIG_ON_START", "False").lower() == "true"
-)
-
-if RESET_CONFIG_ON_START:
-    try:
-        os.remove(f"{DATA_DIR}/config.json")
-        with open(f"{DATA_DIR}/config.json", "w") as f:
-            f.write("{}")
-    except Exception:
-        pass
-
 ####################################
 # Database
 ####################################
@@ -241,6 +260,55 @@ DATABASE_URL = os.environ.get("DATABASE_URL", f"sqlite:///{DATA_DIR}/webui.db")
 if "postgres://" in DATABASE_URL:
    DATABASE_URL = DATABASE_URL.replace("postgres://", "postgresql://")

+DATABASE_POOL_SIZE = os.environ.get("DATABASE_POOL_SIZE", 0)
+
+if DATABASE_POOL_SIZE == "":
+    DATABASE_POOL_SIZE = 0
+else:
+    try:
+        DATABASE_POOL_SIZE = int(DATABASE_POOL_SIZE)
+    except Exception:
+        DATABASE_POOL_SIZE = 0
+
+DATABASE_POOL_MAX_OVERFLOW = os.environ.get("DATABASE_POOL_MAX_OVERFLOW", 0)
+
+if DATABASE_POOL_MAX_OVERFLOW == "":
+    DATABASE_POOL_MAX_OVERFLOW = 0
+else:
+    try:
+        DATABASE_POOL_MAX_OVERFLOW = int(DATABASE_POOL_MAX_OVERFLOW)
+    except Exception:
+        DATABASE_POOL_MAX_OVERFLOW = 0
+
+DATABASE_POOL_TIMEOUT = os.environ.get("DATABASE_POOL_TIMEOUT", 30)
+
+if DATABASE_POOL_TIMEOUT == "":
+    DATABASE_POOL_TIMEOUT = 30
+else:
+    try:
+        DATABASE_POOL_TIMEOUT = int(DATABASE_POOL_TIMEOUT)
+    except Exception:
+        DATABASE_POOL_TIMEOUT = 30
+
+DATABASE_POOL_RECYCLE = os.environ.get("DATABASE_POOL_RECYCLE", 3600)
+
+if DATABASE_POOL_RECYCLE == "":
+    DATABASE_POOL_RECYCLE = 3600
+else:
+    try:
+        DATABASE_POOL_RECYCLE = int(DATABASE_POOL_RECYCLE)
+    except Exception:
+        DATABASE_POOL_RECYCLE = 3600
+
+RESET_CONFIG_ON_START = (
+    os.environ.get("RESET_CONFIG_ON_START", "False").lower() == "true"
+)
+
+####################################
+# REDIS
+####################################
+
+REDIS_URL = os.environ.get("REDIS_URL", "redis://localhost:6379/0")

 ####################################
 # WEBUI_AUTH (Required for security)
@@ -276,3 +344,41 @@ 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_URL)
+
+AIOHTTP_CLIENT_TIMEOUT = os.environ.get("AIOHTTP_CLIENT_TIMEOUT", "")
+
+if AIOHTTP_CLIENT_TIMEOUT == "":
+    AIOHTTP_CLIENT_TIMEOUT = None
+else:
+    try:
+        AIOHTTP_CLIENT_TIMEOUT = int(AIOHTTP_CLIENT_TIMEOUT)
+    except Exception:
+        AIOHTTP_CLIENT_TIMEOUT = 300
+
+AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST = os.environ.get(
+    "AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST", "3"
+)
+
+if AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST == "":
+    AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST = None
+else:
+    try:
+        AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST = int(
+            AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST
+        )
+    except Exception:
+        AIOHTTP_CLIENT_TIMEOUT_OPENAI_MODEL_LIST = 3
+
+####################################
+# OFFLINE_MODE
+####################################
+
+OFFLINE_MODE = os.environ.get("OFFLINE_MODE", "false").lower() == "true"
@@ -7,3 +7,9 @@ def get_existing_tables():
    inspector = Inspector.from_engine(con)
    tables = set(inspector.get_table_names())
    return tables
+
+
+def get_revision_id():
+    import uuid
+
+    return str(uuid.uuid4()).replace("-", "")[:12]
@@ -0,0 +1,151 @@
+"""Migrate tags
+
+Revision ID: 1af9b942657b
+Revises: 242a2047eae0
+Create Date: 2024-10-09 21:02:35.241684
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.sql import table, select, update, column
+from sqlalchemy.engine.reflection import Inspector
+
+import json
+
+revision = "1af9b942657b"
+down_revision = "242a2047eae0"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # Setup an inspection on the existing table to avoid issues
+    conn = op.get_bind()
+    inspector = Inspector.from_engine(conn)
+
+    # Clean up potential leftover temp table from previous failures
+    conn.execute(sa.text("DROP TABLE IF EXISTS _alembic_tmp_tag"))
+
+    # Check if the 'tag' table exists
+    tables = inspector.get_table_names()
+
+    # Step 1: Modify Tag table using batch mode for SQLite support
+    if "tag" in tables:
+        # Get the current columns in the 'tag' table
+        columns = [col["name"] for col in inspector.get_columns("tag")]
+
+        # Get any existing unique constraints on the 'tag' table
+        current_constraints = inspector.get_unique_constraints("tag")
+
+        with op.batch_alter_table("tag", schema=None) as batch_op:
+            # Check if the unique constraint already exists
+            if not any(
+                constraint["name"] == "uq_id_user_id"
+                for constraint in current_constraints
+            ):
+                # Create unique constraint if it doesn't exist
+                batch_op.create_unique_constraint("uq_id_user_id", ["id", "user_id"])
+
+            # Check if the 'data' column exists before trying to drop it
+            if "data" in columns:
+                batch_op.drop_column("data")
+
+            # Check if the 'meta' column needs to be created
+            if "meta" not in columns:
+                # Add the 'meta' column if it doesn't already exist
+                batch_op.add_column(sa.Column("meta", sa.JSON(), nullable=True))
+
+    tag = table(
+        "tag",
+        column("id", sa.String()),
+        column("name", sa.String()),
+        column("user_id", sa.String()),
+        column("meta", sa.JSON()),
+    )
+
+    # Step 2: Migrate tags
+    conn = op.get_bind()
+    result = conn.execute(sa.select(tag.c.id, tag.c.name, tag.c.user_id))
+
+    tag_updates = {}
+    for row in result:
+        new_id = row.name.replace(" ", "_").lower()
+        tag_updates[row.id] = new_id
+
+    for tag_id, new_tag_id in tag_updates.items():
+        print(f"Updating tag {tag_id} to {new_tag_id}")
+        if new_tag_id == "pinned":
+            # delete tag
+            delete_stmt = sa.delete(tag).where(tag.c.id == tag_id)
+            conn.execute(delete_stmt)
+        else:
+            # Check if the new_tag_id already exists in the database
+            existing_tag_query = sa.select(tag.c.id).where(tag.c.id == new_tag_id)
+            existing_tag_result = conn.execute(existing_tag_query).fetchone()
+
+            if existing_tag_result:
+                # Handle duplicate case: the new_tag_id already exists
+                print(
+                    f"Tag {new_tag_id} already exists. Removing current tag with ID {tag_id} to avoid duplicates."
+                )
+                # Option 1: Delete the current tag if an update to new_tag_id would cause duplication
+                delete_stmt = sa.delete(tag).where(tag.c.id == tag_id)
+                conn.execute(delete_stmt)
+            else:
+                update_stmt = sa.update(tag).where(tag.c.id == tag_id)
+                update_stmt = update_stmt.values(id=new_tag_id)
+                conn.execute(update_stmt)
+
+    # Add columns `pinned` and `meta` to 'chat'
+    op.add_column("chat", sa.Column("pinned", sa.Boolean(), nullable=True))
+    op.add_column(
+        "chat", sa.Column("meta", sa.JSON(), nullable=False, server_default="{}")
+    )
+
+    chatidtag = table(
+        "chatidtag", column("chat_id", sa.String()), column("tag_name", sa.String())
+    )
+    chat = table(
+        "chat",
+        column("id", sa.String()),
+        column("pinned", sa.Boolean()),
+        column("meta", sa.JSON()),
+    )
+
+    # Fetch existing tags
+    conn = op.get_bind()
+    result = conn.execute(sa.select(chatidtag.c.chat_id, chatidtag.c.tag_name))
+
+    chat_updates = {}
+    for row in result:
+        chat_id = row.chat_id
+        tag_name = row.tag_name.replace(" ", "_").lower()
+
+        if tag_name == "pinned":
+            # Specifically handle 'pinned' tag
+            if chat_id not in chat_updates:
+                chat_updates[chat_id] = {"pinned": True, "meta": {}}
+            else:
+                chat_updates[chat_id]["pinned"] = True
+        else:
+            if chat_id not in chat_updates:
+                chat_updates[chat_id] = {"pinned": False, "meta": {"tags": [tag_name]}}
+            else:
+                tags = chat_updates[chat_id]["meta"].get("tags", [])
+                tags.append(tag_name)
+
+                chat_updates[chat_id]["meta"]["tags"] = list(set(tags))
+
+    # Update chats based on accumulated changes
+    for chat_id, updates in chat_updates.items():
+        update_stmt = sa.update(chat).where(chat.c.id == chat_id)
+        update_stmt = update_stmt.values(
+            meta=updates.get("meta", {}), pinned=updates.get("pinned", False)
+        )
+        conn.execute(update_stmt)
+    pass
+
+
+def downgrade():
+    pass
@@ -0,0 +1,82 @@
+"""Update chat table
+
+Revision ID: 242a2047eae0
+Revises: 6a39f3d8e55c
+Create Date: 2024-10-09 21:02:35.241684
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.sql import table, select, update
+
+import json
+
+revision = "242a2047eae0"
+down_revision = "6a39f3d8e55c"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # Step 1: Rename current 'chat' column to 'old_chat'
+    op.alter_column("chat", "chat", new_column_name="old_chat", existing_type=sa.Text)
+
+    # Step 2: Add new 'chat' column of type JSON
+    op.add_column("chat", sa.Column("chat", sa.JSON(), nullable=True))
+
+    # Step 3: Migrate data from 'old_chat' to 'chat'
+    chat_table = table(
+        "chat",
+        sa.Column("id", sa.String, primary_key=True),
+        sa.Column("old_chat", sa.Text),
+        sa.Column("chat", sa.JSON()),
+    )
+
+    # - Selecting all data from the table
+    connection = op.get_bind()
+    results = connection.execute(select(chat_table.c.id, chat_table.c.old_chat))
+    for row in results:
+        try:
+            # Convert text JSON to actual JSON object, assuming the text is in JSON format
+            json_data = json.loads(row.old_chat)
+        except json.JSONDecodeError:
+            json_data = None  # Handle cases where the text cannot be converted to JSON
+
+        connection.execute(
+            sa.update(chat_table)
+            .where(chat_table.c.id == row.id)
+            .values(chat=json_data)
+        )
+
+    # Step 4: Drop 'old_chat' column
+    op.drop_column("chat", "old_chat")
+
+
+def downgrade():
+    # Step 1: Add 'old_chat' column back as Text
+    op.add_column("chat", sa.Column("old_chat", sa.Text(), nullable=True))
+
+    # Step 2: Convert 'chat' JSON data back to text and store in 'old_chat'
+    chat_table = table(
+        "chat",
+        sa.Column("id", sa.String, primary_key=True),
+        sa.Column("chat", sa.JSON()),
+        sa.Column("old_chat", sa.Text()),
+    )
+
+    connection = op.get_bind()
+    results = connection.execute(select(chat_table.c.id, chat_table.c.chat))
+    for row in results:
+        text_data = json.dumps(row.chat) if row.chat is not None else None
+        connection.execute(
+            sa.update(chat_table)
+            .where(chat_table.c.id == row.id)
+            .values(old_chat=text_data)
+        )
+
+    # Step 3: Remove the new 'chat' JSON column
+    op.drop_column("chat", "chat")
+
+    # Step 4: Rename 'old_chat' back to 'chat'
+    op.alter_column("chat", "old_chat", new_column_name="chat", existing_type=sa.Text)
@@ -0,0 +1,81 @@
+"""Update tags
+
+Revision ID: 3ab32c4b8f59
+Revises: 1af9b942657b
+Create Date: 2024-10-09 21:02:35.241684
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.sql import table, select, update, column
+from sqlalchemy.engine.reflection import Inspector
+
+import json
+
+revision = "3ab32c4b8f59"
+down_revision = "1af9b942657b"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    conn = op.get_bind()
+    inspector = Inspector.from_engine(conn)
+
+    # Inspecting the 'tag' table constraints and structure
+    existing_pk = inspector.get_pk_constraint("tag")
+    unique_constraints = inspector.get_unique_constraints("tag")
+    existing_indexes = inspector.get_indexes("tag")
+
+    print(f"Primary Key: {existing_pk}")
+    print(f"Unique Constraints: {unique_constraints}")
+    print(f"Indexes: {existing_indexes}")
+
+    with op.batch_alter_table("tag", schema=None) as batch_op:
+        # Drop existing primary key constraint if it exists
+        if existing_pk and existing_pk.get("constrained_columns"):
+            pk_name = existing_pk.get("name")
+            if pk_name:
+                print(f"Dropping primary key constraint: {pk_name}")
+                batch_op.drop_constraint(pk_name, type_="primary")
+
+        # Now create the new primary key with the combination of 'id' and 'user_id'
+        print("Creating new primary key with 'id' and 'user_id'.")
+        batch_op.create_primary_key("pk_id_user_id", ["id", "user_id"])
+
+        # Drop unique constraints that could conflict with the new primary key
+        for constraint in unique_constraints:
+            if (
+                constraint["name"] == "uq_id_user_id"
+            ):  # Adjust this name according to what is actually returned by the inspector
+                print(f"Dropping unique constraint: {constraint['name']}")
+                batch_op.drop_constraint(constraint["name"], type_="unique")
+
+        for index in existing_indexes:
+            if index["unique"]:
+                if not any(
+                    constraint["name"] == index["name"]
+                    for constraint in unique_constraints
+                ):
+                    # You are attempting to drop unique indexes
+                    print(f"Dropping unique index: {index['name']}")
+                    batch_op.drop_index(index["name"])
+
+
+def downgrade():
+    conn = op.get_bind()
+    inspector = Inspector.from_engine(conn)
+
+    current_pk = inspector.get_pk_constraint("tag")
+
+    with op.batch_alter_table("tag", schema=None) as batch_op:
+        # Drop the current primary key first, if it matches the one we know we added in upgrade
+        if current_pk and "pk_id_user_id" == current_pk.get("name"):
+            batch_op.drop_constraint("pk_id_user_id", type_="primary")
+
+        # Restore the original primary key
+        batch_op.create_primary_key("pk_id", ["id"])
+
+        # Since primary key on just 'id' is restored, we now add back any unique constraints if necessary
+        batch_op.create_unique_constraint("uq_id_user_id", ["id", "user_id"])
@@ -0,0 +1,67 @@
+"""Update folder table and change DateTime to BigInteger for timestamp fields
+
+Revision ID: 4ace53fd72c8
+Revises: af906e964978
+Create Date: 2024-10-23 03:00:00.000000
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+
+revision = "4ace53fd72c8"
+down_revision = "af906e964978"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # Perform safe alterations using batch operation
+    with op.batch_alter_table("folder", schema=None) as batch_op:
+        # Step 1: Remove server defaults for created_at and updated_at
+        batch_op.alter_column(
+            "created_at",
+            server_default=None,  # Removing server default
+        )
+        batch_op.alter_column(
+            "updated_at",
+            server_default=None,  # Removing server default
+        )
+
+        # Step 2: Change the column types to BigInteger for created_at
+        batch_op.alter_column(
+            "created_at",
+            type_=sa.BigInteger(),
+            existing_type=sa.DateTime(),
+            existing_nullable=False,
+            postgresql_using="extract(epoch from created_at)::bigint",  # Conversion for PostgreSQL
+        )
+
+        # Change the column types to BigInteger for updated_at
+        batch_op.alter_column(
+            "updated_at",
+            type_=sa.BigInteger(),
+            existing_type=sa.DateTime(),
+            existing_nullable=False,
+            postgresql_using="extract(epoch from updated_at)::bigint",  # Conversion for PostgreSQL
+        )
+
+
+def downgrade():
+    # Downgrade: Convert columns back to DateTime and restore defaults
+    with op.batch_alter_table("folder", schema=None) as batch_op:
+        batch_op.alter_column(
+            "created_at",
+            type_=sa.DateTime(),
+            existing_type=sa.BigInteger(),
+            existing_nullable=False,
+            server_default=sa.func.now(),  # Restoring server default on downgrade
+        )
+        batch_op.alter_column(
+            "updated_at",
+            type_=sa.DateTime(),
+            existing_type=sa.BigInteger(),
+            existing_nullable=False,
+            server_default=sa.func.now(),  # Restoring server default on downgrade
+            onupdate=sa.func.now(),  # Restoring onupdate behavior if it was there
+        )
@@ -0,0 +1,80 @@
+"""Add knowledge table
+
+Revision ID: 6a39f3d8e55c
+Revises: c0fbf31ca0db
+Create Date: 2024-10-01 14:02:35.241684
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy.sql import table, column, select
+import json
+
+
+revision = "6a39f3d8e55c"
+down_revision = "c0fbf31ca0db"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # Creating the 'knowledge' table
+    print("Creating knowledge table")
+    knowledge_table = op.create_table(
+        "knowledge",
+        sa.Column("id", sa.Text(), primary_key=True),
+        sa.Column("user_id", sa.Text(), nullable=False),
+        sa.Column("name", sa.Text(), nullable=False),
+        sa.Column("description", sa.Text(), nullable=True),
+        sa.Column("data", sa.JSON(), nullable=True),
+        sa.Column("meta", sa.JSON(), nullable=True),
+        sa.Column("created_at", sa.BigInteger(), nullable=False),
+        sa.Column("updated_at", sa.BigInteger(), nullable=True),
+    )
+
+    print("Migrating data from document table to knowledge table")
+    # Representation of the existing 'document' table
+    document_table = table(
+        "document",
+        column("collection_name", sa.String()),
+        column("user_id", sa.String()),
+        column("name", sa.String()),
+        column("title", sa.Text()),
+        column("content", sa.Text()),
+        column("timestamp", sa.BigInteger()),
+    )
+
+    # Select all from existing document table
+    documents = op.get_bind().execute(
+        select(
+            document_table.c.collection_name,
+            document_table.c.user_id,
+            document_table.c.name,
+            document_table.c.title,
+            document_table.c.content,
+            document_table.c.timestamp,
+        )
+    )
+
+    # Insert data into knowledge table from document table
+    for doc in documents:
+        op.get_bind().execute(
+            knowledge_table.insert().values(
+                id=doc.collection_name,
+                user_id=doc.user_id,
+                description=doc.name,
+                meta={
+                    "legacy": True,
+                    "document": True,
+                    "tags": json.loads(doc.content or "{}").get("tags", []),
+                },
+                name=doc.title,
+                created_at=doc.timestamp,
+                updated_at=doc.timestamp,  # using created_at for both created_at and updated_at in project
+            )
+        )
+
+
+def downgrade():
+    op.drop_table("knowledge")
@@ -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"),
@@ -0,0 +1,51 @@
+"""Add feedback table
+
+Revision ID: af906e964978
+Revises: c29facfe716b
+Create Date: 2024-10-20 17:02:35.241684
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+
+# Revision identifiers, used by Alembic.
+revision = "af906e964978"
+down_revision = "c29facfe716b"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # ### Create feedback table ###
+    op.create_table(
+        "feedback",
+        sa.Column(
+            "id", sa.Text(), primary_key=True
+        ),  # Unique identifier for each feedback (TEXT type)
+        sa.Column(
+            "user_id", sa.Text(), nullable=True
+        ),  # ID of the user providing the feedback (TEXT type)
+        sa.Column(
+            "version", sa.BigInteger(), default=0
+        ),  # Version of feedback (BIGINT type)
+        sa.Column("type", sa.Text(), nullable=True),  # Type of feedback (TEXT type)
+        sa.Column("data", sa.JSON(), nullable=True),  # Feedback data (JSON type)
+        sa.Column(
+            "meta", sa.JSON(), nullable=True
+        ),  # Metadata for feedback (JSON type)
+        sa.Column(
+            "snapshot", sa.JSON(), nullable=True
+        ),  # snapshot data for feedback (JSON type)
+        sa.Column(
+            "created_at", sa.BigInteger(), nullable=False
+        ),  # Feedback creation timestamp (BIGINT representing epoch)
+        sa.Column(
+            "updated_at", sa.BigInteger(), nullable=False
+        ),  # Feedback update timestamp (BIGINT representing epoch)
+    )
+
+
+def downgrade():
+    # ### Drop feedback table ###
+    op.drop_table("feedback")
@@ -0,0 +1,32 @@
+"""Update file table
+
+Revision ID: c0fbf31ca0db
+Revises: ca81bd47c050
+Create Date: 2024-09-20 15:26:35.241684
+
+"""
+
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "c0fbf31ca0db"
+down_revision: Union[str, None] = "ca81bd47c050"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.add_column("file", sa.Column("hash", sa.Text(), nullable=True))
+    op.add_column("file", sa.Column("data", sa.JSON(), nullable=True))
+    op.add_column("file", sa.Column("updated_at", sa.BigInteger(), nullable=True))
+
+
+def downgrade():
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_column("file", "updated_at")
+    op.drop_column("file", "data")
+    op.drop_column("file", "hash")
@@ -0,0 +1,79 @@
+"""Update file table path
+
+Revision ID: c29facfe716b
+Revises: c69f45358db4
+Create Date: 2024-10-20 17:02:35.241684
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+import json
+from sqlalchemy.sql import table, column
+from sqlalchemy import String, Text, JSON, and_
+
+
+revision = "c29facfe716b"
+down_revision = "c69f45358db4"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # 1. Add the `path` column to the "file" table.
+    op.add_column("file", sa.Column("path", sa.Text(), nullable=True))
+
+    # 2. Convert the `meta` column from Text/JSONField to `JSON()`
+    # Use Alembic's default batch_op for dialect compatibility.
+    with op.batch_alter_table("file", schema=None) as batch_op:
+        batch_op.alter_column(
+            "meta",
+            type_=sa.JSON(),
+            existing_type=sa.Text(),
+            existing_nullable=True,
+            nullable=True,
+            postgresql_using="meta::json",
+        )
+
+    # 3. Migrate legacy data from `meta` JSONField
+    # Fetch and process `meta` data from the table, add values to the new `path` column as necessary.
+    # We will use SQLAlchemy core bindings to ensure safety across different databases.
+
+    file_table = table(
+        "file", column("id", String), column("meta", JSON), column("path", Text)
+    )
+
+    # Create connection to the database
+    connection = op.get_bind()
+
+    # Get the rows where `meta` has a path and `path` column is null (new column)
+    # Loop through each row in the result set to update the path
+    results = connection.execute(
+        sa.select(file_table.c.id, file_table.c.meta).where(
+            and_(file_table.c.path.is_(None), file_table.c.meta.isnot(None))
+        )
+    ).fetchall()
+
+    # Iterate over each row to extract and update the `path` from `meta` column
+    for row in results:
+        if "path" in row.meta:
+            # Extract the `path` field from the `meta` JSON
+            path = row.meta.get("path")
+
+            # Update the `file` table with the new `path` value
+            connection.execute(
+                file_table.update()
+                .where(file_table.c.id == row.id)
+                .values({"path": path})
+            )
+
+
+def downgrade():
+    # 1. Remove the `path` column
+    op.drop_column("file", "path")
+
+    # 2. Revert the `meta` column back to Text/JSONField
+    with op.batch_alter_table("file", schema=None) as batch_op:
+        batch_op.alter_column(
+            "meta", type_=sa.Text(), existing_type=sa.JSON(), existing_nullable=True
+        )
@@ -0,0 +1,50 @@
+"""Add folder table
+
+Revision ID: c69f45358db4
+Revises: 3ab32c4b8f59
+Create Date: 2024-10-16 02:02:35.241684
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+
+revision = "c69f45358db4"
+down_revision = "3ab32c4b8f59"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    op.create_table(
+        "folder",
+        sa.Column("id", sa.Text(), nullable=False),
+        sa.Column("parent_id", sa.Text(), nullable=True),
+        sa.Column("user_id", sa.Text(), nullable=False),
+        sa.Column("name", sa.Text(), nullable=False),
+        sa.Column("items", sa.JSON(), nullable=True),
+        sa.Column("meta", sa.JSON(), nullable=True),
+        sa.Column("is_expanded", sa.Boolean(), default=False, nullable=False),
+        sa.Column(
+            "created_at", sa.DateTime(), server_default=sa.func.now(), nullable=False
+        ),
+        sa.Column(
+            "updated_at",
+            sa.DateTime(),
+            nullable=False,
+            server_default=sa.func.now(),
+            onupdate=sa.func.now(),
+        ),
+        sa.PrimaryKeyConstraint("id", "user_id"),
+    )
+
+    op.add_column(
+        "chat",
+        sa.Column("folder_id", sa.Text(), nullable=True),
+    )
+
+
+def downgrade():
+    op.drop_column("chat", "folder_id")
+
+    op.drop_table("folder")
@@ -0,0 +1,319 @@
+/* HTML and Body */
+@font-face {
+	font-family: 'NotoSans';
+	src: url('fonts/NotoSans-Variable.ttf');
+}
+
+@font-face {
+	font-family: 'NotoSansJP';
+	src: url('fonts/NotoSansJP-Variable.ttf');
+}
+
+@font-face {
+	font-family: 'NotoSansKR';
+	src: url('fonts/NotoSansKR-Variable.ttf');
+}
+
+@font-face {
+	font-family: 'NotoSansSC';
+	src: url('fonts/NotoSansSC-Variable.ttf');
+}
+
+@font-face {
+	font-family: 'NotoSansSC-Regular';
+	src: url('fonts/NotoSansSC-Regular.ttf');
+}
+
+html {
+	font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', 'NotoSans', 'NotoSansJP', 'NotoSansKR',
+		'NotoSansSC', 'STSong-Light', 'MSung-Light', 'HeiseiMin-W3', 'HYSMyeongJo-Medium', Roboto,
+		'Helvetica Neue', Arial, sans-serif;
+	font-size: 14px; /* Default font size */
+	line-height: 1.5;
+}
+
+*,
+*::before,
+*::after {
+	box-sizing: inherit;
+}
+
+body {
+	margin: 0;
+	color: #212529;
+	background-color: #fff;
+	width: auto;
+}
+
+/* Typography */
+h1,
+h2,
+h3,
+h4,
+h5,
+h6 {
+	font-weight: 500;
+	margin: 0;
+}
+
+h1 {
+	font-size: 2.5rem;
+}
+
+h2 {
+	font-size: 2rem;
+}
+
+h3 {
+	font-size: 1.75rem;
+}
+
+h4 {
+	font-size: 1.5rem;
+}
+
+h5 {
+	font-size: 1.25rem;
+}
+
+h6 {
+	font-size: 1rem;
+}
+
+p {
+	margin-top: 0;
+	margin-bottom: 1rem;
+}
+
+/* Grid System */
+.container {
+	width: 100%;
+	padding-right: 15px;
+	padding-left: 15px;
+	margin-right: auto;
+	margin-left: auto;
+}
+
+/* Utilities */
+.text-center {
+	text-align: center;
+}
+
+/* Additional Text Utilities */
+.text-muted {
+	color: #6c757d; /* Muted text color */
+}
+
+/* Small Text */
+small {
+	font-size: 80%; /* Smaller font size relative to the base */
+	color: #6c757d; /* Lighter text color for secondary information */
+	margin-bottom: 0;
+	margin-top: 0;
+}
+
+/* Strong Element Styles */
+strong {
+	font-weight: bolder; /* Ensures the text is bold */
+	color: inherit; /* Inherits the color from its parent element */
+}
+
+/* link */
+a {
+	color: #007bff;
+	text-decoration: none;
+	background-color: transparent;
+}
+
+a:hover {
+	color: #0056b3;
+	text-decoration: underline;
+}
+
+/* General styles for lists */
+ol,
+ul,
+li {
+	padding-left: 40px; /* Increase padding to move bullet points to the right */
+	margin-left: 20px; /* Indent lists from the left */
+}
+
+/* Ordered list styles */
+ol {
+	list-style-type: decimal; /* Use numbers for ordered lists */
+	margin-bottom: 10px; /* Space after each list */
+}
+
+ol li {
+	margin-bottom: 0.5rem; /* Space between ordered list items */
+}
+
+/* Unordered list styles */
+ul {
+	list-style-type: disc; /* Use bullets for unordered lists */
+	margin-bottom: 10px; /* Space after each list */
+}
+
+ul li {
+	margin-bottom: 0.5rem; /* Space between unordered list items */
+}
+
+/* List item styles */
+li {
+	margin-bottom: 5px; /* Space between list items */
+	line-height: 1.5; /* Line height for better readability */
+}
+
+/* Nested lists */
+ol ol,
+ol ul,
+ul ol,
+ul ul {
+	padding-left: 20px;
+	margin-left: 30px; /* Further indent nested lists */
+	margin-bottom: 0; /* Remove extra margin at the bottom of nested lists */
+}
+
+/* Code blocks */
+pre {
+	background-color: #f4f4f4;
+	padding: 10px;
+	overflow-x: auto;
+	max-width: 100%; /* Ensure it doesn't overflow the page */
+	width: 80%; /* Set a specific width for a container-like appearance */
+	margin: 0 1em; /* Center the pre block */
+	box-sizing: border-box; /* Include padding in the width */
+	border: 1px solid #ccc; /* Optional: Add a border for better definition */
+	border-radius: 4px; /* Optional: Add rounded corners */
+}
+
+code {
+	font-family: 'Courier New', Courier, monospace;
+	background-color: #f4f4f4;
+	padding: 2px 4px;
+	border-radius: 4px;
+	box-sizing: border-box; /* Include padding in the width */
+}
+
+.message {
+	margin-top: 8px;
+	margin-bottom: 8px;
+	max-width: 100%;
+	overflow-wrap: break-word;
+}
+
+/* Table Styles */
+table {
+	width: 100%;
+	margin-bottom: 1rem;
+	color: #212529;
+	border-collapse: collapse; /* Removes the space between borders */
+}
+
+th,
+td {
+	margin: 0;
+	padding: 0.75rem;
+	vertical-align: top;
+	border-top: 1px solid #dee2e6;
+}
+
+thead th {
+	vertical-align: bottom;
+	border-bottom: 2px solid #dee2e6;
+}
+
+tbody + tbody {
+	border-top: 2px solid #dee2e6;
+}
+
+/* markdown-section styles */
+.markdown-section blockquote,
+.markdown-section h1,
+.markdown-section h2,
+.markdown-section h3,
+.markdown-section h4,
+.markdown-section h5,
+.markdown-section h6,
+.markdown-section p,
+.markdown-section pre,
+.markdown-section table,
+.markdown-section ul {
+	/* Give most block elements margin top and bottom */
+	margin-top: 1rem;
+}
+
+/* Remove top margin if it's the first child */
+.markdown-section blockquote:first-child,
+.markdown-section h1:first-child,
+.markdown-section h2:first-child,
+.markdown-section h3:first-child,
+.markdown-section h4:first-child,
+.markdown-section h5:first-child,
+.markdown-section h6:first-child,
+.markdown-section p:first-child,
+.markdown-section pre:first-child,
+.markdown-section table:first-child,
+.markdown-section ul:first-child {
+	margin-top: 0;
+}
+
+/* Remove top margin of <ul> following a <p> */
+.markdown-section p + ul {
+	margin-top: 0;
+}
+
+/* Remove bottom margin of <p> if it is followed by a <ul> */
+/* Note: :has is not supported in CSS, so you would need JavaScript for this behavior */
+.markdown-section p {
+	margin-bottom: 0;
+}
+
+/* Add a rule to reset margin-bottom for <p> not followed by <ul> */
+.markdown-section p + ul {
+	margin-top: 0;
+}
+
+/* List item styles */
+.markdown-section li {
+	padding: 2px;
+}
+
+.markdown-section li p {
+	margin-bottom: 0;
+	padding: 0;
+}
+
+/* Avoid margins for nested lists */
+.markdown-section li > ul {
+	margin-top: 0;
+	margin-bottom: 0;
+}
+
+/* Table styles */
+.markdown-section table {
+	width: 100%;
+	border-collapse: collapse;
+	margin: 1rem 0;
+}
+
+.markdown-section th,
+.markdown-section td {
+	border: 1px solid #ddd;
+	padding: 0.5rem;
+	text-align: left;
+}
+
+.markdown-section th {
+	background-color: #f2f2f2;
+}
+
+.markdown-section pre {
+	padding: 10px;
+	margin: 10px;
+}
+
+.markdown-section pre code {
+	position: relative;
+	color: rgb(172, 0, 95);
+}
@@ -0,0 +1,164 @@
+import os
+import boto3
+from botocore.exceptions import ClientError
+import shutil
+
+
+from typing import BinaryIO, Tuple, Optional, Union
+
+from open_webui.constants import ERROR_MESSAGES
+from open_webui.config import (
+    STORAGE_PROVIDER,
+    S3_ACCESS_KEY_ID,
+    S3_SECRET_ACCESS_KEY,
+    S3_BUCKET_NAME,
+    S3_REGION_NAME,
+    S3_ENDPOINT_URL,
+    UPLOAD_DIR,
+)
+
+
+import boto3
+from botocore.exceptions import ClientError
+from typing import BinaryIO, Tuple, Optional
+
+
+class StorageProvider:
+    def __init__(self, provider: Optional[str] = None):
+        self.storage_provider: str = provider or STORAGE_PROVIDER
+
+        self.s3_client = None
+        self.s3_bucket_name: Optional[str] = None
+
+        if self.storage_provider == "s3":
+            self._initialize_s3()
+
+    def _initialize_s3(self) -> None:
+        """Initializes the S3 client and bucket name if using S3 storage."""
+        self.s3_client = boto3.client(
+            "s3",
+            region_name=S3_REGION_NAME,
+            endpoint_url=S3_ENDPOINT_URL,
+            aws_access_key_id=S3_ACCESS_KEY_ID,
+            aws_secret_access_key=S3_SECRET_ACCESS_KEY,
+        )
+        self.bucket_name = S3_BUCKET_NAME
+
+    def _upload_to_s3(self, file_path: str, filename: str) -> Tuple[bytes, str]:
+        """Handles uploading of the file to S3 storage."""
+        if not self.s3_client:
+            raise RuntimeError("S3 Client is not initialized.")
+
+        try:
+            self.s3_client.upload_file(file_path, self.bucket_name, filename)
+            return open(file_path, "rb").read(), file_path
+        except ClientError as e:
+            raise RuntimeError(f"Error uploading file to S3: {e}")
+
+    def _upload_to_local(self, contents: bytes, filename: str) -> Tuple[bytes, str]:
+        """Handles uploading of the file to local storage."""
+        file_path = f"{UPLOAD_DIR}/{filename}"
+        with open(file_path, "wb") as f:
+            f.write(contents)
+        return contents, file_path
+
+    def _get_file_from_s3(self, file_path: str) -> str:
+        """Handles downloading of the file from S3 storage."""
+        if not self.s3_client:
+            raise RuntimeError("S3 Client is not initialized.")
+
+        try:
+            bucket_name, key = file_path.split("//")[1].split("/")
+            local_file_path = f"{UPLOAD_DIR}/{key}"
+            self.s3_client.download_file(bucket_name, key, local_file_path)
+            return local_file_path
+        except ClientError as e:
+            raise RuntimeError(f"Error downloading file from S3: {e}")
+
+    def _get_file_from_local(self, file_path: str) -> str:
+        """Handles downloading of the file from local storage."""
+        return file_path
+
+    def _delete_from_s3(self, filename: str) -> None:
+        """Handles deletion of the file from S3 storage."""
+        if not self.s3_client:
+            raise RuntimeError("S3 Client is not initialized.")
+
+        try:
+            self.s3_client.delete_object(Bucket=self.bucket_name, Key=filename)
+        except ClientError as e:
+            raise RuntimeError(f"Error deleting file from S3: {e}")
+
+    def _delete_from_local(self, filename: str) -> None:
+        """Handles deletion of the file from local storage."""
+        file_path = f"{UPLOAD_DIR}/{filename}"
+        if os.path.isfile(file_path):
+            os.remove(file_path)
+        else:
+            print(f"File {file_path} not found in local storage.")
+
+    def _delete_all_from_s3(self) -> None:
+        """Handles deletion of all files from S3 storage."""
+        if not self.s3_client:
+            raise RuntimeError("S3 Client is not initialized.")
+
+        try:
+            response = self.s3_client.list_objects_v2(Bucket=self.bucket_name)
+            if "Contents" in response:
+                for content in response["Contents"]:
+                    self.s3_client.delete_object(
+                        Bucket=self.bucket_name, Key=content["Key"]
+                    )
+        except ClientError as e:
+            raise RuntimeError(f"Error deleting all files from S3: {e}")
+
+    def _delete_all_from_local(self) -> None:
+        """Handles deletion of all files from local storage."""
+        if os.path.exists(UPLOAD_DIR):
+            for filename in os.listdir(UPLOAD_DIR):
+                file_path = os.path.join(UPLOAD_DIR, filename)
+                try:
+                    if os.path.isfile(file_path) or os.path.islink(file_path):
+                        os.unlink(file_path)  # Remove the file or link
+                    elif os.path.isdir(file_path):
+                        shutil.rmtree(file_path)  # Remove the directory
+                except Exception as e:
+                    print(f"Failed to delete {file_path}. Reason: {e}")
+        else:
+            print(f"Directory {UPLOAD_DIR} not found in local storage.")
+
+    def upload_file(self, file: BinaryIO, filename: str) -> Tuple[bytes, str]:
+        """Uploads a file either to S3 or the local file system."""
+        contents = file.read()
+        if not contents:
+            raise ValueError(ERROR_MESSAGES.EMPTY_CONTENT)
+        contents, file_path = self._upload_to_local(contents, filename)
+
+        if self.storage_provider == "s3":
+            return self._upload_to_s3(file_path, filename)
+        return contents, file_path
+
+    def get_file(self, file_path: str) -> str:
+        """Downloads a file either from S3 or the local file system and returns the file path."""
+        if self.storage_provider == "s3":
+            return self._get_file_from_s3(file_path)
+        return self._get_file_from_local(file_path)
+
+    def delete_file(self, filename: str) -> None:
+        """Deletes a file either from S3 or the local file system."""
+        if self.storage_provider == "s3":
+            self._delete_from_s3(filename)
+
+        # Always delete from local storage
+        self._delete_from_local(filename)
+
+    def delete_all_files(self) -> None:
+        """Deletes all files from the storage."""
+        if self.storage_provider == "s3":
+            self._delete_all_from_s3()
+
+        # Always delete from local storage
+        self._delete_all_from_local()
+
+
+Storage = StorageProvider(provider=STORAGE_PROVIDER)
@@ -1,105 +0,0 @@
-from test.util.abstract_integration_test import AbstractPostgresTest
-from test.util.mock_user import mock_webui_user
-
-
-class TestDocuments(AbstractPostgresTest):
-    BASE_PATH = "/api/v1/documents"
-
-    def setup_class(cls):
-        super().setup_class()
-        from open_webui.apps.webui.models.documents import Documents
-
-        cls.documents = Documents
-
-    def test_documents(self):
-        # Empty database
-        assert len(self.documents.get_docs()) == 0
-        with mock_webui_user(id="2"):
-            response = self.fast_api_client.get(self.create_url("/"))
-        assert response.status_code == 200
-        assert len(response.json()) == 0
-
-        # Create a new document
-        with mock_webui_user(id="2"):
-            response = self.fast_api_client.post(
-                self.create_url("/create"),
-                json={
-                    "name": "doc_name",
-                    "title": "doc title",
-                    "collection_name": "custom collection",
-                    "filename": "doc_name.pdf",
-                    "content": "",
-                },
-            )
-        assert response.status_code == 200
-        assert response.json()["name"] == "doc_name"
-        assert len(self.documents.get_docs()) == 1
-
-        # Get the document
-        with mock_webui_user(id="2"):
-            response = self.fast_api_client.get(self.create_url("/doc?name=doc_name"))
-        assert response.status_code == 200
-        data = response.json()
-        assert data["collection_name"] == "custom collection"
-        assert data["name"] == "doc_name"
-        assert data["title"] == "doc title"
-        assert data["filename"] == "doc_name.pdf"
-        assert data["content"] == {}
-
-        # Create another document
-        with mock_webui_user(id="2"):
-            response = self.fast_api_client.post(
-                self.create_url("/create"),
-                json={
-                    "name": "doc_name 2",
-                    "title": "doc title 2",
-                    "collection_name": "custom collection 2",
-                    "filename": "doc_name2.pdf",
-                    "content": "",
-                },
-            )
-        assert response.status_code == 200
-        assert response.json()["name"] == "doc_name 2"
-        assert len(self.documents.get_docs()) == 2
-
-        # Get all documents
-        with mock_webui_user(id="2"):
-            response = self.fast_api_client.get(self.create_url("/"))
-        assert response.status_code == 200
-        assert len(response.json()) == 2
-
-        # Update the first document
-        with mock_webui_user(id="2"):
-            response = self.fast_api_client.post(
-                self.create_url("/doc/update?name=doc_name"),
-                json={"name": "doc_name rework", "title": "updated title"},
-            )
-        assert response.status_code == 200
-        data = response.json()
-        assert data["name"] == "doc_name rework"
-        assert data["title"] == "updated title"
-
-        # Tag the first document
-        with mock_webui_user(id="2"):
-            response = self.fast_api_client.post(
-                self.create_url("/doc/tags"),
-                json={
-                    "name": "doc_name rework",
-                    "tags": [{"name": "testing-tag"}, {"name": "another-tag"}],
-                },
-            )
-        assert response.status_code == 200
-        data = response.json()
-        assert data["name"] == "doc_name rework"
-        assert data["content"] == {
-            "tags": [{"name": "testing-tag"}, {"name": "another-tag"}]
-        }
-        assert len(self.documents.get_docs()) == 2
-
-        # Delete the first document
-        with mock_webui_user(id="2"):
-            response = self.fast_api_client.delete(
-                self.create_url("/doc/delete?name=doc_name rework")
-            )
-        assert response.status_code == 200
-        assert len(self.documents.get_docs()) == 1
@@ -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 is not None:
+        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
@@ -0,0 +1,261 @@
+import base64
+import logging
+import mimetypes
+import uuid
+
+import aiohttp
+from authlib.integrations.starlette_client import OAuth
+from authlib.oidc.core import UserInfo
+from fastapi import (
+    HTTPException,
+    status,
+)
+from starlette.responses import RedirectResponse
+
+from open_webui.apps.webui.models.auths import Auths
+from open_webui.apps.webui.models.users import Users
+from open_webui.config import (
+    DEFAULT_USER_ROLE,
+    ENABLE_OAUTH_SIGNUP,
+    OAUTH_MERGE_ACCOUNTS_BY_EMAIL,
+    OAUTH_PROVIDERS,
+    ENABLE_OAUTH_ROLE_MANAGEMENT,
+    OAUTH_ROLES_CLAIM,
+    OAUTH_EMAIL_CLAIM,
+    OAUTH_PICTURE_CLAIM,
+    OAUTH_USERNAME_CLAIM,
+    OAUTH_ALLOWED_ROLES,
+    OAUTH_ADMIN_ROLES,
+    WEBHOOK_URL,
+    JWT_EXPIRES_IN,
+    AppConfig,
+)
+from open_webui.constants import ERROR_MESSAGES
+from open_webui.env import WEBUI_SESSION_COOKIE_SAME_SITE, WEBUI_SESSION_COOKIE_SECURE
+from open_webui.utils.misc import parse_duration
+from open_webui.utils.utils import get_password_hash, create_token
+from open_webui.utils.webhook import post_webhook
+
+log = logging.getLogger(__name__)
+
+auth_manager_config = AppConfig()
+auth_manager_config.DEFAULT_USER_ROLE = DEFAULT_USER_ROLE
+auth_manager_config.ENABLE_OAUTH_SIGNUP = ENABLE_OAUTH_SIGNUP
+auth_manager_config.OAUTH_MERGE_ACCOUNTS_BY_EMAIL = OAUTH_MERGE_ACCOUNTS_BY_EMAIL
+auth_manager_config.ENABLE_OAUTH_ROLE_MANAGEMENT = ENABLE_OAUTH_ROLE_MANAGEMENT
+auth_manager_config.OAUTH_ROLES_CLAIM = OAUTH_ROLES_CLAIM
+auth_manager_config.OAUTH_EMAIL_CLAIM = OAUTH_EMAIL_CLAIM
+auth_manager_config.OAUTH_PICTURE_CLAIM = OAUTH_PICTURE_CLAIM
+auth_manager_config.OAUTH_USERNAME_CLAIM = OAUTH_USERNAME_CLAIM
+auth_manager_config.OAUTH_ALLOWED_ROLES = OAUTH_ALLOWED_ROLES
+auth_manager_config.OAUTH_ADMIN_ROLES = OAUTH_ADMIN_ROLES
+auth_manager_config.WEBHOOK_URL = WEBHOOK_URL
+auth_manager_config.JWT_EXPIRES_IN = JWT_EXPIRES_IN
+
+
+class OAuthManager:
+    def __init__(self):
+        self.oauth = OAuth()
+        for provider_name, provider_config in OAUTH_PROVIDERS.items():
+            self.oauth.register(
+                name=provider_name,
+                client_id=provider_config["client_id"],
+                client_secret=provider_config["client_secret"],
+                server_metadata_url=provider_config["server_metadata_url"],
+                client_kwargs={
+                    "scope": provider_config["scope"],
+                },
+                redirect_uri=provider_config["redirect_uri"],
+            )
+
+    def get_client(self, provider_name):
+        return self.oauth.create_client(provider_name)
+
+    def get_user_role(self, user, user_data):
+        if user and Users.get_num_users() == 1:
+            # If the user is the only user, assign the role "admin" - actually repairs role for single user on login
+            return "admin"
+        if not user and Users.get_num_users() == 0:
+            # If there are no users, assign the role "admin", as the first user will be an admin
+            return "admin"
+
+        if auth_manager_config.ENABLE_OAUTH_ROLE_MANAGEMENT:
+            oauth_claim = auth_manager_config.OAUTH_ROLES_CLAIM
+            oauth_allowed_roles = auth_manager_config.OAUTH_ALLOWED_ROLES
+            oauth_admin_roles = auth_manager_config.OAUTH_ADMIN_ROLES
+            oauth_roles = None
+            role = "pending"  # Default/fallback role if no matching roles are found
+
+            # Next block extracts the roles from the user data, accepting nested claims of any depth
+            if oauth_claim and oauth_allowed_roles and oauth_admin_roles:
+                claim_data = user_data
+                nested_claims = oauth_claim.split(".")
+                for nested_claim in nested_claims:
+                    claim_data = claim_data.get(nested_claim, {})
+                oauth_roles = claim_data if isinstance(claim_data, list) else None
+
+            # If any roles are found, check if they match the allowed or admin roles
+            if oauth_roles:
+                # If role management is enabled, and matching roles are provided, use the roles
+                for allowed_role in oauth_allowed_roles:
+                    # If the user has any of the allowed roles, assign the role "user"
+                    if allowed_role in oauth_roles:
+                        role = "user"
+                        break
+                for admin_role in oauth_admin_roles:
+                    # If the user has any of the admin roles, assign the role "admin"
+                    if admin_role in oauth_roles:
+                        role = "admin"
+                        break
+        else:
+            if not user:
+                # If role management is disabled, use the default role for new users
+                role = auth_manager_config.DEFAULT_USER_ROLE
+            else:
+                # If role management is disabled, use the existing role for existing users
+                role = user.role
+
+        return role
+
+    async def handle_login(self, provider, request):
+        if provider not in OAUTH_PROVIDERS:
+            raise HTTPException(404)
+        # If the provider has a custom redirect URL, use that, otherwise automatically generate one
+        redirect_uri = OAUTH_PROVIDERS[provider].get("redirect_uri") or request.url_for(
+            "oauth_callback", provider=provider
+        )
+        client = self.get_client(provider)
+        if client is None:
+            raise HTTPException(404)
+        return await client.authorize_redirect(request, redirect_uri)
+
+    async def handle_callback(self, provider, request, response):
+        if provider not in OAUTH_PROVIDERS:
+            raise HTTPException(404)
+        client = self.get_client(provider)
+        try:
+            token = await client.authorize_access_token(request)
+        except Exception as e:
+            log.warning(f"OAuth callback error: {e}")
+            raise HTTPException(400, detail=ERROR_MESSAGES.INVALID_CRED)
+        user_data: UserInfo = token["userinfo"]
+        if not user_data:
+            user_data: UserInfo = await client.userinfo(token=token)
+        if not user_data:
+            log.warning(f"OAuth callback failed, user data is missing: {token}")
+            raise HTTPException(400, detail=ERROR_MESSAGES.INVALID_CRED)
+
+        sub = user_data.get("sub")
+        if not sub:
+            log.warning(f"OAuth callback failed, sub is missing: {user_data}")
+            raise HTTPException(400, detail=ERROR_MESSAGES.INVALID_CRED)
+        provider_sub = f"{provider}@{sub}"
+        email_claim = auth_manager_config.OAUTH_EMAIL_CLAIM
+        email = user_data.get(email_claim, "").lower()
+        # We currently mandate that email addresses are provided
+        if not email:
+            log.warning(f"OAuth callback failed, email is missing: {user_data}")
+            raise HTTPException(400, detail=ERROR_MESSAGES.INVALID_CRED)
+
+        # Check if the user exists
+        user = Users.get_user_by_oauth_sub(provider_sub)
+
+        if not user:
+            # If the user does not exist, check if merging is enabled
+            if auth_manager_config.OAUTH_MERGE_ACCOUNTS_BY_EMAIL:
+                # Check if the user exists by email
+                user = Users.get_user_by_email(email)
+                if user:
+                    # Update the user with the new oauth sub
+                    Users.update_user_oauth_sub_by_id(user.id, provider_sub)
+
+        if user:
+            determined_role = self.get_user_role(user, user_data)
+            if user.role != determined_role:
+                Users.update_user_role_by_id(user.id, determined_role)
+
+        if not user:
+            # If the user does not exist, check if signups are enabled
+            if auth_manager_config.ENABLE_OAUTH_SIGNUP:
+                # Check if an existing user with the same email already exists
+                existing_user = Users.get_user_by_email(
+                    user_data.get("email", "").lower()
+                )
+                if existing_user:
+                    raise HTTPException(400, detail=ERROR_MESSAGES.EMAIL_TAKEN)
+
+                picture_claim = auth_manager_config.OAUTH_PICTURE_CLAIM
+                picture_url = user_data.get(picture_claim, "")
+                if picture_url:
+                    # Download the profile image into a base64 string
+                    try:
+                        async with aiohttp.ClientSession() as session:
+                            async with session.get(picture_url) as resp:
+                                picture = await resp.read()
+                                base64_encoded_picture = base64.b64encode(
+                                    picture
+                                ).decode("utf-8")
+                                guessed_mime_type = mimetypes.guess_type(picture_url)[0]
+                                if guessed_mime_type is None:
+                                    # assume JPG, browsers are tolerant enough of image formats
+                                    guessed_mime_type = "image/jpeg"
+                                picture_url = f"data:{guessed_mime_type};base64,{base64_encoded_picture}"
+                    except Exception as e:
+                        log.error(
+                            f"Error downloading profile image '{picture_url}': {e}"
+                        )
+                        picture_url = ""
+                if not picture_url:
+                    picture_url = "/user.png"
+                username_claim = auth_manager_config.OAUTH_USERNAME_CLAIM
+
+                role = self.get_user_role(None, user_data)
+
+                user = Auths.insert_new_auth(
+                    email=email,
+                    password=get_password_hash(
+                        str(uuid.uuid4())
+                    ),  # Random password, not used
+                    name=user_data.get(username_claim, "User"),
+                    profile_image_url=picture_url,
+                    role=role,
+                    oauth_sub=provider_sub,
+                )
+
+                if auth_manager_config.WEBHOOK_URL:
+                    post_webhook(
+                        auth_manager_config.WEBHOOK_URL,
+                        auth_manager_config.WEBHOOK_MESSAGES.USER_SIGNUP(user.name),
+                        {
+                            "action": "signup",
+                            "message": auth_manager_config.WEBHOOK_MESSAGES.USER_SIGNUP(
+                                user.name
+                            ),
+                            "user": user.model_dump_json(exclude_none=True),
+                        },
+                    )
+            else:
+                raise HTTPException(
+                    status.HTTP_403_FORBIDDEN, detail=ERROR_MESSAGES.ACCESS_PROHIBITED
+                )
+
+        jwt_token = create_token(
+            data={"id": user.id},
+            expires_delta=parse_duration(auth_manager_config.JWT_EXPIRES_IN),
+        )
+
+        # Set the cookie token
+        response.set_cookie(
+            key="token",
+            value=jwt_token,
+            httponly=True,  # Ensures the cookie is not accessible via JavaScript
+            samesite=WEBUI_SESSION_COOKIE_SAME_SITE,
+            secure=WEBUI_SESSION_COOKIE_SECURE,
+        )
+
+        # Redirect back to the frontend with the JWT token
+        redirect_url = f"{request.base_url}auth#token={jwt_token}"
+        return RedirectResponse(url=redirect_url)
+
+
+oauth_manager = OAuthManager()
@@ -0,0 +1,183 @@
+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_messages_openai_to_ollama(messages: list[dict]) -> list[dict]:
+    ollama_messages = []
+
+    for message in messages:
+        # Initialize the new message structure with the role
+        new_message = {"role": message["role"]}
+
+        content = message.get("content", [])
+
+        # Check if the content is a string (just a simple message)
+        if isinstance(content, str):
+            # If the content is a string, it's pure text
+            new_message["content"] = content
+        else:
+            # Otherwise, assume the content is a list of dicts, e.g., text followed by an image URL
+            content_text = ""
+            images = []
+
+            # Iterate through the list of content items
+            for item in content:
+                # Check if it's a text type
+                if item.get("type") == "text":
+                    content_text += item.get("text", "")
+
+                # Check if it's an image URL type
+                elif item.get("type") == "image_url":
+                    img_url = item.get("image_url", {}).get("url", "")
+                    if img_url:
+                        # If the image url starts with data:, it's a base64 image and should be trimmed
+                        if img_url.startswith("data:"):
+                            img_url = img_url.split(",")[-1]
+                        images.append(img_url)
+
+            # Add content text (if any)
+            if content_text:
+                new_message["content"] = content_text.strip()
+
+            # Add images (if any)
+            if images:
+                new_message["images"] = images
+
+        # Append the new formatted message to the result
+        ollama_messages.append(new_message)
+
+    return ollama_messages
+
+
+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"] = convert_messages_openai_to_ollama(
+        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
@@ -0,0 +1,139 @@
+from datetime import datetime
+from io import BytesIO
+from pathlib import Path
+from typing import Dict, Any, List
+
+from markdown import markdown
+
+import site
+from fpdf import FPDF
+
+from open_webui.env import STATIC_DIR, FONTS_DIR
+from open_webui.apps.webui.models.chats import ChatTitleMessagesForm
+
+
+class PDFGenerator:
+    """
+    Description:
+    The `PDFGenerator` class is designed to create PDF documents from chat messages.
+    The process involves transforming markdown content into HTML and then into a PDF format
+
+    Attributes:
+    - `form_data`: An instance of `ChatTitleMessagesForm` containing title and messages.
+
+    """
+
+    def __init__(self, form_data: ChatTitleMessagesForm):
+        self.html_body = None
+        self.messages_html = None
+        self.form_data = form_data
+
+        self.css = Path(STATIC_DIR / "assets" / "pdf-style.css").read_text()
+
+    def format_timestamp(self, timestamp: float) -> str:
+        """Convert a UNIX timestamp to a formatted date string."""
+        try:
+            date_time = datetime.fromtimestamp(timestamp)
+            return date_time.strftime("%Y-%m-%d, %H:%M:%S")
+        except (ValueError, TypeError) as e:
+            # Log the error if necessary
+            return ""
+
+    def _build_html_message(self, message: Dict[str, Any]) -> str:
+        """Build HTML for a single message."""
+        role = message.get("role", "user")
+        content = message.get("content", "")
+        timestamp = message.get("timestamp")
+
+        model = message.get("model") if role == "assistant" else ""
+
+        date_str = self.format_timestamp(timestamp) if timestamp else ""
+
+        # extends pymdownx extension to convert markdown to html.
+        # - https://facelessuser.github.io/pymdown-extensions/usage_notes/
+        html_content = markdown(content, extensions=["pymdownx.extra"])
+
+        html_message = f"""
+              <div class="message">
+                  <small> {date_str} </small>
+                  <div>
+                      <h2>
+                          <strong>{role.title()}</strong>
+                          <small class="text-muted">{model}</small>
+                      </h2>
+                  </div>
+                  <div class="markdown-section">
+                      {html_content}
+                  </div>
+              </div>
+          """
+        return html_message
+
+    def _generate_html_body(self) -> str:
+        """Generate the full HTML body for the PDF."""
+        return f"""
+        <html>
+            <head>
+                <meta charset="UTF-8">
+                <meta name="viewport" content="width=device-width, initial-scale=1.0">
+            </head>
+            <body>
+                <div class="container"> 
+                    <div class="text-center">
+                        <h1>{self.form_data.title}</h1>
+                    </div>
+                    <div>
+                        {self.messages_html}
+                    </div>
+                </div>
+            </body>
+        </html>
+        """
+
+    def generate_chat_pdf(self) -> bytes:
+        """
+        Generate a PDF from chat messages.
+        """
+        try:
+            global FONTS_DIR
+
+            pdf = FPDF()
+            pdf.add_page()
+
+            # 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"
+            # When running using `pip install -e .` the static directory is in the site packages.
+            # This path only works if `open-webui serve` is run from the root of this project.
+            if not FONTS_DIR.exists():
+                FONTS_DIR = Path("./backend/static/fonts")
+
+            pdf.add_font("NotoSans", "", f"{FONTS_DIR}/NotoSans-Regular.ttf")
+            pdf.add_font("NotoSans", "b", f"{FONTS_DIR}/NotoSans-Bold.ttf")
+            pdf.add_font("NotoSans", "i", f"{FONTS_DIR}/NotoSans-Italic.ttf")
+            pdf.add_font("NotoSansKR", "", f"{FONTS_DIR}/NotoSansKR-Regular.ttf")
+            pdf.add_font("NotoSansJP", "", f"{FONTS_DIR}/NotoSansJP-Regular.ttf")
+            pdf.add_font("NotoSansSC", "", f"{FONTS_DIR}/NotoSansSC-Regular.ttf")
+
+            pdf.set_font("NotoSans", size=12)
+            pdf.set_fallback_fonts(["NotoSansKR", "NotoSansJP", "NotoSansSC"])
+
+            pdf.set_auto_page_break(auto=True, margin=15)
+
+            # Build HTML messages
+            messages_html_list: List[str] = [
+                self._build_html_message(msg) for msg in self.form_data.messages
+            ]
+            self.messages_html = "<div>" + "".join(messages_html_list) + "</div>"
+
+            # Generate full HTML body
+            self.html_body = self._generate_html_body()
+
+            pdf.write_html(self.html_body)
+
+            # Save the pdf with name .pdf
+            pdf_bytes = pdf.output()
+
+            return bytes(pdf_bytes)
+        except Exception as e:
+            raise e
@@ -0,0 +1,31 @@
+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"
+        yield line
+
+    yield "data: [DONE]\n\n"
@@ -104,5 +104,9 @@ def json_schema_to_pydantic_type(json_schema: dict[str, Any]) -> Any:
        return Optional[Any]  # Use Optional[Any] for nullable fields
    elif type_ == "literal":
        return Literal[literal_eval(json_schema.get("enum"))]
+    elif type_ == "optional":
+        inner_schema = json_schema.get("items", {"type": "string"})
+        inner_type = json_schema_to_pydantic_type(inner_schema)
+        return Optional[inner_type]
    else:
        raise ValueError(f"Unsupported JSON schema type: {type_}")
@@ -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)?$"
+    match = re.match(pattern, value, re.IGNORECASE)
+    if not match:
+        value = "max-age=31536000;includeSubDomains"
+    return {"Strict-Transport-Security": value}
+
+
+# Set X-Frame-Options response header
+def set_xframe(value: str):
+    pattern = r"^(DENY|SAMEORIGIN)$"
+    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}
@@ -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)
@@ -66,6 +67,55 @@ def title_generation_template(
        replacement_function,
        template,
    )
+    return template
+
+
+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)
+
+        # 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 get_messages_content(messages[: int(start_length)])
+        elif end_length is not None:
+            return get_messages_content(messages[-int(end_length) :])
+        elif middle_length is not None:
+            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"{{MESSAGES}}|{{MESSAGES:START:(\d+)}}|{{MESSAGES:END:(\d+)}}|{{MESSAGES:MIDDLETRUNCATE:(\d+)}}",
+        replacement_function,
+        template,
+    )
+
+    return template
+
+
+# {{prompt:middletruncate:8000}}
+
+
+def title_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,
@@ -79,36 +129,47 @@ def title_generation_template(
    return template


-def search_query_generation_template(
+def tags_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,
+        **(
+            {"user_name": user.get("name"), "user_location": user.get("location")}
+            if user
+            else {}
+        ),
+    )
+    return template
+
+
+def emoji_generation_template(
    template: str, prompt: str, user: Optional[dict] = None
 ) -> 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
-        elif start_length is not None:
-            return prompt[: int(start_length)]
-        elif end_length is not None:
-            return prompt[-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}"
-        return ""
-
-    template = re.sub(
-        r"{{prompt}}|{{prompt:start:(\d+)}}|{{prompt:end:(\d+)}}|{{prompt:middletruncate:(\d+)}}",
-        replacement_function,
+    template = replace_prompt_variable(template, prompt)
+    template = prompt_template(
        template,
+        **(
+            {"user_name": user.get("name"), "user_location": user.get("location")}
+            if user
+            else {}
+        ),
    )

+    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,
        **(
@@ -7,7 +7,7 @@ import jwt
 from open_webui.apps.webui.models.users import Users
 from open_webui.constants import ERROR_MESSAGES
 from open_webui.env import WEBUI_SECRET_KEY
-from fastapi import Depends, HTTPException, Request, status
+from fastapi import Depends, HTTPException, Request, Response, status
 from fastapi.security import HTTPAuthorizationCredentials, HTTPBearer
 from passlib.context import CryptContext

@@ -1,6 +1,6 @@
 fastapi==0.111.0
 uvicorn[standard]==0.30.6
-pydantic==2.8.2
+pydantic==2.9.2
 python-multipart==0.0.9

 Flask==3.0.3
@@ -11,7 +11,8 @@ python-jose==3.3.0
 passlib[bcrypt]==1.7.4

 requests==2.32.3
-aiohttp==3.10.5
+aiohttp==3.10.8
+async-timeout

 sqlalchemy==2.0.32
 alembic==1.13.2
@@ -36,19 +37,29 @@ tiktoken

 langchain==0.2.15
 langchain-community==0.2.12
-langchain-chroma==0.1.2
+langchain-chroma==0.1.4

 fake-useragent==1.5.1
-chromadb==0.5.5
-sentence-transformers==3.0.1
+chromadb==0.5.9
+pymilvus==2.4.7
+qdrant-client~=1.12.0
+
+sentence-transformers==3.2.0
+colbert-ai==0.2.21
+einops==0.8.0
+
+
+ftfy==6.2.3
 pypdf==4.3.1
+xhtml2pdf==0.2.16
+pymdown-extensions==10.11.2
 docx2txt==0.8
 python-pptx==1.0.0
 unstructured==0.15.9
 nltk==3.9.1
 Markdown==3.7
 pypandoc==1.13
-pandas==2.2.2
+pandas==2.2.3
 openpyxl==3.1.5
 pyxlsb==1.0.10
 xlrd==2.0.1
@@ -73,9 +84,11 @@ pytube==15.0.0

 extract_msg
 pydub
-duckduckgo-search~=6.2.11
+duckduckgo-search~=6.2.13

 ## Tests
 docker~=7.1.0
 pytest~=8.3.2
 pytest-docker~=3.1.1
+
+googleapis-common-protos==1.63.2
@@ -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 '*'
--- a/Show More
+++ b/Show More