- Removed outdated dependencies: unstructured-client and langchain-unstructured.
- Added new versions for unstructured-client (0.42.3), unstructured[all-docs] (0.18.31), and langchain-unstructured (1.0.1).
- Added gitingest as a dependency to streamline the ingestion of GitHub repositories.
- Refactored GitHubConnector to utilize gitingest for efficient repository digest generation, reducing API calls.
- Updated GitHub indexer to process entire repository digests, enhancing performance and simplifying the indexing process.
- Modified GitHub connect form to indicate that the Personal Access Token is optional for public repositories.
- Introduced a new notifications table with relevant fields and indexes.
- Configured Electric SQL replication for the notifications, search_source_connectors, and documents tables.
- Centralized Electric SQL user credentials in the migration script for better management.
- Ensured idempotency in the migration process for creating users and publications.
- Removed all references to the Serper API from the codebase, including related components, hooks, and schemas.
- Updated the `pyproject.toml` to reflect the new version of `chonkie` and other dependencies.
- Cleaned up the configuration and connector management to streamline the application.
- Added BACKEND_URL to .env.example for optional OAuth callback configuration when behind a reverse proxy with HTTPS.
- Updated fastapi-users dependency to version 15.0.3 in pyproject.toml.
- Updated argon2-cffi and email-validator versions in uv.lock.
- Enhanced Google OAuth integration in app.py to handle secure contexts for CSRF cookies.
- Updated GoogleLoginButton component to include credentials in fetch requests.
- Updated documentation to reflect the new BACKEND_URL variable in both docker and manual installation guides.
- Updated langchain-unstructured to version 1.0.0 and langchain to 1.2.0.
- Updated litellm to version 1.80.10 and grpcio versions based on Python version.
- Updated fastuuid to version 0.14.0 and langgraph to 1.0.5.
- Updated openai to version 2.12.0 and langgraph-sdk to 0.3.0.
- Added uuid-utils package version 0.12.0.
- Refactored imports in utils.py, chats_routes.py, and query_service.py to use langchain_core.messages.
- Removed outdated dependencies: langchain-community, langchain-unstructured, langgraph, langchain-litellm, and litellm.
- Added updated versions of langchain-community (0.3.31), langchain-unstructured (0.1.6), langgraph (1.0.1), langchain-litellm (0.3.5), and litellm (1.77.5).
- Updated langchain (0.3.27 to 0.3.80) and langchain-community (0.3.25 to 0.3.31) in uv.lock.
- Updated langgraph (0.4.8 to 1.0.1) and langchain-litellm (0.2.3 to 0.3.5) in uv.lock.
- Updated xxhash version from 3.5.0 to 3.6.0 in uv.lock.
- Add STT service with CPU-optimized Faster-Whisper
- Add API endpoints for transcription and model management
- Add React audio recorder component
- Support multiple Whisper models (tiny to large-v3)
- Include error handling for corrupted/invalid files
- Tested with real speech audio (99% accuracy)
- No external API dependencies, fully offline
- Added DOCLING as third ETL_SERVICE option (alongside UNSTRUCTURED/LLAMACLOUD)
- Implemented add_received_file_document_using_docling function
- Added Docling processing logic in documents_routes.py
- Enhanced chunking with configurable overlap support
- Added comprehensive document processing service
- Supports both CPU and GPU processing with user selection
Addresses #161 - Add Docling Support as an ETL_SERVICE
Follows same pattern as LlamaCloud integration (PR #123)
