Skip to content

Conversation

@kovtcharov
Copy link
Contributor

Chat SDK / LLM Client

  • Fix 404 error on /api/v1/completions by switching default endpoint to /api/v1/chat/completions
  • Add structured messages support to LLMClient for all backends (Claude, chat, OpenAI)
  • Add _history_to_messages() helper to convert chat history to OpenAI-compatible format
  • Add no_history parameter to send_stream() for API consistency with send()
  • Add send_messages() and send_messages_stream() methods for explicit message array support

RAG Improvements

  • Add pymupdf to [rag] dependencies for PDF image extraction
  • Add RAG availability check for /index command with helpful error message
  • Add embeddings validation to catch malformed responses early
  • Improve error message when embedding model is not loaded (shows lemonade-server pull command)

Developer Experience

  • Update all user-facing help text to use uv pip install (recommended in docs)
  • Add unit tests to CI/CD workflow (test_chat_sdk.yml)

- Added a new workflow to run Chat SDK unit tests without requiring a server.
- Implemented the no_history parameter in send() and send_stream() methods to allow sending messages without affecting chat history.
- Introduced unit tests for the _history_to_messages() method to ensure proper message conversion from chat history.
- Added tests for send_messages() and send_messages_stream() methods to validate message handling and system prompt integration.
- Replaced 'pip install' with 'uv pip install' in various error messages and installation instructions across multiple files.
- Ensured consistency in messaging for installing dependencies related to evaluation, RAG, audio, and other components.
- Improved user guidance for dependency installation in the CLI and agent modules.
- Included 'pymupdf' in the RAG dependencies within setup.py.
- Improved error handling in RAGSDK to check for connection issues and empty responses from the embedding model.
- Added validation for embedding shapes to ensure correct data format and provide clearer error messages when issues arise.
@kovtcharov kovtcharov requested a review from itomek December 16, 2025 05:31
@kovtcharov kovtcharov self-assigned this Dec 16, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants