ray-project/llm-applications

Covers end-to-end RAG pipeline scaling across document loading, chunking, embedding, and indexing using Ray's distributed compute framework. Implements hybrid LLM routing to dynamically select between OpenAI and Anyscale open-source models, with built-in evaluation metrics for both component-level and end-to-end quality optimization. Includes production deployment patterns with vector database integration and multi-GPU serving configurations.

1,853 stars. No commits in the last 6 months.