Glass-box RAG

Ask questions about the Kubernetes documentation and watch every step of the answer: hybrid search, rank fusion, reranking, the streamed answer and the passage behind each citation. Built without a RAG framework: pgvector + Postgres full-text fused with Reciprocal Rank Fusion, cross-encoder reranking (Hit@5 0.915 on 94 reviewed questions), citations validated against the retrieved passages (zero invented), and a FastAPI streaming API with rate limiting and a daily LLM budget.

For more projects, visit my GitHub.