What it is
A memory layer for professional teams. Upload documents, then ask questions across everything your team knows, with every answer linked to its source.
How it works
A full RAG pipeline: structure-preserving chunking, embeddings, and hybrid retrieval over Qdrant (dense vectors plus BM25), followed by reranking and citation-grounded answers.
Built with Python, FastAPI, Qdrant, Gemini, and Next.js. It won 2nd prize at the hackathon.