# FluxIndex

A .NET 10 RAG library for building semantic search and retrieval systems, combining vector and keyword hybrid search with pluggable storage backends.

FluxIndex builds hybrid search (vector plus BM25) through a builder API, with batch indexing and embedding generation, cached query embeddings, and search-result caching that invalidates automatically on new writes.

It supports cross-encoder reranking, either local (no API key) or through an OpenAI-compatible endpoint, with automatic scaling between logit and probability score ranges so a relevance threshold keeps the same meaning across different rerankers. It also supports graph traversal over the index, including breadth-first and depth-first search, shortest and strongest path, and PageRank-style importance scoring.

Storage backends include SQLite, PostgreSQL, Neo4j and Qdrant, and it integrates with MCP servers for use by AI agents.

Website: https://github.com/iyulab/FluxIndex