GH Repository · infiniflow
infinity
The AI-native database built for LLM applications, providing incredibly fast hybrid search of dense vector, sparse vector, tensor (multi-vector), and full-text.
- stars
- 4,708
- 30-day movement
- +21/day
- Related entries
- 60
- Connections
- 2
pythonC++multi-vectorvectortensor-databasehnswai-nativehybrid-searchragpython/uvvector-searchinformation-retrivalvector-databasebm25vectordatabasenearest-neighbor-searchapproximate-nearest-neighbor-searchembeddingfull-text-searchsearch-enginecpp20cpp20-modules
Infinity is an open-source database designed for LLM applications, combining dense vector, sparse vector, tensor (multi-vector), and full-text search in one system. It is implemented in C++ (C++20) and supports hybrid search across these index types, including HNSW and BM25.
Use it when your RAG or LLM application needs fast hybrid retrieval over embeddings and text without stitching together separate vector and keyword search systems.
Use it to
- Run hybrid vector and full-text search for RAG pipelines
- Store and query dense and sparse embeddings
- Perform multi-vector (tensor) retrieval
- Combine BM25 keyword search with nearest-neighbor search
- Self-host a search backend for LLM applications
For Developers building RAG and LLM retrieval applications
- Role
- rag
- Language
- C++
- Licence
- Apache-2.0
- Forks
- 445
- Open issues
- 87
- Last push
- 2026-09-09
- Latest release
- v0.1.0 · 2024-04-30
topicsvector-databasehybrid-searchragfull-text-searchnearest-neighbor-searchembeddings