infinity

Infinity addresses the challenge of fast, hybrid search across multiple data types including dense vectors, sparse vectors, tensors, and full text, targeting LLM applications such as retrieval-augmented generation, question answering, and recommenders. Unlike typical vector databases that focus solely on dense embeddings, Infinity integrates hybrid search with rerankers like ColBERT and supports multi-vector and tensor data natively. It offers very low latency (0.1 ms for vector queries, 1 ms for full-text) and high throughput on million-scale datasets, packaged as a single binary with a Python API, easing deployment and developer adoption. With 4,500 stars and recent commits, it shows active maintenance but limited growth so far, suggesting it is still gaining traction. Teams requiring mature, widely adopted vector search solutions with extensive community support might prefer alternatives like Pinecone or Weaviate instead.

Choose Infinity when you need extremely low-latency hybrid search combining dense and sparse embeddings plus full-text search in a single system, especially for LLM-driven retrieval augmented generation workflows. Consider alternatives like Vespa or Qdrant if you need broader ecosystem integrations or more mature community support.

Infinity is suited for engineering teams building internal RAG search systems over large document corpora, conversational AI platforms requiring fast multi-modal retrieval, or recommendation engines that combine vector and keyword signals. Its Python-friendly API and Docker deployment target AI teams embedding search tightly with LLM applications.

Latest commit: 2026-05-13. Activity data is based on daily RepoPi snapshots of the GitHub repository.