The best vector databases, according to hackers
10 vector databases recommended in Hacker News comments over the past year, ranked by how many separate people recommend each one.
All vector databases
- QdrantPeople recommend Qdrant as an easy-to-run vector database with solid performance, especially for self-hosting and local RAG setups. Critics report shard failures and out-of-memory crashes at large scale.22recommend
- ChromaOpen-source Apache 2.0 database for AI search, favored for similarity search over documents and quick setup. Critics say results can disappoint at larger scale.17recommend
- sqlite-vecFree SQLite extension for storing and searching vectors locally, valued for simplicity and speed on small datasets. Critics note it is brute-force, so latency grows linearly with dataset size.17recommend
- turbopufferHosted database for vector and keyword search at large scale, recommended for its low cost and fast queries on very large datasets.13recommend
- LanceDBFree open-source embedded vector database people use for RAG and production work. Critics say inserts slow at scale and it is heavy for small needs.12recommend
- FAISSFree library for fast similarity search over vector collections, used for prototyping and local retrieval. Critics say it lacks metadata, update and operational tooling and is not suited to production.9recommend
- zvecPeople recommend it as a fast, lightweight embedded vector database that runs in-process and is simple to use. The most common caveat is that speed depends on local disk performance.8recommend
- Milvus vector databaseOpen-source vector database recommended for hybrid search, RAG and billions of vectors. Critics point to recent security flaws.7recommend
- USearchFree vector search library that people find fast at large scale and happy working with vectors on disk.7recommend
- PineconeManaged vector database praised for hybrid dense and sparse search and cost-effectiveness for specialized use. Critics say dedicated vector databases are often overkill.6recommend