Weaviate
A self-hosted vector database that combines semantic and keyword search with RAG. Run it locally with Docker, on Kubernetes, or in a managed cloud service.
Weaviate is a self-hosted vector database for developers building search applications, RAG systems, recommendation engines, and chatbots. It stores data objects alongside their vector embeddings, so applications can search by meaning and filter results using structured data. You can run the database locally with Docker, deploy it on Kubernetes, or use the hosted Weaviate Cloud service.
Search combines vector similarity with keyword matching, including BM25, and supports image search. Built-in reranking helps order retrieved results, while generative search connects retrieval to an LLM for answers and summaries. These capabilities share a query interface, reducing the separate systems an application needs for retrieval and response generation.
You can supply your own embeddings or let integrated models generate them during import. Supported integrations include OpenAI, Cohere, HuggingFace, and Google. A Docker deployment can pair the database with a local embedding model; integrations with external model services send model requests outside that deployment. Query Agent turns natural-language questions into database queries, and Engram supports personalized AI experiences that adapt to individual users over time.
For shared production deployments, Weaviate supports multi-tenancy, replication, horizontal scaling, and role-based access control. Vector compression reduces memory use, and collection-level expiry rules remove stale objects automatically.
