Weaviate
Open source vector database combining semantic search, keyword search, and filters, available self-hosted or managed
Overview
Weaviate is an open source vector database that combines vector search, keyword search, and structured filtering in a single hybrid search engine. The source code is free and self-hostable without limit, in Docker or Kubernetes. For those who do not want to manage infrastructure, the vendor offers Weaviate Cloud, a managed service with a permanent free plan limited to one cluster per user, a Flex plan starting at $45/month billed by usage with no commitment, and Premium plans (shared or dedicated) starting at $400/month as prepaid contracts. Built-in embedding models (Weaviate Embeddings) are billed separately, between $0.025 and $0.065 per million tokens depending on the model chosen.
Weaviate is not translated into French: the website, documentation, and management console are only in English. The vendor does not document specific GDPR guarantees for the free Weaviate Cloud plan; however, Flex and Premium plans allow you to choose the region on AWS, GCP, or Azure, including in Europe. The REST and GraphQL API is comprehensive, with official integrations to LangChain, LlamaIndex, Haystack, OpenAI, Cohere, and Hugging Face. The main limitation is the learning curve: configuring a schema, choosing the right vectorizers, and sizing a self-hosted cluster requires DevOps skills that not all small teams possess.
Our verdict
Best for development teams wanting hybrid search (vector and keyword) with the freedom of self-hosting or a managed service. Not for you if your team lacks in-house DevOps skills: configuring and sizing a Weaviate cluster requires more preparation than simpler vector databases like Chroma.