Chroma

If you're building RAG apps or anything that needs semantic search over documents, this gives you a local vector database with a clean Python API. Four core functions (add, query, get, delete), built-in embedding support for OpenAI and HuggingFace, and metadata filtering that actually works. It's legitimately popular at 24k+ GitHub stars and integrates cleanly with LangChain and LlamaIndex. The persistent client mode means your embeddings survive restarts, which is table stakes but worth noting. Best for prototyping and self-hosted setups. If you need managed infrastructure or are already at scale, look at Pinecone or Weaviate instead.