LangChain Chroma

LangChain Chroma is an adapter that connects the LangChain framework to Chroma, an open-source, serverless vector database optimized for rapid prototyping. Using Chroma.from_documents, developers break text into chunks, create embeddings, and persist them on disk in a SQLite file (chroma.sqlite3) alongside the binary HNSW index files. Query-time calls like similarity_search return the top-k vectors in milliseconds, powering Retrieval-Augmented Generation (RAG) pipelines without external cloud costs. Namespace isolation supports multi-tenant SaaS, while in-process operation keeps the data inside your own infrastructure, with no transfer to an external provider. Because Chroma implements LangChain’s VectorStore interface, teams can replace it with Pinecone, FAISS, or Milvus with a single line of code, scaling from a laptop to a cluster as traffic grows. Observability callbacks record search latency and the number of results returned, while recall can only be measured against a labelled set in a separate evaluation, and the where and where_document parameters filter by metadata and by document text, while hybrid BM25 and vector retrieval has to be assembled separately, for example with an EnsembleRetriever over a BM25Retriever and the vector store retriever. Thus, LangChain Chroma provides zero configuration and low latency for chatbots, code assistants, and analytics agents.

Work with us

Ready to put agentic AI to work?

Book a free 45-minute consultation. We'll map one real process worth automating with production-grade AI.