Chroma

Chroma is designed to be the simplest way to add vector search to an LLM application. The default mode runs entirely in-memory with no setup, and switching to persistent or client-server mode requires only changing one parameter.

It handles embedding generation automatically — you provide text strings, Chroma calls the embedding function (OpenAI by default) and stores the vectors. At query time you pass a query string and get back the nearest documents.

pip install chromadb

import chromadb

client = chromadb.Client()  # in-memory
# client = chromadb.PersistentClient(path="./chroma_db")  # on-disk

collection = client.create_collection("docs")
collection.add(documents=["LangChain is an LLM framework", "Ollama runs models locally"], ids=["1", "2"])
results = collection.query(query_texts=["What runs LLMs locally?"], n_results=1)
print(results)