Choosing a vector database is one of the first decisions in any RAG project. Three options dominate developer usage: Pinecone , Chroma , and Weaviate . Here's a practical comparison.

Quick Overview

Pinecone Chroma Weaviate

Hosting Managed cloud only Local or cloud Self-hosted or cloud

Setup time 5 minutes 1 minute 10–20 minutes

Free tier Yes (1 index) Yes (local) Yes (self-hosted)

Best for Production scale Development, prototyping Hybrid search, complex queries

Chroma: Start Here

Chroma runs in-process — no server needed. Perfect for development and small projects.

pip install chromadb

Enter fullscreen mode

Exit fullscreen mode

import chromadb from chromadb.utils import embedding_functions

client = chromadb . PersistentClient ( " ./chroma_db " )

ef = embedding_functions . SentenceTransformerEmbeddingFunction ( model_name = " all-MiniLM-L6-v2 " )

collection = client . get_or_create_collection ( " docs " , embedding_function = ef )

collection .…