Reference · Glossary
Pinecone
Last updated
A **managed, serverless vector database** — you send it embeddings and metadata, it handles indexing, scaling, and fast nearest-neighbor search without you running any infrastructure.
#When to use
Production RAG apps that need low-ops scaling, metadata filtering, and namespaces to separate tenants — without self-hosting a search cluster.
#When not to
Tiny personal projects or prototypes where an in-process store (like Chroma) is faster to start with and avoids an external service and its cost.
#Example
from pinecone import Pinecone
pc = Pinecone(api_key="...")
index = pc.Index("docs")
index.upsert(vectors=[("chunk-1", [0.12, 0.03, ...], {"source": "faq.md"})])
matches = index.query(vector=[0.10, 0.05, ...], top_k=3, include_metadata=True)