Pinecone Deep Dive

Pinecone Deep Dive

Getting Started

1. Create account at pinecone.io 2. Create index with dimension (e.g., 1536) 3. Install SDK: pip install pinecone-client

Core Operations

Initialize: import pinecone pc = pinecone.Pinecone(api_key="...") index = pc.Index("my-index")

Upsert (Insert/Update): index.upsert(vectors=[ {"id": "doc1", "values": embedding, "metadata": {...}} ])

Query: results = index.query( vector=query_embedding, top_k=10, include_metadata=True )

Namespaces

Partition data within an index:

Metadata Filtering

Combine vector search with filters: index.query( vector=embedding, filter={"category": "science", "year": {"$gte": 2020}} )

Best Practices