Pinecone

What is Pinecone?

Pinecone is a fully managed, cloud-native vector database designed for building high-performance AI applications at scale. It enables fast and accurate similarity search across billions of high-dimensional vectors, making it ideal for semantic search, recommendation systems, and Retrieval-Augmented Generation (RAG) applications.

Pinecone is commonly used in these AI scenarios:

Architecture Overview

Pinecone's architecture is divided into two main components: the Control Plane and the Data Plane, providing separation of concerns for management and data operations.

Core Concepts

{
  "id": "doc-123",
  "values": [0.1, 0.2, 0.3, ...],  // Dense vector (1536 dims for OpenAI)
  "sparse_values": {"indices": [1, 5], "values": [0.5, 0.3]},  // Optional
  "metadata": {"category": "tech", "date": "2024-01-15"}
}

Serverless vs Pod-Based

Pinecone offers two deployment architectures, each suited for different use cases:

Key Features

Getting Started

# Install: pip install pinecone-client

from pinecone import Pinecone

# Initialize client
pc = Pinecone(api_key="YOUR_API_KEY")

# Create serverless index
pc.create_index(
    name="my-index",
    dimension=1536,  # OpenAI embedding dimension
    metric="cosine",
    spec=ServerlessSpec(cloud="aws", region="us-east-1")
)

# Connect to index
index = pc.Index("my-index")

# Upsert vectors
index.upsert(
    vectors=[
        {"id": "doc1", "values": [0.1, 0.2, ...], "metadata": {"category": "tech"}},
        {"id": "doc2", "values": [0.3, 0.4, ...], "metadata": {"category": "science"}}
    ],
    namespace="my-namespace"
)

# Query with metadata filter
results = index.query(
    vector=[0.1, 0.2, ...],
    top_k=5,
    filter={"category": {"$eq": "tech"}},
    namespace="my-namespace",
    include_metadata=True
)

# RAG with Pinecone + OpenAI
from openai import OpenAI
from pinecone import Pinecone

openai_client = OpenAI()
pc = Pinecone(api_key="YOUR_PINECONE_KEY")
index = pc.Index("knowledge-base")

def rag_query(question: str) -> str:
    # 1. Embed the question
    embedding = openai_client.embeddings.create(
        model="text-embedding-3-small",
        input=question
    ).data[0].embedding

    # 2. Search Pinecone for relevant context
    results = index.query(vector=embedding, top_k=3, include_metadata=True)
    context = "\n".join([r.metadata["text"] for r in results.matches])

    # 3. Generate answer with context
    response = openai_client.chat.completions.create(
        model="gpt-4",
        messages=[
            {"role": "system", "content": f"Answer based on context:\n{context}"},
            {"role": "user", "content": question}
        ]
    )
    return response.choices[0].message.content

Performance & Scaling

Best Practices

• Use namespaces: Isolate data by tenant, category, or time period
• Batch upserts: Send up to 100 vectors per upsert for efficiency
• Metadata filtering: Add relevant metadata to narrow search scope
• Choose the right metric: Cosine for normalized embeddings, Euclidean for absolute distances
• Monitor warm namespaces: Frequently queried namespaces stay cached for faster response
• Use hybrid search: Combine semantic + keyword for better recall

Integrations

Use Cases

Learn More

Related Topics

• RAG (Retrieval-Augmented Generation)
• Embeddings
• Chroma Vector Database
• LangChain
• LlamaIndex
• Semantic Kernel