Prismix
Pinecone Vector DB 9 min read

Pinecone Guide 2025: Vector Database for AI & RAG Applications

A practical guide to Pinecone for developers building RAG pipelines, semantic search, and AI-powered retrieval — from first index to production query.

What Is Pinecone?

Pinecone is a fully managed vector database. Instead of storing rows and columns, it stores embedding vectors — high-dimensional numerical representations of text, images, or audio generated by machine learning models. You can then search for vectors that are semantically similar to a query vector, even if they don't share exact keywords.

The dominant use case is RAG (Retrieval-Augmented Generation): you embed your documents, store them in Pinecone, and at query time retrieve the most relevant chunks to pass as context to Claude, GPT-4o, or any LLM. This lets the model answer questions about your private data without fine-tuning.

Python SDK Quickstart

# Install

pip install pinecone openai

# Create index, upsert vectors, query

from pinecone import Pinecone, ServerlessSpec
from openai import OpenAI

pc = Pinecone(api_key="YOUR_PINECONE_API_KEY")
oai = OpenAI(api_key="YOUR_OPENAI_API_KEY")

# Create a serverless index (one-time setup)
pc.create_index(
    name="my-rag-index",
    dimension=1536,          # must match your embedding model
    metric="cosine",
    spec=ServerlessSpec(cloud="aws", region="us-east-1")
)

index = pc.Index("my-rag-index")

# Embed and upsert documents
texts = ["Claude is an AI assistant by Anthropic.", "GPT-4o is OpenAI's flagship model."]
embeddings = oai.embeddings.create(model="text-embedding-3-small", input=texts)

vectors = [
    {"id": f"doc-{i}", "values": e.embedding, "metadata": {"text": texts[i]}}
    for i, e in enumerate(embeddings.data)
]
index.upsert(vectors=vectors)

# Query: find the most similar document
query = "Who made Claude?"
q_embed = oai.embeddings.create(model="text-embedding-3-small", input=[query])
results = index.query(vector=q_embed.data[0].embedding, top_k=3, include_metadata=True)

for match in results["matches"]:
    print(f"Score: {match['score']:.3f} | {match['metadata']['text']}")

Get your API key: Sign up at app.pinecone.io, go to API Keys in the left sidebar, and create a key. The free tier includes 2 GB of serverless storage.

Core Concepts

Indexes

An index is a named collection of vectors with a fixed dimension and distance metric. You cannot change dimension or metric after creation. Use one index per embedding model.

Namespaces

Namespaces partition vectors within an index — useful for multi-tenant apps where each user has their own data. Query with namespace="user-123" to scope results.

Metadata filtering

Attach JSON metadata to each vector and filter at query time: filter={"category": {"$eq": "finance"}}. Filtering happens before similarity scoring — only filtered vectors are considered.

Distance metrics

Cosine — measures angle between vectors, independent of magnitude (best for text). Euclidean — measures straight-line distance (best for dense image embeddings). Dotproduct — product of magnitudes (use when your model normalizes for this).

Common Embedding Dimensions

Model Provider Dimension
text-embedding-3-small OpenAI 1536
text-embedding-3-large OpenAI 3072
embed-english-v3.0 Cohere 1024
textembedding-gecko Google 768
multilingual-e5-large Microsoft / HuggingFace 1024

Pinecone Pricing

Tier Price Notes
Serverless (free) $0 2 GB storage, 5 indexes, no credit card
Serverless (pay-as-you-go) ~$0.033/GB/mo storage + read/write units Scales to zero, variable traffic
Pod-based (p1.x1) ~$70/mo Dedicated, predictable latency, 1M 768d vectors

Pinecone vs Alternatives

Database Type Best for Free tier
Pinecone Managed cloud Production RAG, fastest setup 2 GB serverless
Weaviate Open-source / cloud Hybrid search (vector + keyword), self-host Sandbox (14 days)
Qdrant Open-source / cloud High performance, self-host, filtering 1 GB cloud free
Chroma Open-source, local Local dev, prototyping, Python notebooks Fully free (local)
pgvector PostgreSQL extension Already on Postgres, small-medium scale Free (self-host)

For more on RAG pipelines with Pinecone, see LlamaIndex guide and LangChain guide.

Monitor Pinecone Status

Pinecone has had outages affecting RAG pipelines in production. Prismix tracks live Pinecone status and sends instant alerts — so your on-call team knows before your users do.

Pinecone Troubleshooting Guide →