How to beginner · 4 min read

How to store embeddings in Pinecone

Q: How to store embeddings in Pinecone

To store embeddings in Pinecone, first generate vector embeddings using an embedding model like text-embedding-3-small from OpenAI. Then, initialize a Pinecone index and use the index.upsert() method to insert vectors with unique IDs. This enables fast similarity search and retrieval.

Quick answer

To store embeddings in Pinecone, first generate vector embeddings using an embedding model like text-embedding-3-small from OpenAI. Then, initialize a Pinecone index and use the index.upsert() method to insert vectors with unique IDs. This enables fast similarity search and retrieval.

PREREQUISITES

Python 3.8+
OpenAI API key
Pinecone API key
pip install openai>=1.0 pinecone-client

Setup

Install the required Python packages and set environment variables for your OPENAI_API_KEY and PINECONE_API_KEY. Initialize the Pinecone client and create or connect to an index.

bash

pip install openai pinecone-client

Step by step

This example shows how to generate embeddings with OpenAI, create a Pinecone index, and store embeddings with unique IDs.

python

import os
from openai import OpenAI
from pinecone import Pinecone

# Initialize OpenAI client
openai_client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

# Initialize Pinecone client
pc = Pinecone(api_key=os.environ["PINECONE_API_KEY"])

# Create or connect to an index
index_name = "example-embedding-index"
if index_name not in pc.list_indexes():
    pc.create_index(index_name, dimension=1536)  # dimension must match embedding size
index = pc.Index(index_name)

# Sample texts to embed
texts = ["Hello world", "Pinecone vector database", "Store embeddings efficiently"]

# Generate embeddings using OpenAI
response = openai_client.embeddings.create(
    model="text-embedding-3-small",
    input=texts
)
embeddings = [data.embedding for data in response.data]

# Prepare vectors for upsert (id, vector, metadata)
vectors = [(f"vec{i}", emb, {"text": texts[i]}) for i, emb in enumerate(embeddings)]

# Upsert vectors into Pinecone index
index.upsert(vectors=vectors)

# Query example: find top 2 similar vectors to a query
query_response = openai_client.embeddings.create(
    model="text-embedding-3-small",
    input=["Vector databases"]
)
query_vector = query_response.data[0].embedding

results = index.query(vector=query_vector, top_k=2, include_metadata=True)

print("Query results:")
for match in results.matches:
    print(f"ID: {match.id}, Score: {match.score}, Text: {match.metadata['text']}")

output

Query results:
ID: vec1, Score: 0.92, Text: Pinecone vector database
ID: vec2, Score: 0.85, Text: Store embeddings efficiently

Common variations

Use async Pinecone client for high throughput applications.
Store additional metadata with vectors for richer filtering.
Use different embedding models like text-embedding-3-large or local sentence-transformers.
Batch upserts for large datasets to optimize performance.

Troubleshooting

If you get dimension mismatch errors, verify the embedding vector size matches the Pinecone index dimension.
Ensure your Pinecone API key has permissions and the environment is correctly set.
Check network connectivity if pinecone.init() fails.
Use pinecone.list_indexes() to confirm your index exists before upserting.

✅

Key Takeaways

Always match embedding vector dimension with Pinecone index dimension to avoid errors.
Use index.upsert() to efficiently store embeddings with unique IDs and optional metadata.
Generate embeddings with OpenAI or other models before storing in Pinecone for similarity search.
Batch operations improve performance when handling large embedding datasets.
Verify API keys and environment settings to prevent connection issues.

Verified 2026-04 · text-embedding-3-small, gpt-4o

Verify ↗