Pinecone Python Client (Legacy)

6.0.0 · deprecated · verified Sat Apr 11

The `pinecone-client` library is the **DEPRECATED** Python client for the Pinecone vector database, which allows users to store, search, and manage high-dimensional vectors. This package is no longer maintained, and its last version is 6.0.0. All users should migrate to the new `pinecone` package for continued support, updates, and new features. The new `pinecone` package follows a quarterly release cadence aligned with Pinecone's API versioning.

Warnings

breaking The `pinecone-client` package is DEPRECATED. The official Python client has been renamed to `pinecone`. Continuing to use `pinecone-client` will result in no further updates, bug fixes, or new features, and may lead to conflicts if both packages are installed.
Fix: Uninstall `pinecone-client` and install the new package: `pip uninstall pinecone-client && pip install pinecone`. Update import statements from `import pinecone` to `from pinecone import Pinecone` and instantiate the client `pc = Pinecone(...)` instead of `pinecone.init(...)`.
breaking The `pinecone.init()` function is no longer available in the new `pinecone` package (and was deprecated in later `pinecone-client` versions). Attempts to call it will result in an `AttributeError`.
Fix: Migrate to the class-based initialization: `from pinecone import Pinecone; pc = Pinecone(api_key='YOUR_API_KEY')`.
breaking The Python compatibility has changed. While `pinecone-client` supported Python 3.9+, the new `pinecone` SDK officially requires Python 3.10 or greater.
Fix: Upgrade your Python environment to 3.10 or later before migrating to the `pinecone` package.
gotcha Installing both `pinecone-client` and `pinecone` in the same environment can lead to confusing interactions and unexpected behavior.
Fix: Always uninstall `pinecone-client` completely before installing `pinecone`. Run `pip uninstall pinecone-client` first.
deprecated The `openapi_config` parameter for client initialization was deprecated in `pinecone-client` v5.x and removed in favor of direct parameters such as `proxy_url`, `proxy_headers`, and `ssl_ca_certs`.
Fix: Pass proxy and SSL configuration directly as keyword arguments to the `Pinecone` constructor (e.g., `Pinecone(api_key=..., proxy_url='...')`).

Install

pip install pinecone-client Deprecated Client Installation
pip install "pinecone-client[grpc]" Deprecated Client with gRPC (for performance)

Imports

Pinecone
```
from pinecone import Pinecone
```
The `pinecone.init()` function is deprecated and removed in the new `pinecone` package. Users must instantiate the `Pinecone` class directly.

Quickstart

This quickstart demonstrates how to initialize the **new** `pinecone` client, create a serverless index, upsert vectors, and query the index. It is crucial to use the `pinecone` package for all new development as `pinecone-client` is deprecated. Replace `YOUR_API_KEY` with your actual Pinecone API key and adjust the dimension and region as needed.

import os
from pinecone import Pinecone, ServerlessSpec, CloudProvider, AwsRegion

# Initialize the Pinecone client (using the new 'pinecone' package)
api_key = os.environ.get('PINECONE_API_KEY', 'YOUR_API_KEY')
# Ensure you have a Pinecone API Key set as an environment variable or replace 'YOUR_API_KEY'

pc = Pinecone(api_key=api_key)

index_name = 'my-first-index'
dimension = 1536  # Example dimension, choose based on your embedding model

# Create a serverless index (if it doesn't exist)
if index_name not in pc.list_indexes().names():
    pc.create_index(
        name=index_name,
        dimension=dimension,
        metric='cosine', # or 'euclidean', 'dotproduct'
        spec=ServerlessSpec(
            cloud=CloudProvider.AWS, 
            region=AwsRegion.US_EAST_1
        )
    )
print(f"Index '{index_name}' created or already exists.")

# Connect to the index
index = pc.Index(name=index_name)

# Example: Upserting data (replace with your actual vectors/text)
vectors_to_upsert = [
    {"id": "vec1", "values": [0.1]*dimension, "metadata": {"genre": "fiction"}},
    {"id": "vec2", "values": [0.2]*dimension, "metadata": {"genre": "non-fiction"}}
]
index.upsert(vectors=vectors_to_upsert)
print(f"Upserted {len(vectors_to_upsert)} vectors to '{index_name}'.")

# Example: Querying the index
query_vector = [0.15]*dimension
query_results = index.query(vector=query_vector, top_k=1, include_metadata=True)
print("Query results:", query_results)

# Example: Deleting the index (uncomment to run)
# pc.delete_index(index_name)
# print(f"Index '{index_name}' deleted.")

view raw JSON →