Pinecone Python SDK

The Pinecone Python SDK provides a client for the Pinecone vector database. Use it to create and manage indexes, upsert and query records, and run inference operations from Python.

Requires Python 3.10+.

Upgrading from 9.x? create and configure moved from spec=/dimension= to schema=/deployment=; see the v10 migration guide for the field-by-field mapping.

Installation

pip install pinecone

Quick start

An index declares its fields as a schema. Declaring a schema makes it a document index: you read and write it through index.documents, and each record is a JSON document whose fields you named yourself.

from pinecone import DenseVectorQuery, Pinecone

pc = Pinecone(api_key="your-api-key")  # or omit and set PINECONE_API_KEY

# Create an index. This blocks until the index is ready.
pc.indexes.create(
    name="movie-recommendations",
    schema={"fields": {"your_vector": {"type": "dense_vector", "dimension": 3, "metric": "cosine"}}},
    deployment={"deployment_type": "managed", "cloud": "aws", "region": "us-east-1"},
)

# Get a data-plane handle for that index
index = pc.index("movie-recommendations")

# Upsert documents. Each one needs an `_id`; every other key is either a field
# you declared in the schema or arbitrary metadata.
index.documents.upsert(
    namespace="movies-en",
    documents=[
        {"_id": "movie-001", "your_vector": [0.1, 0.2, 0.3], "title": "Arrival"},
        {"_id": "movie-002", "your_vector": [0.4, 0.5, 0.6], "title": "Interstellar"},
    ],
)

# Search. `score_by` names the field to compare against.
results = index.documents.search(
    namespace="movies-en",
    top_k=5,
    score_by=[DenseVectorQuery(field="your_vector", values=[0.1, 0.2, 0.3])],
    include_fields=["title"],
)
for doc in results.matches:
    print(doc.id, doc.score)

Upserts apply asynchronously, so a document may not be visible to the next search immediately.

Indexes created another way use other methods: an index created with the deprecated top-level dimension=/metric= uses index.upsert / index.query, and a legacy integrated index (pc.indexes.create_for_model(...)) uses index.upsert_records / index.search. For new indexes that embed text for you, give a document index a field with integrated embedding. See the quickstart and the documentation.

Async usage

The SDK provides an async client for use with asyncio. Its index() is a coroutine, and the handle it returns is a context manager:

import asyncio

from pinecone import AsyncPinecone, DenseVectorQuery


async def main():
    async with AsyncPinecone(api_key="your-api-key") as pc:
        index = await pc.index("movie-recommendations")
        async with index:
            results = await index.documents.search(
                namespace="movies-en",
                top_k=5,
                score_by=[DenseVectorQuery(field="your_vector", values=[0.1, 0.2, 0.3])],
                include_fields=["title"],
            )
            for doc in results.matches:
                print(doc.id, doc.score)


asyncio.run(main())

Configuration

from pinecone import Pinecone

pc = Pinecone(
    api_key="your-api-key",  # or omit and set PINECONE_API_KEY
    timeout=60,  # request timeout in seconds; index handles inherit it
)

Setting PINECONE_DEBUG=1 before importing pinecone sets the pinecone logger to DEBUG. The SDK installs no log handler, so configure one to see the output:

import logging

logging.basicConfig()

Development

Clone the repository and install the dev dependency group with uv, which is what CI does:

uv sync --group dev

uv run pytest tests/unit/ -x -v      # tests
uv run mypy --strict pinecone/       # type checking
uv run ruff check --fix              # linting
uv run ruff format                   # formatting

The unit suite must leave the working tree clean: git status is expected to report no changes after a bare uv run pytest tests/unit. Use tmp_path / tmp_path_factory if a test genuinely needs a file on disk.

Integration suites that hit a real backend are opt-in because they cost money; each module documents its own gate. Before any release that touches retry logic, HTTP transport, the AIMD adaptive-concurrency limiter, or the batch-upsert path, run tests/integration/test_retry_smoke.py with PINECONE_API_KEY and PINECONE_RETRY_SMOKE=1 set.

License

Apache-2.0. See LICENSE for details.