The official Pinecone Python client.
For more information, see the docs at https://www.pinecone.io/docs/
As part of an overall move to stop exposing generated code in the package's public interface, an obscure configuration property (openapi_config) was removed in favor of individual configuration options such as proxy_url, proxy_headers, and ssl_ca_certs. All of these properties were available in v3 and v4 releases of the SDK, with deprecation notices shown to affected users.
It is no longer necessary to install a separate plugin, pinecone-plugin-inference, to try out the Inference API; that plugin is now installed by default in the v5 SDK. See usage instructions below.
-
Upgrading to
4.x: For this upgrade you are unlikely to be impacted by breaking changes unless you are using thegrpcextras (see install steps below). Read full details in these v4 Release Notes. -
Upgrading to
3.x: Many things were changed in the v3 client to pave the way for Pinecone's new Serverless index offering. These changes are covered in detail in the v3 Migration Guide. Serverless indexes are only available in3.xrelease versions or greater.
Many of the brief examples shown in this README are using very small vectors to keep the documentation concise, but most real world usage will involve much larger embedding vectors. To see some more realistic examples of how this client can be used, explore some of our many Jupyter notebooks in the examples repository.
The Pinecone Python client is compatible with Python 3.8 and greater.
There are two flavors of the Pinecone python client. The default client installed from PyPI as pinecone-client has a minimal set of dependencies and interacts with Pinecone via HTTP requests.
If you are aiming to maximimize performance, you can install additional gRPC dependencies to access an alternate client implementation that relies on gRPC for data operations. See the guide on tuning performance.
# Install the latest version
pip3 install pinecone-client
# Install the latest version, with extra grpc dependencies
pip3 install "pinecone-client[grpc]"
# Install a specific version
pip3 install pinecone-client==5.0.0
# Install a specific version, with grpc extras
pip3 install "pinecone-client[grpc]"==5.0.0# Install the latest version
poetry add pinecone-client
# Install the latest version, with grpc extras
poetry add pinecone-client --extras grpc
# Install a specific version
poetry add pinecone-client==5.0.0
# Install a specific version, with grpc extras
poetry add pinecone-client==5.0.0 --extras grpcBefore you can use the Pinecone SDK, you must sign up for an account and find your API key in the Pinecone console dashboard at https://app.pinecone.io.
The Pinecone class is your main entry point into the Pinecone python SDK. If you have set your API Key in the PINECONE_API_KEY environment variable, you can instantiate the client with no other arguments.
from pinecone import Pinecone
pc = Pinecone() # This reads the PINECONE_API_KEY env varIf you prefer to pass configuration in code, for example if you have a complex application that needs to interact with multiple different Pinecone projects, the constructor accepts a keyword argument for api_key.
If you pass configuration in this way, you can have full control over what name to use for the environment variable, sidestepping any issues that would result
from two different client instances both needing to read the same PINECONE_API_KEY variable that the client implicitly checks for.
Configuration passed with keyword arguments takes precedence over environment variables.
import os
from pinecone import Pinecone
pc = Pinecone(api_key=os.environ.get('CUSTOM_VAR'))If your network setup requires you to interact with Pinecone via a proxy, you will need
to pass additional configuration using optional keyword parameters. These optional parameters are forwarded to urllib3, which is the underlying library currently used by the Pinecone client to make HTTP requests. You may find it helpful to refer to the urllib3 documentation on working with proxies while troubleshooting these settings.
Here is a basic example:
from pinecone import Pinecone
pc = Pinecone(
api_key='YOUR_API_KEY',
proxy_url='https://your-proxy.com'
)
pc.list_indexes()If your proxy requires authentication, you can pass those values in a header dictionary using the proxy_headers parameter.
from pinecone import Pinecone
import urllib3 import make_headers
pc = Pinecone(
api_key='YOUR_API_KEY',
proxy_url='https://your-proxy.com',
proxy_headers=make_headers(proxy_basic_auth='username:password')
)
pc.list_indexes()By default the Pinecone Python client will perform SSL certificate verification using the CA bundle maintained by Mozilla in the certifi package.
If your proxy server is using a self-signed certificate, you will need to pass the path to the certificate in PEM format using the ssl_ca_certs parameter.
from pinecone import Pinecone
import urllib3 import make_headers
pc = Pinecone(
api_key="YOUR_API_KEY",
proxy_url='https://your-proxy.com',
proxy_headers=make_headers(proxy_basic_auth='username:password'),
ssl_ca_certs='path/to/cert-bundle.pem'
)
pc.list_indexes()If you would like to disable SSL verification, you can pass the ssl_verify
parameter with a value of False. We do not recommend going to production with SSL verification disabled.
from pinecone import Pinecone
import urllib3 import make_headers
pc = Pinecone(
api_key='YOUR_API_KEY',
proxy_url='https://your-proxy.com',
proxy_headers=make_headers(proxy_basic_auth='username:password'),
ssl_ca_certs='path/to/cert-bundle.pem',
ssl_verify=False
)
pc.list_indexes()If you've followed instructions above to install with optional grpc extras, you can unlock some performance improvements by working with an alternative version of the client imported from the pinecone.grpc subpackage.
import os
from pinecone.grpc import PineconeGRPC
pc = PineconeGRPC(api_key=os.environ.get('PINECONE_API_KEY'))
# From here on, everything is identical to the REST-based client.
index = pc.Index(host='my-index-8833ca1.svc.us-east1-gcp.pinecone.io')
index.upsert(vectors=[])
index.query(vector=[...], top_key=10)The following example creates a serverless index in the us-west-2
region of AWS. For more information on serverless and regional availability, see Understanding indexes.
from pinecone import Pinecone, ServerlessSpec
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
pc.create_index(
name='my-index',
dimension=1536,
metric='euclidean',
deletion_protection='enabled',
spec=ServerlessSpec(
cloud='aws',
region='us-west-2'
)
)The following example creates an index without a metadata configuration. By default, Pinecone indexes all metadata.
from pinecone import Pinecone, PodSpec
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
pc.create_index(
name="example-index",
dimension=1536,
metric="cosine",
deletion_protection='enabled',
spec=PodSpec(
environment='us-west-2',
pod_type='p1.x1'
)
)Pod indexes support many optional configuration fields. For example, the following example creates an index that only indexes the "color" metadata field. Queries against this index cannot filter based on any other metadata field.
from pinecone import Pinecone, PodSpec
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
metadata_config = {
"indexed": ["color"]
}
pc.create_index(
"example-index-2",
dimension=1536,
spec=PodSpec(
environment='us-west-2',
pod_type='p1.x1',
metadata_config=metadata_config
)
)The following example returns all indexes in your project.
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
for index in pc.list_indexes():
print(index['name'])The following example returns information about the index example-index.
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
index_description = pc.describe_index("example-index")The following example deletes the index named example-index. Only indexes which are not protected by deletion protection may be deleted.
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
pc.delete_index("example-index")The following example changes the number of replicas for example-index.
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
new_number_of_replicas = 4
pc.configure_index("example-index", replicas=new_number_of_replicas)If you would like to enable deletion protection, which prevents an index from being deleted, the configure_index method also handles that via an optional deletion_protection keyword argument.
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
# To enable deletion protection
pc.configure_index("example-index", deletion_protection='enabled')
# Disable deletion protection
pc.configure_index("example-index", deletion_protection='disabled')
# Call describe index to verify the configuration change has been applied
desc = pc.describe_index("example-index")
print(desc.deletion_protection)The following example returns statistics about the index example-index.
import os
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
index = pc.Index(host=os.environ.get('INDEX_HOST'))
index_stats_response = index.describe_index_stats()The following example upserts vectors to example-index.
import os
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
index = pc.Index(host=os.environ.get('INDEX_HOST'))
upsert_response = index.upsert(
vectors=[
("vec1", [0.1, 0.2, 0.3, 0.4], {"genre": "drama"}),
("vec2", [0.2, 0.3, 0.4, 0.5], {"genre": "action"}),
],
namespace="example-namespace"
)The following example queries the index example-index with metadata
filtering.
import os
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
# Find your index host by calling describe_index
# through the Pinecone web console
index = pc.Index(host=os.environ.get('INDEX_HOST'))
query_response = index.query(
namespace="example-namespace",
vector=[0.1, 0.2, 0.3, 0.4],
top_k=10,
include_values=True,
include_metadata=True,
filter={
"genre": {"$in": ["comedy", "documentary", "drama"]}
}
)The following example deletes vectors by ID.
import os
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
# Find your index host by calling describe_index
# through the Pinecone web console
index = pc.Index(host=os.environ.get('INDEX_HOST'))
delete_response = index.delete(ids=["vec1", "vec2"], namespace="example-namespace")The following example fetches vectors by ID.
import os
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
# Find your index host by calling describe_index
# through the Pinecone web console
index = pc.Index(host=os.environ.get('INDEX_HOST'))
fetch_response = index.fetch(ids=["vec1", "vec2"], namespace="example-namespace")The following example updates vectors by ID.
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
# Find your index host by calling describe_index
# through the Pinecone web console
index = pc.Index(host=os.environ.get('INDEX_HOST'))
update_response = index.update(
id="vec1",
values=[0.1, 0.2, 0.3, 0.4],
set_metadata={"genre": "drama"},
namespace="example-namespace"
)The list and list_paginated methods can be used to list vector ids matching a particular id prefix.
With clever assignment of vector ids, this can be used to help model hierarchical relationships between
different vectors such as when there are embeddings for multiple chunks or fragments related to the
same document.
The list method returns a generator that handles pagination on your behalf.
from pinecone import Pinecone
pc = Pinecone(api_key='xxx')
index = pc.Index(host='hosturl')
# To iterate over all result pages using a generator function
namespace = 'foo-namespace'
for ids in index.list(prefix='pref', limit=3, namespace=namespace):
print(ids) # ['pref1', 'pref2', 'pref3']
# Now you can pass this id array to other methods, such as fetch or delete.
vectors = index.fetch(ids=ids, namespace=namespace)There is also an option to fetch each page of results yourself with list_paginated.
from pinecone import Pinecone
pc = Pinecone(api_key='xxx')
index = pc.Index(host='hosturl')
# For manual control over pagination
results = index.list_paginated(
prefix='pref',
limit=3,
namespace='foo',
pagination_token='eyJza2lwX3Bhc3QiOiI5IiwicHJlZml4IjpudWxsfQ=='
)
print(results.namespace) # 'foo'
print([v.id for v in results.vectors]) # ['pref1', 'pref2', 'pref3']
print(results.pagination.next) # 'eyJza2lwX3Bhc3QiOiI5IiwicHJlZml4IjpudWxsfQ=='
print(results.usage) # { 'read_units': 1 }The following example creates the collection example-collection from
example-index.
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
pc.create_collection(
name="example-collection",
source="example-index"
)The following example returns a list of the collections in the current project.
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
active_collections = pc.list_collections()The following example returns a description of the collection
example-collection.
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
collection_description = pc.describe_collection("example-collection")The following example deletes the collection example-collection.
from pinecone import Pinecone
pc = Pinecone(api_key='<<PINECONE_API_KEY>>')
pc.delete_collection("example-collection")The Pinecone SDK now supports creating embeddings via the Inference API.
from pinecone import Pinecone
pc = Pinecone(api_key="YOUR_API_KEY")
model = "multilingual-e5-large"
# Embed documents
text = [
"Turkey is a classic meat to eat at American Thanksgiving.",
"Many people enjoy the beautiful mosques in Turkey.",
]
text_embeddings = pc.inference.embed(
model=model,
inputs=text,
parameters={"input_type": "passage", "truncate": "END"},
)
# Upsert documents into Pinecone index
# Embed a query
query = ["How should I prepare my turkey?"]
query_embeddings = pc.inference.embed(
model=model,
inputs=query,
parameters={"input_type": "query", "truncate": "END"},
)
# Send query to Pinecone index to retrieve similar documentsIf you'd like to make a contribution, or get setup locally to develop the Pinecone python client, please see our contributing guide
![dependabot[bot] avatar](https://avatars.githubusercontent.com/in/29110?v=4 "dependabot[bot]")
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent