Skip to content

Latest commit

 

History

History
275 lines (205 loc) · 6.7 KB

README.md

File metadata and controls

275 lines (205 loc) · 6.7 KB

Header

pinecone

Pub Publish

pinecone is an unofficial Dart client for your managed Pinecone vector database instance.

Is this package actively maintained?

Yes! This package is used in production applications and is actively maintained!

API Reference

For detailed API information, please see the official Pinecone API reference. This package is a thin wrapper around the official API and aims to have full parity with the official API.

This Dart client was generated using the openapi_spec package by mirroring the implementation details defined in the Pinecone API.

Client Instance

To create a client instance you simply need the API key for your Pinecone project. You can find your API key in the Pinecone console.

final client = PineconeClient(
  apiKey: '123-456-789',
);

Index Operations

All index operations require an environment parameter which is used to route to the appropriate host:

https://controller.{environment}.pinecone.io

You can find your environment name in the Pinecone console.

List Indexes

Official Documentation: list_indexes

List<String> indexes = await client.listIndexes(
  environment: environment,
);

Create Index

Official Documentation: create_index

await client.createIndex(
  environment: environment,
  request: CreateIndexRequest(
    name: indexName,
    dimension: dimension,
    metric: SearchMetric.cosine,
    pods: 1,
    replicas: 1,
    podType: PodType.p1x1,
  ),
);

Describe Index

Official Documentation: describe_index

Index index = await client.describeIndex(
  environment: environment,
  indexName: indexName,
);

Delete Index

Official Documentation: delete_index

await client.deleteIndex(
  environment: environment,
  indexName: indexName,
);

Configure Index

Official Documentation: configure_index

await client.configureIndex(
  environment: environment,
  indexName: indexName,
  request: ConfigureIndexRequest(
    replicas: 2,
    podType: PodType.p2x2,
  )
);

List Collections

Official Documentation: list_collections

List<String> collections await client.listCollections(
  environment: environment,
);

Create Collection

Official Documentation: create_collection

await client.createCollection(
  environment: environment,
  request: CreateCollectionRequest(
    name: collectionName,
    source: indexName,
  ),
);

Describe Collection

Official Documentation: describe_collection

Collection collection = await client.describeCollection(
  environment: environment,
  collectionName: collectionName,
);

Delete Collection

Official Documentation: delete_collection

await client.deleteCollection(
  environment: environment,
  collectionName: collectionName,
);

Vector Operations

All vector operations require the indexName, projectId, and environment parameters which determine the appropriate host:

https://{indexName}-{projectId}.svc.{environment}.pinecone.io

For convenience, each of these components can be retrieved from the Index object:

final Index index = await client.describeIndex(
  indexName: indexName,
);

final indexName =  index.name;
final projectId =  index.projectId;
final environment = index.environment;

To retrieve the full host URL, you can use the status property:

final String host = index.status.host;

Describe Index Stats

Official Documentation: describe_index_stats

IndexStats indexStats = await client.describeIndexStats(
  indexName: index.name,
  projectId: index.projectId,
  environment: index.environment,
);

Query

Official Documentation: query

QueryResponse queryResponse = await client.queryVectors(
  indexName: index.name,
  projectId: index.projectId,
  environment: index.environment,
  request: QueryRequest(
    vector: queryVector,
  ),
);

Delete

Official Documentation: delete

await client.deleteVectors(
  indexName: index.name,
  projectId: index.projectId,
  environment: index.environment,
  request: DeleteRequest(
    ids: ['vector-0', 'vector-1', 'vector-2'],
  ),
);

Fetch

Official Documentation: fetch

FetchResponse fetchResponse = await client.fetchVectors(
  indexName: index.name,
  projectId: index.projectId,
  environment: index.environment,
  ids: ['vector-0', 'vector-1', 'vector-2'],
);

Update

Official Documentation: update

await client.updateVector(
  indexName: index.name,
  projectId: index.projectId,
  environment: index.environment,
  request: UpdateRequest(
    id: 'vector-5',
    namespace: namespaceName,
    values: List.generate(dimension, (k) => 999.9),
    setMetadata: {'test-meta': 'new-meta-value'},
  ),
);

Upsert

Official Documentation: upsert

UpsertResponse upsertResponse = await client.upsertVectors(
  indexName: index.name,
  projectId: index.projectId,
  environment: index.environment,
  request: UpsertRequest(
    namespace: namespaceName,
    vectors: [
      for (var i = 0; i < 10; i++)
        Vector(
          id: 'vector-$i',
          values: List.generate(dimension, (k) => (k + i).toDouble()),
          metadata: {'test-meta': 'test-value-$i'},
        ),
    ],
  ),
);

Contributing

Please see the pinecone Github repository. Feel free to open an issue to report bugs or request new features.