Typing issue with Pinecone

[dejoma]

Checklist

• [X] I added a very descriptive title to this issue.
• [X] I included a link to the documentation page I am referring to (if applicable).

Issue with current documentation:

If you update all pinecone/langchain packages to their latest versions:

↑ @langchain/community 0.2.4 → 0.2.6
↑ @pinecone-database/pinecone 2.2.1 → 2.2.2
↑ langchain 0.2.4 → 0.2.5

# having @langchain/pinecone@0.0.7

Then the following code starts failing (in build-time):

// My index with or without custom type; doesn't matter
const pinecone = new Pinecone();
const pineconeIndex = pinecone.Index<{
  helpMePlease: string;
}>('index-prod-eu');

const vectorStore = await PineconeStore.fromExistingIndex(embeddings, {
    pineconeIndex, // <- 
    namespace: 'books'
  });

Error with custom type:

Error without custom type:

Idea or request for content:

No response

[dosubot[bot]]

To resolve the TypeScript type mismatch error (TS2322) when using Pinecone and Langchain packages with the latest versions, ensure that the pineconeIndex property in PineconeStoreParams is correctly typed as Index<RecordMetadata> | undefined. The provided code already defines pineconeIndex as an optional property of type PineconeIndex in the PineconeStoreParams interface:

export interface PineconeStoreParams extends AsyncCallerParams {
  pineconeIndex?: PineconeIndex;
  textKey?: string;
  namespace?: string;
  filter?: PineconeMetadata;
  pineconeConfig?: {
    indexName: ConstructorParameters<typeof PineconeIndex>[0];
    config: ConstructorParameters<typeof PineconeIndex>[1];
    namespace?: string;
    indexHostUrl?: string;
    additionalHeaders?: HTTPHeaders;
  };
}

In the PineconeStore class constructor, the pineconeIndex is correctly assigned based on the provided parameters:

constructor(embeddings: EmbeddingsInterface, params: PineconeStoreParams) {
  super(embeddings, params);
  this.embeddings = embeddings;

  const {
    namespace,
    pineconeIndex,
    textKey,
    filter,
    pineconeConfig,
    ...asyncCallerArgs
  } = params;
  this.namespace = namespace;
  if (!pineconeIndex && !pineconeConfig) {
    throw new Error("pineconeConfig or pineconeIndex must be provided.");
  }
  if (pineconeIndex && pineconeConfig) {
    throw new Error(
      "Only one of pineconeConfig or pineconeIndex can be provided."
    );
  }

  if (pineconeIndex) {
    this.pineconeIndex = pineconeIndex;
  } else if (pineconeConfig) {
    this.pineconeIndex = new PineconeIndex(
      pineconeConfig.indexName,
      {
        ...pineconeConfig.config,
        sourceTag: "langchainjs",
      },
      pineconeConfig.namespace,
      pineconeConfig.indexHostUrl,
      pineconeConfig.additionalHeaders
    );
  }

  this.textKey = textKey ?? "text";
  this.filter = filter;
  this.caller = new AsyncCaller(asyncCallerArgs);
}

Ensure that the PineconeIndex type is correctly imported and used. The PineconeIndex type should be imported from the @pinecone-database/pinecone package:

import { Index as PineconeIndex } from "@pinecone-database/pinecone";

Additionally, refer to the test cases provided in the libs/langchain-pinecone/src/tests/vectorstores.test.ts file to ensure that your implementation aligns with the expected usage. For example, the test cases show how to instantiate PineconeStore with the pineconeIndex property:

const store = new PineconeStore(embeddings, { pineconeIndex: client as any });

Make sure that the PineconeIndex type is correctly defined in the @pinecone-database/pinecone package and that you are using the latest versions of both Pinecone and Langchain packages. If the error persists, it might be due to a version mismatch or an incorrect type definition in the Pinecone package ^[1][2].

To continue talking to Dosu, mention @dosu.

[jacoblee93]

Hey @dejoma, sorry you're running into this - could you share your lockfile? I'm unable to repro with my local setup.

Additionally (a bit of a shot in the dark) - can you try explicitly resolving your @langchain/core to a single version?

https://js.langchain.com/v0.2/docs/how_to/installation/#installing-integration-packages