Use official pinecone npm package

[fraserxu]

#104

I was able to replace the pinecone client with official one easily without much change. Tested to index and search both works fine.

index

const pinecone = new PineconeClient()
await pinecone.init({
  environment: "us-west1-gcp",
  apiKey: process.env.PINECONE_API_KEY,
})
const index = pinecone.Index("my-index")
await PineconeStore.fromDocuments(
  index,
  docs,
  new OpenAIEmbeddings({
    openAIApiKey: process.env.OPEN_API_KEY,
  })
)

query

const pinecone = new PineconeClient()

await pinecone.init({
  environment: "us-west1-gcp",
  apiKey: process.env.PINECONE_API_KEY,
})

const index = pinecone.Index("my-index")
const vectorStore = await PineconeStore.fromExistingIndex(
  index,
  new OpenAIEmbeddings({
    openAIApiKey: process.env.OPEN_API_KEY,
  })
)

const model = new OpenAI({
  openAIApiKey: process.env.OPEN_API_KEY,
})
const chain = VectorDBQAChain.fromLLM(model, vectorStore)
const response = await chain.call({
  query: "what does the doc say about pinecone",
})

TODO:

• update doc
• fix type error

[rschwabco]

@fraserxu We have a PR out moving the Pinecone client from Axios to fetch, should be merged shortly. Small breaking changes, like not using data in the response.

[rschwabco]

@fraserxu We have a PR out moving the Pinecone client from Axios to fetch, should be merged shortly. Small breaking changes, like not using data in the response.

PR was merge, v0.0.7 is now available.

[fraserxu]

Thanks @rschwabco 🎉 I'll get the new version in and update my pr accordingly. 🙇

[fraserxu]

PR was merge, v0.0.7 is now available.

I see 0.0.8 is available and just synced up with it.

[fraserxu]

Have fixed the build too.

[fraserxu]

Can you please take a look at this pull request and let me know what you think @hwchase17 ? Thanks!

[hwchase17]

lgtm! dont mind the breaking changes that much given the earliness of the package