ValueError: client should be an instance of pinecone.Index, got <class 'pinecone.data.index.Index'>

Question

Please help me fix this:

import os
from pinecone import Pinecone, ServerlessSpec
from langchain.vectorstores import Pinecone as PineconeLangchain


os.environ["PINECONE_API_KEY"] = PINECONE_API_KEY


pc = Pinecone(api_key=PINECONE_API_KEY)


index_name = "medchat"


if index_name not in pc.list_indexes().names():
    pc.create_index(
        name=index_name,
        dimension=384,
        metric="cosine",
        spec=ServerlessSpec(
            cloud="aws",
            region=PINECONE_API_ENV
        )
    )


index = pc.Index(index_name)


docsearch = PineconeLangchain.from_existing_index(index_name, embeddings)


qa = RetrievalQA.from_chain_type(
    llm=llm,
    chain_type="stuff",
    retriever=docsearch.as_retriever(search_kwargs={'k': 2}),
    return_source_documents=True,
    chain_type_kwargs=chain_type_kwargs
)

Error:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-33-f1e2e49d4e0a> in <cell line: 0>()
     28 
     29 
---> 30 docsearch = PineconeLangchain.from_existing_index(index_name, embeddings)
     31 
     32 

2 frames
/usr/local/lib/python3.11/dist-packages/langchain_community/vectorstores/pinecone.py in __init__(self, index, embedding, text_key, namespace, distance_strategy)
     71             )
     72         if not isinstance(index, pinecone.Index):
---> 73             raise ValueError(
     74                 f"client should be an instance of pinecone.Index, " f"got {type(index)}"
     75             )

ValueError: client should be an instance of pinecone.Index, got <class 'pinecone.data.index.Index'>

The code was working correctly 2~3 weeks ago, but I checked today and this error was persistently coming up. I can see the problem is in these bits:

index = pc.Index(index_name)


docsearch = PineconeLangchain.from_existing_index(index_name, embeddings)

What's the problem?

Rob Davidson · Accepted Answer · 2025-02-21 16:37:51Z

Just had this problem myself. It seems that Pinecone has updated its data structures and Langchain is yet to update theirs to match.

The answer seems to be to use pip install langchain_pinecone and then create a PineconeVectorStore instead of the langchain.vectorstores option.

Tutorial from Pinecone can be found here

Some example code:

from pinecone import Pinecone
from langchain_pinecone import PineconeVectorStore

# Same as your example

os.environ["PINECONE_API_KEY"] = PINECONE_API_KEY


pc = Pinecone(api_key=PINECONE_API_KEY)


index_name = "medchat"


if index_name not in pc.list_indexes().names():
    pc.create_index(
        name=index_name,
        dimension=384,
        metric="cosine",
        spec=ServerlessSpec(
            cloud="aws",
            region=PINECONE_API_ENV
        )
    )


index = pc.Index(index_name)

Then choose a different vectorstore Note1: this uses the pinecone apikey again - must call up a pinecone instantiation from scratch. Note2: You've not said how you're creating your embeddings so I'm just copying your code here but i've tried this on lanchain's OpenAIEmbeddings

docsearch = PineconeVectorStore(pinecone_api_key = PINECONE_API_KEY, index_name=index_name, embedding=embeddings)

And then this is the same as your code

qa = RetrievalQA.from_chain_type(
    llm=llm,
    chain_type="stuff",
    retriever=docsearch.as_retriever(search_kwargs={'k': 2}),
    return_source_documents=True,
    chain_type_kwargs=chain_type_kwargs
)

This worked for me to remove the error

Collectives™ on Stack Overflow

ValueError: client should be an instance of pinecone.Index, got <class 'pinecone.data.index.Index'>

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related