BaseCohere Rerank Argument Mismatch

[robuno]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

from langchain.retrievers.document_compressors import CohereRerank
from langchain.retrievers import ContextualCompressionRetriever
import cohere
from cohere import Client

class CustomCohereRerank(CohereRerank):
  class Config(CohereRerank.Config):
    arbitrary_types_allowed = True

CustomCohereRerank.update_forward_refs()
compressor = CustomCohereRerank(cohere_api_key=os.environ["COHERE_API_KEY"],
                                client=Client,
                                top_n=4)

query = "How do you combine multiple adapters?"
compressed_docs = compression_retriever.get_relevant_documents(query)
pretty_print_docs(compressed_docs)

Description

I am trying to use Langchain library to use Cohere Rerank by following this doc with HuggingFace LLM models, embeddings and FAISS db. A few days ago, It worked perfect and I obtained results but I am getting TypeError at these days:

[<ipython-input-41-f5aced2b34a0>](https://localhost:8080/#) in <cell line: 1>()
----> 1 compressed_docs = compression_retriever.get_relevant_documents(
      2     query
      3 )
      4 pretty_print_docs(compressed_docs)

4 frames
[/usr/local/lib/python3.10/dist-packages/langchain/retrievers/document_compressors/cohere_rerank.py](https://localhost:8080/#) in rerank(self, documents, query, model, top_n, max_chunks_per_doc)
     77         model = model or self.model
     78         top_n = top_n if (top_n is None or top_n > 0) else self.top_n
---> 79         results = self.client.rerank(
     80             query, docs, model, top_n=top_n, max_chunks_per_doc=max_chunks_per_doc
     81         )

TypeError: BaseCohere.rerank() takes 1 positional argument but 4 positional arguments (and 2 keyword-only arguments) were given

The results that I obtained a few days ago:

I have tried with some version combination of Cohere and LangChain to solve that but they did not work.

System Info

langchain==0.1.11
langchain-community==0.0.29
langchain-core==0.1.33
langchain-text-splitters==0.0.1

cohere==5.1.1
faiss-gpu==1.7.2

[dosubot[bot]]

Based on the error message you provided, it seems like the rerank method in the CohereRerank class is receiving more arguments than it expects. This method is called within the get_relevant_documents function, which is where your error is occurring.

The rerank method in the CohereRerank class expects the following arguments:

1. documents: A sequence of documents to rerank. This can be a sequence of strings, Document objects, or dictionaries.
2. query: The query string to use for reranking.
3. model (optional): The model to use for re-ranking. Defaults to self.model if not provided.
4. top_n (optional): The number of results to return. If None, returns all results. Defaults to self.top_n. If provided as -1, it also defaults to self.top_n.
5. max_chunks_per_doc (optional): The maximum number of chunks derived from a document.

The error message suggests that rerank is receiving 4 positional arguments and 2 keyword-only arguments, which is more than it expects. This could be due to the way you're initializing CustomCohereRerank or calling get_relevant_documents.

Here are some steps to troubleshoot and resolve the issue:

1. Verify Document Types: Ensure that all items in the documents sequence are either Document objects, strings, or dictionaries. If you are passing custom objects or other types, you should convert them to one of the supported types.

2. Check Document Object Construction: If you are using Document objects, ensure they are correctly constructed and contain the page_content attribute. The Document class should be from the langchain_core.documents module, as expected by the CohereRerank class.

3. Review Cohere and LangChain Versions: Although you mentioned trying different version combinations, ensure that the versions of Cohere and LangChain you are using are compatible with each other. Incompatibilities between library versions could lead to unexpected behavior.

4. Debugging: Add print statements or use a debugger to inspect the documents parameter just before it's passed to the compress_documents or rerank method. This can help you verify the types and contents of the items in the sequence.

If after these steps the issue persists, consider providing more specific details about the TypeError (e.g., the exact error message and stack trace). This information could offer more clues about the root cause of the problem.

For more information, you can refer to the CohereRerank class in the LangChain codebase here.

Sources

• libs/langchain/langchain/retrievers/document_compressors/cohere_rerank.py

About Dosu

This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

If you want to continue the conversation, start your reply with @dosu-bot.

[fjzznjsh]

@dosu-bot step3, when i use langchain==0.1.13, what kind of version should i use for cohere

[fjzznjsh]

@robuno same problem, have you fixed it?

[robuno]

@fjzznjsh Thank you for your addition. No, I haven't found a solution yet.

[ddl-john-alexander]

There's a PR here. #19486 I referenced this conversation but you all may want to comment.

[taaha3244]

Same problem

[bhargav-11]

pip install langchain==0.1.14

[robuno]

It is working currently, latest package versions I checked:

cohere==5.4.0
langchain==0.1.20
langchain-community==0.0.38
langchain-core==0.1.52
langchain-openai==0.1.6
langchain-text-splitters==0.0.1