Error when using VespaRetriever with Langchain

[soph-ma]

Hi everyone,
I am following this tutorial for retrieving search results using Vespa and Langchain Vespa | 🦜️🔗 Langchain

I've done everything exactly as the tutorial says and I have such error:

Traceback (most recent call last):
  File "C:\some\path\lchain.py", line 13, in <module>
    retriever = VespaRetriever(vespa_app, vespa_query_body, vespa_content_field)
TypeError: Serializable.__init__() takes 1 positional argument but 4 were given

So, I've read langchain documentation langchain.retrievers.vespa_retriever.VespaRetriever — 🦜🔗 LangChain 0.0.266 and I am passing keyword arguments that are required by VespaRetriever:

retriever = VespaRetriever(app=vespa_app, body=vespa_query_body, content_field=vespa_content_field, metadata_fields=[])

And such error appeared:

pydantic.errors.ConfigError: field "app" not yet prepared so type is still a ForwardRef, you might need to call VespaRetriever.update_forward_refs().

Could anyone help me solve this? Could this be a bug? Calling update_forward_refs() does not help.

I am using Python 3.10.11, pyvespa 0.34.0 and langchain 0.0.250

[dosubot[bot]]

🤖

Hello,

From the context provided, it seems like there might be a discrepancy between the version of LangChain you're using (0.0.250) and the latest version (0.0.277). The required arguments for the VespaRetriever class in the LangChain codebase are url and content_field, which are passed to the from_params class method to instantiate a VespaRetriever object.

Here is the relevant code:

@classmethod
def from_params(
    cls,
    url: str,
    content_field: str,
    *,
    k: Optional[int] = None,
    metadata_fields: Union[Sequence[str], Literal["*"]] = (),
    sources: Union[Sequence[str], Literal["*"], None] = None,
    _filter: Optional[str] = None,
    yql: Optional[str] = None,
    **kwargs: Any,
) -> VespaRetriever:
    """Instantiate retriever from params.

    Args:
        url (str): Vespa app URL.
        content_field (str): Field in results to return as Document page_content.
        ...
    Returns:
        VespaRetriever: Instantiated VespaRetriever.
    """
    ...
    return cls(
        app=app,
        body=body,
        content_field=content_field,
        metadata_fields=metadata_fields,
    )

In this method, url is the URL of the Vespa application and content_field is the field in the results to return as Document page_content.

Unfortunately, I couldn't find any specific changes made to the VespaRetriever class between version 0.0.250 and the latest version 0.0.277. It's possible that the issue you're encountering is due to changes made in these versions.

As for the pydantic.errors.ConfigError you're encountering, it seems like the VespaRetriever class is expecting a ForwardRef for the app field, but it's not prepared yet. This might be due to the way the VespaRetriever class is defined or instantiated.

I would recommend trying to update your LangChain version to the latest one (0.0.277) and see if the issue persists. If it does, please provide more details about how you're instantiating the VespaRetriever class and the values you're passing to it.

I hope this helps! Let me know if you have any other questions.

Sources

• libs/langchain/langchain/retrievers/vespa_retriever.py

---

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.

[soph-ma]

I tried upgrading langchain version to 0.0.277, but the error persists.

Here is the full code:

om langchain.retrievers.vespa_retriever import VespaRetriever
from vespa.application import Vespa

vespa_app = Vespa(url="http://some/vespa/url")

vespa_query_body = {
    "yql": "select chunks from search_schema where userQuery()",
    "hits": 5,
    "ranking": "bm25",
    "locale": "en-us",
}

vespa_content_field = "text"
retriever = VespaRetriever(app=vespa_app, 
                           body=vespa_query_body, 
                           content_field=vespa_content_field, 
                           metadata_fields=[])
retriever.get_relevant_documents("what is vespa?")

[aazizisoufiane]

I have the same issue, have you found a solution?

[soph-ma]

No :(

[aazizisoufiane]

create a new env and install the older version pip install --upgrade langchain==0.0.224, it works for me

[aazizisoufiane]

thanks to Vespa community on Slack: add VespaRetriever.update_forward_refs(Vespa=Vespa) before VespaRetriever