YandexGPT client returns an error when invoked

[nktpv]

Checked other resources

• I added a very descriptive title to this issue.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.
• I am sure that this is a bug in LangChain rather than my code.
• The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).

Example Code

pip install -U langchain langchain-community yandexcloud

from langchain_community.llms import YandexGPT
from langchain_core.prompts import PromptTemplate
from langchain.chains import LLMChain

os.environ['YC_API_KEY'] = 'xxxxxxx'
os.environ['YC_FOLDER_ID'] = 'xxxxxxx'
llm = YandexGPT()

template = "What is the capital of {country}?"
prompt = PromptTemplate.from_template(template)
llm_chain = LLMChain(prompt=prompt, llm=llm)
country = "Russia"

llm_chain.invoke(country)

raises an error:

Error Message and Stack Trace (if applicable)

WARNING:langchain_community.llms.yandex:Retrying langchain_community.llms.yandex.completion_with_retry.<locals>._completion_with_retry in 1.0 seconds as it raised _MultiThreadedRendezvous: <_MultiThreadedRendezvous of RPC that terminated with:
	status = StatusCode.UNAUTHENTICATED
	details = "IAM token or API key has to be passed in request"
	debug_error_string = "UNKNOWN:Error received from peer ipv4:158.160.54.160:443 {created_time:"2024-08-20T07:41:09.86231059+00:00", grpc_status:16, grpc_message:"IAM token or API key has to be passed in request"}"
>.
WARNING:langchain_community.llms.yandex:Retrying langchain_community.llms.yandex.completion_with_retry.<locals>._completion_with_retry in 2.0 seconds as it raised _MultiThreadedRendezvous: <_MultiThreadedRendezvous of RPC that terminated with:
	status = StatusCode.UNAUTHENTICATED
	details = "IAM token or API key has to be passed in request"
	debug_error_string = "UNKNOWN:Error received from peer ipv4:158.160.54.160:443 {created_time:"2024-08-20T07:41:11.58718279+00:00", grpc_status:16, grpc_message:"IAM token or API key has to be passed in request"}"
>.
WARNING:langchain_community.llms.yandex:Retrying langchain_community.llms.yandex.completion_with_retry.<locals>._completion_with_retry in 4.0 seconds as it raised _MultiThreadedRendezvous: <_MultiThreadedRendezvous of RPC that terminated with:
	status = StatusCode.UNAUTHENTICATED
	details = "IAM token or API key has to be passed in request"
	debug_error_string = "UNKNOWN:Error received from peer ipv4:158.160.54.160:443 {created_time:"2024-08-20T07:41:14.337100847+00:00", grpc_status:16, grpc_message:"IAM token or API key has to be passed in request"}"
>.
WARNING:langchain_community.llms.yandex:Retrying langchain_community.llms.yandex.completion_with_retry.<locals>._completion_with_retry in 8.0 seconds as it raised _MultiThreadedRendezvous: <_MultiThreadedRendezvous of RPC that terminated with:
	status = StatusCode.UNAUTHENTICATED
	details = "IAM token or API key has to be passed in request"
	debug_error_string = "UNKNOWN:Error received from peer ipv4:158.160.54.160:443 {grpc_message:"IAM token or API key has to be passed in request", grpc_status:16, created_time:"2024-08-20T07:41:19.077443222+00:00"}"
>.
WARNING:langchain_community.llms.yandex:Retrying langchain_community.llms.yandex.completion_with_retry.<locals>._completion_with_retry in 16.0 seconds as it raised _MultiThreadedRendezvous: <_MultiThreadedRendezvous of RPC that terminated with:
	status = StatusCode.UNAUTHENTICATED
	details = "IAM token or API key has to be passed in request"
	debug_error_string = "UNKNOWN:Error received from peer ipv4:158.160.54.160:443 {created_time:"2024-08-20T07:41:27.822608895+00:00", grpc_status:16, grpc_message:"IAM token or API key has to be passed in request"}"
>.
---------------------------------------------------------------------------
_MultiThreadedRendezvous                  Traceback (most recent call last)
[<ipython-input-11-91d22ce65899>](https://localhost:8080/#) in <cell line: 16>()
     14 country = "Russia"
     15 
---> 16 llm_chain.invoke(country)

22 frames
[/usr/local/lib/python3.10/dist-packages/grpc/_channel.py](https://localhost:8080/#) in _next(self)
    967                     raise StopIteration()
    968                 elif self._state.code is not None:
--> 969                     raise self
    970 
    971 

_MultiThreadedRendezvous: <_MultiThreadedRendezvous of RPC that terminated with:
	status = StatusCode.UNAUTHENTICATED
	details = "IAM token or API key has to be passed in request"
	debug_error_string = "UNKNOWN:Error received from peer ipv4:158.160.54.160:443 {grpc_message:"IAM token or API key has to be passed in request", grpc_status:16, created_time:"2024-08-20T07:41:44.933096923+00:00"}"
>

Description

I’m trying to use the langchain library to invoke a YandexGPT model. I expected to see a successful response from the model. Instead, I receive an StatusCode.UNAUTHENTICATED error with the message.

Previously, I encountered a similar issue where the error was AttributeError: 'YandexGPT' object has no attribute '_grpc_metadata'. I resolved this by modifying the _BaseYandexGPT class to make _grpc_metadata accessible in child classes, which fixed the issue. However, recent changes to the repository reverted my fix by reintroducing the _grpc_metadata parameter as protected and setting its default value to None.

This change has led to the current issue, where the authentication details are not being passed correctly, resulting in the StatusCode.UNAUTHENTICATED error. The original fix, which made _grpc_metadata accessible, resolved the issue and allowed the code to work as expected. Reverting these changes has introduced a new error without solving the original problem.

System Info

System Information
------------------
> OS:  Linux
> OS Version:  #1 SMP PREEMPT_DYNAMIC Thu Jun 27 21:05:47 UTC 2024
> Python Version:  3.10.12 (main, Jul 29 2024, 16:56:48) [GCC 11.4.0]

Package Information
-------------------
> langchain_core: 0.2.33
> langchain: 0.2.14
> langchain_community: 0.2.12
> langsmith: 0.1.99
> langchain_text_splitters: 0.2.2

Optional packages not installed
-------------------------------
> langgraph
> langserve

[mkhludnev]

may #25617 fix it?

[mkhludnev]

Hi @nktpv , could you please check if #25617 fixes this issue?

[nktpv]

Hi @nktpv , could you please check if #25617 fixes this issue?

yes, this PR contains a solution to the issue described

[mkhludnev]

@nktpv thanks for confirming.
Can you consult me about Pydantic? I have to drop leading underscore changing from private field, but what's the better approach?

[nktpv]

@nktpv thanks for confirming. Can you consult me about Pydantic? I have to drop leading underscore changing from private field, but what's the better approach?

I believe your solution is the best possible one, because to make the grpc_metadata field visible to subclasses, you can only either remove the protected access modifier or directly assign a value to the field through cls. Since the second option isn’t great, I think we can stay with your approach to solving the problem.