Failed to Load AWS Credentials for Authentication in Virtual Environment

[harshakoneru7]

When upgrading to langchain-aws==0.2.31, an error occurs during AWS client authentication. The issue is specific to virtual environments and does not appear when using version 0.2.30. The error message indicates that credentials could not be loaded to authenticate with the AWS client, and a region must be specified.

Error details

1 validation error for ChatBedrock Value error, Could not load credentials to authenticate with AWS client. 
Please check that the specified profile name and/or its credentials are valid. 
Service error: You must specify a region. [type=value_error, input_value={'client': <botocore.clie...se_converse_api': False}, input_type=dict]
For further information visit [https://errors.pydantic.dev/2.11/v/value_error]

Workaround

Downgrading to langchain-aws==0.2.30 resolves the issue, as the error does not occur in this version. Alternatively, explicitly specifying the region in the ChatBedrock configuration resolves the problem, as the component fails to retrieve the region from the ~/.aws/config file in version 0.2.31.

[michaelnchin]

Hi @harshakoneru7, thank you for filing an issue.

I'm unable to reproduce the same issue using ChatBedrock locally, with langchain-aws==0.2.31 installed in a fresh environment. For reference, my ~/.aws/config file looks like this:

[default]
region = us-east-1
output = json

I'm not too certain what the root cause of your error could be, as the AWS client generation logic was not modified between this release and v0.2.30.

If possible, could you provide a self-contained reproducer code sample, in addition to the stack trace from the error?

[vikrant-mah-dhimate]

@michaelnchin Please find the details below with sample program and output.
cc: @harshakoneru7

langchain-aws==0.2.31
pydantic==2.11.7
pydantic-settings==2.10.1
pydantic_core==2.33.2

import asyncio
import boto3
from langchain_aws import ChatBedrock

# --- Configuration ---
# Replace these placeholder values with your own.
MODEL_ID = "us.meta.llama3-3-70b-instruct-v1:0"
AWS_REGION = "us-east-1"
OUTPUT_GUARDRAIL_ID = "your-guardrail-id"
OUTPUT_GUARDRAIL_VERSION = "DRAFT"

CONFIG = {
    "temperature": 0,
    "max_tokens": 5000,
    "top_p": 1,
}


# --- Main Program Logic ---

async def run_query():
    """Runs a single query against the Bedrock model with guardrails."""
    try:
        # Get the boto3 client using a Session object
        session = boto3.Session(profile_name="test_profile")
        client = session.client("bedrock-runtime", region_name=AWS_REGION)

        # Create and configure the ChatBedrock instance in one step
        llm = ChatBedrock(
            client=client,
            model_id=MODEL_ID,
            model_kwargs=CONFIG,
            # After uncommenting below line, the code works as expected.
            # region_name="us-west-2"  
        )

        # Define the sample input
        sample_input = "What is the capital of France?"

        print(f"Sending prompt to Bedrock: '{sample_input}'")

        # Execute the query and print the response
        response = await llm.ainvoke(sample_input)

        print("\n--- Response from LLM ---")
        print(response.content)

    except Exception as e:
        print(f"\nAn error occurred: {e}")

# Run the asynchronous function
if __name__ == "__main__":
    asyncio.run(run_query())

Output

---------------------------------------------------
Successfully installed langchain_aws-0.2.31
---------------------------------------------------
- python3 test.py
Sending prompt to Bedrock: 'What is the capital of France?'

--- Response from LLM ---
The capital of France is Paris.

---------------------------------------------------
(After commenting region)
---------------------------------------------------
- python3 test.py

An error occurred: 1 validation error for ChatBedrock
  Value error, Could not load credentials to authenticate with AWS client. Please check that the specified profile name and/or its credentials are valid. Service error: You must specify a region. [type=value_error, input_value={'client': <botocore.clie...se_converse_api': False}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.11/v/value_error

----------------------------------------------------
Successfully installed langchain_aws-0.2.30 (commented code)
----------------------------------------------------
- python3 test.py
Sending prompt to Bedrock: 'What is the capital of France?'

--- Response from LLM ---
The capital of France is Paris.

[michaelnchin]

Thank you @vikrant-mah-dhimate, I was able to reproduce with your sample.

langchain-aws v0.2.31 introduced some changes (see: #561) where ChatBedrock and ChatBedrockConverse now also initialize a separate bedrock client instance, controllable via the bedrock_client parameter.

Currently, this client does not have visibility into any configuration set for the original bedrock-runtime client, and so it will fail if it cannot find region_name via the class attribute or environmental variables.

I've opened a PR (#592) to add the bedrock-runtime client's region value as a fallback in such cases. As a workaround for now, you can pass the region_name model parameter directly, or inside of bedrock_client, ex.

client = session.client("bedrock-runtime", region_name=AWS_REGION)
bedrock_client = session.client("bedrock", region_name=AWS_REGION)

# Create and configure the ChatBedrock instance in one step
llm = ChatBedrock(
    client=client,
    bedrock_client=bedrock_client,
    ...
)