ChatBedrock not working with latest nova models (amazon.nova-pro-v1:0 model)

[aashipandya]

I am trying to use the new Amazon Bedrock model amazon.nova-pro-v1:0 with the ChatBedrock integration provided by LangChain-AWS. However, I encounter a validation error when I attempt to invoke the model.

Code to Reproduce
import boto3
from langchain_aws import ChatBedrock
import os

AWS credentials -
aws_access_key = "XYZ"
aws_secret_key = "XYZ"
region_name = "us-east-1"

Bedrock Runtime Client -
model_name = "amazon.nova-pro-v1:0"
bedrock_client = boto3.client(
service_name="bedrock-runtime",
region_name=region_name,
aws_access_key_id=aws_access_key,
aws_secret_access_key=aws_secret_key
)
llm = ChatBedrock(client=bedrock_client, model_id=model_name, model_kwargs=dict(temperature=0))
response = llm.invoke("Create a list of 3 pop songs")
print(response)

Error Output
ValidationError: 1 validation error for ChatBedrockConverse
Value error, Could not load credentials to authenticate with AWS client.
Please check that credentials in the specified profile name are valid.

Although If i try same credentials & same model with client.converse method it works.

response = bedrock_client.converse(
modelId=model_name,
messages=messages,
inferenceConfig={"temperature": 0.0}
)

Environment
• LangChain-AWS version: 0.2.9
• Python version: 3.10
• AWS SDK (boto3) version: 1.35.83

I am not sure if the issue is with credentials or ChatBedrock.

[mlrana]

I have also tried chatbedrock with nova models
throwing me this error
An error occurred (ValidationException) when calling the Converse operation: Invocation of model ID amazon.nova-micro-v1:0 with on-demand throughput isn’t supported. Retry your request with the ID or ARN of an inference profile that contains this model.

Though chatbedrock is working with claude models.

[anujj-ti]

I am also facing similar issues.

It frequently gives this error:

An error occurred (ModelErrorException) when calling the Converse operation: The system encountered an unexpected error during processing. Try your request again.

My code snippet:

            client = ChatBedrock(
                model_id=model_id,
                temperature=temperature,
                max_tokens=max_tokens,
            )

            if tools:
                client = client.bind_tools(
                    tools=[tool.get_schema() for tool in tools],
                    tool_choice=tool_choice
                )

            response: AIMessage = await client.ainvoke(messages)

[jnsaff]

I had to enable the models in AWS Console. After that the invoke part works just fine.

Running into issues with embeddings though.

[clouddev-code]

I tried the following code.

boto3 = "^1.35.93"
langchain-aws = "^0.2.10"

from langchain_aws import ChatBedrockConverse

llm = ChatBedrockConverse(
    model_id="amazon.nova-pro-v1:0",
    temperature=0,
    max_tokens=None,
    region_name="us-east-1"
    # other params...
)

messages = [
    (
        "system",
        "You are a helpful assistant that translates English to French. Translate the user sentence.",
    ),
    ("human", "I love programming."),
]

for chunk in llm.stream(messages):
    print(chunk)

I could get success response.

content=[] additional_kwargs={} response_metadata={} id='run-9d3f8e2b-70ab-4529-a0ce-833a050ac0e1'
content=[{'type': 'text', 'text': 'J', 'index': 0}] additional_kwargs={} response_metadata={} id='run-9d3f8e2b-70ab-4529-a0ce-833a050ac0e1'
content=[{'type': 'text', 'text': "'", 'index': 0}] additional_kwargs={} response_metadata={} id='run-9d3f8e2b-70ab-4529-a0ce-833a050ac0e1'
content=[{'type': 'text', 'text': 'ado', 'index': 0}] additional_kwargs={} response_metadata={} id='run-9d3f8e2b-70ab-4529-a0ce-833a050ac0e1'
content=[{'type': 'text', 'text': 're', 'index': 0}] additional_kwargs={} response_metadata={} id='run-9d3f8e2b-70ab-4529-a0ce-833a050ac0e1'
content=[{'type': 'text', 'text': ' la', 'index': 0}] additional_kwargs={} response_metadata={} id='run-9d3f8e2b-70ab-4529-a0ce-833a050ac0e1'
content=[{'type': 'text', 'text': ' program', 'index': 0}] additional_kwargs={} response_metadata={} id='run-9d3f8e2b-70ab-4529-a0ce-833a050ac0e1'
content=[{'type': 'text', 'text': 'mation', 'index': 0}] additional_kwargs={} response_metadata={} id='run-9d3f8e2b-70ab-4529-a0ce-833a050ac0e1'
content=[{'type': 'text', 'text': '.', 'index': 0}] additional_kwargs={} response_metadata={} id='run-9d3f8e2b-70ab-4529-a0ce-833a050ac0e1'
content=[{'type': 'text', 'text': '', 'index': 0}] additional_kwargs={} response_metadata={} id='run-9d3f8e2b-70ab-4529-a0ce-833a050ac0e1'
content=[{'index': 0}] additional_kwargs={} response_metadata={} id='run-9d3f8e2b-70ab-4529-a0ce-833a050ac0e1'
content=[] additional_kwargs={} response_metadata={'stopReason': 'end_turn'} id='run-9d3f8e2b-70ab-4529-a0ce-833a050ac0e1'
content=[] additional_kwargs={} response_metadata={'metrics': {'latencyMs': 264}} id='run-9d3f8e2b-70ab-4529-a0ce-833a050ac0e1' usage_metadata={'input_tokens': 20, 'output_tokens': 8, 'total_tokens': 28}

[3coins]

@aashipandya
Based on the conversation so far, it seems like this could be either an access or a transient service issue. Are you still facing any issue using Nova models with BedrockConverse?

[kaustubh-darekar]

The issue is resolved with addition of region_name in ChatBedrock method
llm = ChatBedrock(client=bedrock_client,region_name=region_name, model_id=model_name, model_kwargs=dict(temperature=0))