[BUG]

[btobio-switchfly]

Description

I'm having issues using a StringKnowledgeSource with a BedRock model.

Steps to Reproduce

Code below

Expected behavior

no errors, crew should run

Screenshots/Code snippets

ks = StringKnowledgeSource(content=knowledge_string)

session = boto3.Session(
    aws_access_key_id=os.getenv('AWS_ACCESS_KEY_ID'),
    aws_secret_access_key=os.getenv('AWS_SECRET_ACCESS_KEY'),
    aws_session_token=os.getenv('AWS_SESSION_TOKEN'),
    region_name=os.getenv('AWS_REGION_NAME')
)

crew = Crew(
    agents=[agent],
    tasks=[task],
    verbose=False,
    knowledge_sources=[ks],
    embedder={
        "provider": "bedrock",
        "config": {
            "model": "bedrock/amazon.titan-embed-text-v2:0",
            "session": session,
        }
    }
)

Operating System

macOS Sonoma

Python Version

3.11

crewAI Version

0.114.0

crewAI Tools Version

0.42.0

Virtual Environment

Venv

Evidence

Getting the following error:
2025-04-25 13:37:46][ERROR]: Failed to upsert documents: ClientError.__init__() missing 1 required positional argument: 'operation_name'

Possible Solution

Not sure

Additional context

None

[btobio-switchfly]

Reference issue: #2299

[btobio-switchfly]

Is there an ETA for this fix?

[Vidit-Ostwal]

@btobio-switchfly, I am on it.
Can you use a devin's PR once, to send a operation_name.
To check whether it's a right fix?

From what I understand, another exception is being raised which is causing the issue.

[Vidit-Ostwal]

I checked this boto3 documentation, I don't see any operation_name, as a parameters

client(service_name, region_name=None, api_version=None, use_ssl=True, verify=None, endpoint_url=None, aws_access_key_id=None, aws_secret_access_key=None, aws_session_token=None, config=None, aws_account_id=None)[[source]](https://boto3.amazonaws.com/v1/documentation/api/latest/_modules/boto3/session.html#Session.client)

[btobio-switchfly]

Yes, that's exactly why the error doesn't make sense

[Vidit-Ostwal]

@btobio-switchfly, can you force update the chroma db version and check once?

[lucasgomide]

@btobio-switchfly would you mind sharing the stack trace

[btobio-switchfly]

@Vidit-Ostwal I updated chromadb and embed chain along with removing the bedrock/ before the model name and that worked. Not sure if removing bedrock/ would have solved the issue on it's own.

[Vidit-Ostwal]

@Vidit-Ostwal I updated chromadb and embed chain along with removing the bedrock/ before the model name and that worked. Not sure if removing bedrock/ would have solved the issue on it's own.

I am not sure how did this worked out.

Can you share the following

1. Chromadb version
2. Embedchain version
3. Crew along with agents
4. The stack trace, when you were first facing this error.

[btobio-switchfly]

I'll give a summary of the issue. So the operation_name error wasn't the actual error, was just at the bottom of the stack trace. When applying knowledge source and embedder in the crew, only the last error of the stack trace (that is the operation_name error) was shown. However, once I moved the knowledge source and embedder to the agent, the full stack trace for the error was shown.

[lucasgomide]

@btobio-switchfly till not totally clear to me what you made it work. Was the issue addressed by moving knowledge and embedder from a Crew to Agent or upgrading your dependencies?

Your answer would be really valuable for us!