Skip to content

agent_examples.py using UI-TARS and OmniParser #365

@cy3021561

Description

@cy3021561

Hi, I was trying to use this script to test different models, but only OpenAI CUA and Anthropic CUA worked successfully.

When I tested UI-TARS (deployed via TGI on HF, following this guide
) and Omniparser + [other thinking model], both failed.

For UI-TARS as a computer-use model, it failed right after the first turn:
raise APIError( litellm.exceptions.APIError: litellm.APIError: HuggingfaceException - Failed to deserialize the JSON body into the target type: messages[2]: data did not match any variant of untagged enum MessageBody at line 1 column 1985 LiteLLM Retried: 3 times

For the composed model, it seems like the message handling between components didn’t work as expected:

raise BadRequestError( litellm.exceptions.BadRequestError: litellm.BadRequestError: OpenAIException - Missing required parameter: 'tools[0].function'. LiteLLM Retried: 3 times

I’ve been trying to fix this by reading the docs for days, but I still can’t get it working.
Has anyone else run into similar issues? Or is there another test example I could follow?

Any suggestions would be helpful, thanks!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions