Skip to content

Compatibility Issue with External Multimodal Models (Mistral, Gemini) #27

@Ananya-yash

Description

@Ananya-yash

Hello, I have been trying to evaluate modern multimodal models using your framework. I have tested gemini-1.5-flash (from Google AI) and mistralai/mistral-small-3.1-24b-instruct:free (from OpenRouter).

In both cases, the API call is successful, but the evaluation fails with the error: Error parsing tool call: 'image'.

This suggests that the project's agent logic cannot parse the tool-calling output format for image-related tasks from these models. The parsing seems to be tightly coupled to the original Qwen model's output format.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions