Tool Modality

[Jeomon]

Just like input and output modalities, please also include the tool result modality, because, for example, Gemini models support file,image, video, and audio as tool results, while the Anthropic models support text and image, and the Openai models support text only.

[github-actions]

This is a feature request to track a new kind of metadata (tool-result modalities), not a concrete catalog data fix for existing model/provider fields. It needs maintainer review for schema/product changes; no automated fix was opened.