feat(model-profiles): add input/output MIME type fields to ModelProfile

[open-swe]

Adds informational input_mime_types and output_mime_types dict fields to ModelProfile, keyed by models.dev modality names ('image', 'audio', 'pdf', 'video'). Values are lists of IANA media types (e.g. 'image/png'). Both are informational — presence implies a known-good MIME type, but absence is not a guarantee of rejection.

The CLI in langchain-model-profiles previously used isinstance(value, dict) under [overrides] to decide "is this a provider-level field or a model id?". That heuristic breaks once a provider-level field's value is itself a TOML table (as the new MIME maps are). The loader now dispatches against the declared ModelProfile field names: keys matching declared fields are provider-level; anything else is treated as a model id.

profile_augmentations.toml for anthropic, openai, and perplexity is backfilled with documented MIME types, and the generated _profiles.py for each is regenerated via the refresh CLI.

Release Note

Added input_mime_types and output_mime_types fields to model profiles for OpenAI, Anthropic, and Perplexity integrations.

Disclaimer: opened collaboratively by Mason Daugherty and open-swe.

Opened collaboratively by Mason Daugherty and open-swe.

[codspeed-hq]

Merging this PR will not alter performance

✅ 13 untouched benchmarks
⏩ 2 skipped benchmarks¹

---

_{Comparing open-swe/profile-mime-types (988ee53) with master (2086b91)²}

Footnotes

1. 2 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

2. No successful run was found on master (bbd10fe) during the generation of this report, so 2086b91 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩