fix: show only currently supported Groq models in the model dropdown picker

[pianoandglass]

Summary

The model dropdown was displaying models that are either no longer available on groq and also displaying models that don't support vision (image input) in the Context Model field. This fix removes unavailable models, introduces a dedicated visionModels list, and wires the context model picker to that list so only models that can actually process screenshots are offered.

Changes

Component	Details
ModelConfiguration.swift	Removed unavailable Groq models (allam-2-7b, canopylabs/orpheus-*, meta-llama/llama-prompt-guard-2-*) from llmModels. Added visionModels array containing only models that accept image input (qwen/qwen3.6-27b).
SettingsView.swift	Context model picker now uses ModelConfiguration.visionModels instead of llmModels. Updated subtitle to clarify that screenshot analysis requires a vision-capable model.

• Custom model typed by user: The picker still allows free-text input; users who type a vision-capable model not in visionModels will continue to work as before — the list is suggestive, not restrictive.
• Existing user config with a now-removed model: The stored model string is preserved unchanged; only the suggestions list shrinks. No data migration needed.

Summary by CodeRabbit

• New Features

  • Updated the Context Model setting to clearly indicate that screenshot analysis requires an image-capable model.
  • The Context Model dropdown now lists vision-capable models, including newly supported options.

• Improvements

  • Refined the available vision model selection to provide a more focused set of compatible choices.
  • Improved fallback descriptions and model guidance for clearer configuration.

[coderabbitai]

📝 Walkthrough

Walkthrough

The context model dropdown now presents the configured vision model list and explains its requirement for screenshot analysis. The vision model catalog was reduced, while configuration logic remained unchanged.

Changes

Context model selection

Layer / File(s)	Summary
Vision model catalog and configuration comments Sources/ModelConfiguration.swift	The vision model list now contains a reduced set including Qwen and Groq compound variants; nearby comments were revised without changing configuration logic.
Context model dropdown wiring Sources/SettingsView.swift	The Context Model dropdown now uses ModelConfiguration.visionModels, and its subtitle states that screenshot analysis requires an image-capable model.

Estimated code review effort: 2 (Simple) | ~10 minutes

Possibly related PRs

• zachlatta/freeflow#91: Introduces the configurable context-model flow used by this dropdown.
• zachlatta/freeflow#222: Also updates the context model dropdown and model configuration lists.

Suggested reviewers: marcbodea, iris-sfg

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly matches the main change: restricting the model dropdown to currently supported Groq image-capable models.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}