fix: bring back model selection

[k11kirky]

TL;DR

Added model selection capability to the task creation interface, allowing users to choose specific AI models when creating new tasks.

What changed?

• Added a model selector dropdown in the task input interface
• Implemented persistence of the last used model in settings
• Added model parameter to task creation flow
• Created filtering logic to show only models compatible with the selected adapter
• Updated the default gateway model from claude-opus-4-5 to claude-opus-4-6
• Added automatic model selection when switching adapters to ensure compatibility

How to test?

1. Open the task creation interface
2. Verify that a model selector dropdown appears next to the adapter selector
3. Select different adapters (Claude/Codex) and confirm that the model options update accordingly
4. Create tasks with different model selections and verify the selected model is used
5. Restart the application and verify that your last selected model is remembered

Why make this change?

This change gives users more control over their AI interactions by allowing them to explicitly select which model to use for each task. Previously, the system would automatically select a model based on the adapter, but now users can choose specific models with different capabilities or performance characteristics based on their needs.

[k11kirky]

• fix: bring back model selection #844 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[joshsny]

not sure if we need to check what was wrong with the GPT model output on Claude Code being empty?