feat: improve Ollama local model configuration UX

[ParthSareen]

Summary

• Auto-populate default endpoint URL when Ollama is selected
• Fetch and display available models from Ollama API
• Add model dropdown with refresh button for Ollama

Changes

• Auto-populate http://localhost:11434/v1 for Ollama (unless user manually edits)
• Fetch models from /api/tags endpoint on page load and platform switch
• Replace text input with Select dropdown for Ollama model selection
• Add refresh button to re-fetch available models
• Sort model list with configured model at top

Test plan

• Fresh load with Ollama selected → URL auto-populated, models fetched
• Switch to vLLM → URL not auto-populated, text input shown
• Switch back to Ollama → URL restored, models dropdown shown
• Click refresh button → Models re-fetched
• Manually edit URL → Preserved when switching platforms
• Click Reset → URL reset to default, models re-fetched

[a7m-1st]

Hi there @ParthSareen, can you link an issue regarding this and your scope?
Additionally, would be cool if you can provide screenshots of the frontend fields if updated.

We are looking to resolve custom model config too like top-p and others:
https://discord.com/channels/1115015097560076329/1463396464557555891

So if its related maybe we can collab !

[Wendong-Fan]

hey @ParthSareen

The main branch was refactored to use per-platform state instead of single values. I rebased onto main and adapted all Ollama-specific features to work with the new structure