Implemented full support for Nano-GPT as a dynamic provider with three model list options:
- Canonical models list (all): 448 public models without authentication
- Subscription-only list: 221 models requiring subscription
- Personalized list: User's curated models (requires API key)

Key Features:
- Fetches detailed model info using ?detailed=true API parameter
- Displays real context windows, max tokens, and pricing per model
- Proper API key handling (only sent for personalized/subscription endpoints)
- Cache differentiation per model list type using enhanced cache keys
- Handles null values in API responses (context_length, cost_estimate)
- Model dropdown in chat window now shows all nano-gpt models
- Added user guidance text for refreshing model list after changing selection

Technical Changes:
- Updated Zod schema to accept nullable fields from API
- Enhanced modelCache.ts with composite cache keys (nano-gpt:all, etc.)
- Added nano-gpt case to useProviderModels hook for chat window
- Updated React Query keys to include nanoGptApiKey and nanoGptModelList
- Added nano-gpt to router models type definitions

Fixes:
- Schema validation errors for models with null context_length
- API key filtering causing only 221 models on "all" endpoint
- Chat window model dropdown not showing nano-gpt models
- Cache not invalidating when switching between model list types

…no-gpt

- Add nano-gpt provider configuration and settings
- Implement nano-gpt API integration for model fetching
- Add nano-gpt provider UI components

…no-gpt

…sageHandler tests