Changes
PR #2138 by ksylvan: New Claude Fable model + cache OpenAI model discovery and handle provider rate limits
- Add persistent cache for provider model discovery results, improving performance and reliability of provider integrations.
- Serve stale model caches during discovery failures, ensuring continued operation when upstream providers are unavailable.
- Add Claude Fable 5 Anthropic model support, with sampling parameters automatically omitted for compatibility.
- Return concise localized errors for rate-limited model fetches, with updated translations across all supported locales.
- Update Go dependencies for AI provider integrations to keep upstream libraries current.