What's Changed
- fix: use max_tokens for non-OpenAI providers (Mistral, Ollama) (#319) — Fixes HTTP 422 errors when using Mistral, Ollama, or other OpenAI-compatible LLM providers. The app now sends the correct
max_tokensparameter for third-party providers instead of the OpenAI-specificmax_completion_tokens. - UI: sidecast debug tool tweaks (#316) — Added demo mode toggle and custom model picker to the sidecast debug panel; demoted export button to outline style; reordered settings panel.