* LLM - Dont import litellm library until a valid config is entered
* LLM - "Test connect" runs in clean subprocess for better memory management
* LLM - Add missing model_list module and tests
The /settings/llm/models route imports changedetectionio.llm.model_list,
which lists provider models over plain HTTP instead of importing litellm.
It and its unit tests were left untracked in the earlier commits.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* LLM - Upgrade litellm to >=1.106.2, add DeepSeek, Groq, Mistral, Together AI and xAI providers (#4597)
* Update litellm requirement from >=1.40.0 to >=1.106.2
Supersedes #4595.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* LLM - Add DeepSeek, Groq, Mistral, Together AI and xAI to the provider list
All five serve OpenAI's GET /models, so "Load available models" works for
them via model_list without importing litellm. Together's bare-list response
and its image/embedding models are handled, and Mistral's {"detail": ...}
errors are surfaced.
Also re-select Anthropic on page load for saved 'anthropic/claude-...' models,
the form the model list returns.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
