Major Features
-
Structured Output Option 🆕
- New toggle for models that support structured output
- Defines a JSON structure for model responses
- Makes recommendation parsing significantly more consistent compared to text-based formatting
-
TMDB Integration 🎬
- Added optional TMDB integration with api key.
- New info modal with detailed TMDB data when clicking recommendation cards
- Now prefers TMDB for posters and metadata (much faster than Sonarr/Radarr requests)
- History page now uses TMDB data when connected
-
Radarr/Sonarr No Longer Required
- You can now get recommendations so long as at least one media service is connected (Sonarr/Radarr, Plex, Tautulli, Jellyfin, Trakt)
UI Enhancements
- A more refined general theme.
- Recommendations now appear as they're returned from the LLM
- Added automatic compact cards mode when the screen can't fit standard poster layout
- Various UI cleanup and improvements
- PWA support added
- This means the app can be installed to your phone screen or desktop like an application from your browser.
Bug Fixes
- Fixed major bug causing inconsistent filtering of results
- Library/liked/watched content should no longer appear in recommendations
- Fixed issue with failed logins after expired sessions
- Fixed connection issues with Ollama, LM Studio and Google endpoints
- Consistent clearing of conversation history when setting new options
- Fixed bug that could cause users to see the "Configure AI Service" upon initial login even with a saved connection.
Other Improvements
- Adjusted prompt chunking token limit
- Increased watch history limit from 100 to 2000 items
TMDB Info Modal (When clicking a recommendation card)
Full Changelog: 1.2.70...1.2.80