Immaculaterr === [](https://github.com/ohmzi/Immaculaterr/actions/workflows/publish-containers.yml) [](https://github.com/ohmzi/Immaculaterr/releases) [](../LICENSE) [](https://github.com/ohmzi/Immaculaterr/pkgs/container/immaculaterr) [](https://hub.docker.com/r/ohmzii/immaculaterr) Immaculaterr is a **Plex “autopilot”** that watches what you’re watching, generates recommendations, and keeps your library tidy. It builds curated Plex collections (with proper artwork), can optionally send missing titles to Radarr/Sonarr or Seerr, and gives you detailed run reports so you always know what happened. Official Docker images and release artifacts are the supported public distribution channel. The public source repository remains visible for transparency and reference, but the source code is not licensed for general reuse. Major Features Include --- - **Seerr integration (optional centralized request flow)**: - Route missing movie/TV requests to Seerr instead of direct ARR sends. - Works per task card, so you can choose where missing titles go. - Includes Command Center reset control to clear all Seerr requests when needed. - **Plex-triggered automation**: - Finish a movie or episode, and Immaculaterr can turn that watch into fresh collections and great new recommendations right away. - **Scheduler automation**: - Off-peak schedules keep those rows fresh with background refresh, discovery, cleanup, and maintenance. - **Shared persisted job queue**: - Manual runs, schedules, Plex webhooks, and Plex polling all go through one persisted FIFO queue. - Rewind shows queued work, live progress, reports, logs, and run history. - **Managed Plex collections**: - `Based on your recently watched Movie` and `Based on your recently watched Show` - `Change of Movie Taste` and `Change of Show Taste` - `Inspired by your Immaculate Taste in Movies` and `Inspired by your Immaculate Taste in Shows` - `Fresh Out Of The Oven` and `Fresh Out Of The Oven Show` - `Netflix Import Picks` and `Netflix Import: Change of Taste` - `Plex History Picks` and `Plex History: Change of Taste` - Immaculate Taste profiles can also create extra custom-named collections. - **Immaculate Taste profiles**: - Give each profile its own users, media type, filters, collection names, and download route. - Use genre and audio-language filters to build focused lanes like animation, family, or specific-language picks. - **Recommendation engine**: - TMDB is the main source. - Google and OpenAI are optional helpers for widening or refining results. - You can tune the mix between titles available now and future releases. - **Download routing**: - Send missing titles directly to Radarr and Sonarr, or route them through Seerr on a per-task basis. - Observatory can hold items for swipe approval before they are sent. - **Discovery and maintenance jobs**: - `Fresh Out Of The Oven` builds recent-release movie and TV rows for titles a user has not watched yet. - `TMDB Upcoming Movies` finds upcoming movies with filter sets and routes matches to Radarr or Seerr. - `Rotten Tomatoes Upcoming Movies + TV Shows` scrapes fixed Rotten Tomatoes movie and TV pages, uses saved Movies and TV toggles with separate Movie Top 20 and TV Top 10 counts, and lets Run Now use those saved settings directly. - Cleanup and ARR sync jobs help keep Plex, Radarr, and Sonarr tidy after imports and downloads, including season-aware Sonarr monitoring cleanup. - **History imports**: - Netflix CSV import creates dedicated Netflix import collections and feeds the main recommendation system. - Plex watch-history import does the same without needing a CSV. - **Custom posters for managed collections**: - Upload and manage poster overrides in Command Center. - Posters are stored so they survive restarts and updates. - **Built-in admin sign-in and recovery**: - Create the admin login during setup. - Password recovery uses security questions. - **Management pages**: - Use Vault for integrations, Task Manager for jobs, Rewind for reports, and Command Center for resets, posters, user monitoring, and request cleanup. - **More features on the way:** - Email reports on media server health - Windows and macOS support How Recommendations Are Built --- 1. A watch event, manual run, or history import supplies a seed title, and the app builds a richer seed profile from it. 2. TMDB pulls fuller metadata and candidate pools, including standard picks plus wildcard lanes for global-language films and hidden gems. 3. A multi-factor ranking engine scores candidates using similarity, quality, novelty, and indie/popularity signals. 4. Ranking weights change by intent, so latest-watched and change-of-taste runs do not rank titles the same way, and released vs. upcoming mixes can be tuned separately. 5. Final picks are interleaved so core recommendations stay strong while wildcard discoveries add variety. Screenshots --- **Desktop UI**