Discovarr - AI Powered Media Recommendations

Description

Discovarr is a comprehensive media management and automation tool designed to streamline your media consumption and discovery experience. It intelligently integrates with popular media servers like Jellyfin and Plex, download clients Radarr and Sonarr, and leverages the power of Google's Gemini AI to provide personalized media recommendations.

With Discovarr, you can:

• Automatically track your watch history from Jellyfin and Plex.
• Get intelligent media suggestions based on your viewing habits and preferences.
• Easily request new movies and TV shows through Radarr and Sonarr.
• Manage and customize search prompts for AI-driven recommendations.
• Schedule automated tasks for syncing history and processing suggestions.
• View and manage your suggestions, including active and ignored items.

Screenshots

Supported Providers

• Library:
  • Jellyfin
  • Plex
  • Trakt.tv (Watch History Sync Only)
• Request:
  • Radarr (Movies)
  • Sonarr (TV Shows)
  • Jellyseerr
  • Overseer
• LLM:
  • Google Gemini
  • OpenAI
    • Supports all services that adhere to the OpenAI API spec. Requires the Responses API for structured output.
    • Continue to use the separate Ollama provider until they add the Responses API. See ollama/ollama#9659
  • Ollama (for local models)

Key Features

• Media Server Integration:
  • Seamlessly connect to Jellyfin and Plex to fetch user lists, watch history, and library content.
  • Sync watch history across multiple users from both platforms.
  • Support for Trakt.tv to sync watch history.
• Downloader Integration:
  • Integrate with Radarr for movie requests and Sonarr for TV show requests.
  • Supports Jellyseerr and Overseerr.
  • Manage quality profiles for downloads.
• AI-Powered Recommendations:
  • Utilize Google Gemini and OpenAI to generate personalized movie and TV show recommendations.
  • Customize prompts and exclude existing media or ignored titles from suggestions. Prompts support templating with dynamic variables.
  • Leverage your "favorites" from Jellyfin/Plex to influence recommendations.
  • Support for local AI models via Ollama.
  • Integrates with TMDB API to pull metadata for suggested title.
• Automated Workflows:
  • Schedule job to automatically sync watch history.
  • Schedule jobs to process watch history and generate new media suggestions.
  • Manually trigger scheduled jobs as needed.
  • Uses APScheduler and supports cron syntax.
• Search & Discovery:
  • Save and manage custom search prompts for AI recommendations.
  • Run saved searches to get fresh suggestions.
  • Preview rendered prompts before execution.
  • Schedule custom searches to run automatically.
• User Interface:
  • A FastAPI-based backend provides a robust API for all functionalities.
  • A frontend UI to interact with the system. Based on Vite + Vue.js + Tailwind CSS.
• Configuration & Settings:
  • Manage all service integrations and application settings through a persistent database configuration.
  • Settings can be overridden by environment variables.
• Database & Backup:
  • Use SQLite or PostgreSQL for data persistence via the Peewee ORM.
  • Automatic database backups before schema migrations/upgrades.
• Monitoring & Stats:
  • Track token usage for searches.
  • View watch history grouped by user and date range.

Prerequisites

• Jellyfin URL and API key (or other Library Provider)
  • Can use docker container name in the URL if hosted on the same docker network
• Sonarr/Radarr URL and API key (or other Request Provider)
  • Can use docker container name in the URL if hosted on the same docker network
• Gemini API key (or other LLM Provider)
• TMDB API key

Background

The original intent was to write a script to generate a prompt that gave me recommendations that were not in my media library. After I got that working I decided to turn this into a full application. Hope you find it useful as well!

Scheduled Tasks

• The sync_watch_histry job that runs at midnight every night to sync the watch history for each user to the watchhistory table. This job can be triggered manually on the Watch History tab. It uses the app.recent_limit setting to control how many items for each user to sync.
• The recently_watched default search queries the watchistory table, returing a set number of items based on the app.recent_limit setting. It then submits a prompt to the Gemini API using the prompt associated with the search. The {{limit}} prompt template variable is set using the gemini.limit setting. Results are saved to the media table and displayed on the Home tab. The default search is not scheduled initially.

Search

• Custom searches can be created and run ad-hoc, results are displayed and can be automatically saved with the app.auto_media_save setting. Saved searches are saved to the media table automatically.
• Searches support various template variables like {{media_name}} {{limit}} {{all_media}} {{favorites}} that will be auto-populated at runtime. {{media_name}} is used to populate the Source column linking recommendations to the source title. Use the Preview button to preview the generated prompt before submitting.
• The Search button next to items in Watch History or Home will navigate to the Search page and auto-populate the {{media_name}} variable accordingly.

Possible Roadmap

• Integration with other AI providers like OpenAI, Claude, etc.
  • Gemini was choosen due to their generous free tier.
• Overseer and Jellyseer integration.
  • I don't use these but would be willing to consider supporting them.
• SmartCollections using LLM directly or potentially a RAG vector search with LangFlow

Docker Compose Example

• This example compose.yml demonstrates the minimum config for the Discovarr container.
• Environment variables can be set according to your specific setup and will update populate Settings. ENVs will always overwrite Settings in the database.
• Volumes used in the container are /config and /backups. Map /config to a path on your host to persist the database between container restarts.
• The container does not run as root and uses a discovarr user and group with UID/GID 1884 by default. The entrypoint.sh script will create this user/group within the container and ensure proper directory permissions.
• If you want to change the 1884 UID/GID use the PUID and PGID environment variables.

services:
  discovarr:
    image: ghcr.io/sqrlmstr5000/discovarr:latest 
    container_name: discovarr
    restart: unless-stopped
    ports:
      - "8000:8000" 
    environment:
      # Client needs to know where the API is. This will be your host machine IP or hostname since the client is connecting from your browser
      - VITE_DISCOVARR_URL=http://192.168.0.100:8000/api
    volumes:
      - ./config:/config
      - ./cache:/cache

Environment Variables

• These are optional and can be set in the UI
• If specified will overwrite existing settings

- PUID=1884
- PGID=1884
- LOGLEVEL=INFO # DEBUG, INFO, WARNING, ERROR, CRITICAL
- APP_DEFAULT_PROMPT="Your custom default prompt here"
- APP_RECENT_LIMIT=10
- APP_SUGGESTION_LIMIT=10
- APP_TEST_MODE=False
- APP_BACKUP_BEFORE_UPGRADE=True
- APP_AUTO_MEDIA_SAVE=True
- APP_SYSTEM_PROMPT="Your custom system prompt for Gemini"

# Client needs to know where the API is. This will be your host machine IP or hostname since the client is connecting from your browser
- VITE_DISCOVARR_URL=http://192.168.0.100:8000/api

# --- Jellyfin Settings ---
- JELLYFIN_ENABLED=True # Set to True if using Jellyfin
- JELLYFIN_URL=http://jellyfin:8096
- JELLYFIN_API_KEY=your_jellyfin_api_key

# --- Plex Settings ---
- PLEX_ENABLED=True # Set to True if using Plex
- PLEX_URL=http://plex:32400
- PLEX_API_KEY=your_plex_token

# --- Radarr Settings ---
- RADARR_URL=http://radarr:7878
- RADARR_API_KEY=your_radarr_api_key
- RADARR_DEFAULT_QUALITY_PROFILE_ID=1 # Example ID
- RADARR_ROOT_DIR_PATH=/movies

# --- Sonarr Settings ---
- SONARR_URL=http://sonarr:8989
- SONARR_API_KEY=your_sonarr_api_key
- SONARR_DEFAULT_QUALITY_PROFILE_ID=1 # Example ID
- SONARR_ROOT_DIR_PATH=/tv

# --- Gemini AI Settings ---
- GEMINI_API_KEY=your_gemini_api_key
- GEMINI_MODEL=gemini-pro # Or your preferred Gemini model
- GEMINI_THINKING_BUDGET=1024
- GEMINI_TEMPERATURE=0.7

# --- TMDB Settings ---
- TMDB_API_KEY=your_tmdb_api_key

API Documentation

Swagger docs available at http://host:discovarr_port/api/docs

Style

Logo
#FBBF24