Torrent Bot

Self-hosted media stack: a Telegram AI assistant that accepts natural language commands, magnet links, and torrent files, downloads them via qBittorrent, and serves the result over SMB (Windows/Mac file sharing) and Jellyfin (streaming). The bot is powered by Google Gemini and controls qBittorrent and Jellyfin via MCP tools.

Services

Service	Port	Purpose
qBittorrent Web UI	8080	Torrent client + management
Jellyfin	8096	Media streaming (browser, TV apps)
Samba	445	SMB file share for PCs
Telegram bot	—	AI assistant: natural language, magnet links, .torrent files

Prerequisites

• Docker and Docker Compose
• A Telegram bot token from @BotFather
• A Google Gemini API key

Deployment

1. Clone the repository

git clone https://github.com/OksLo/torrentbot.git
cd torrentbot

2. Configure environment

cp .env.example .env

Edit .env and fill in:

BOT_TOKEN=your_telegram_bot_token_here
QBIT_USERNAME=admin
QBIT_PASSWORD=change_me
JELLYFIN_USERNAME=admin
JELLYFIN_PASSWORD=change_me
GEMINI_API_KEY=your_gemini_api_key_here
MCP_HTTP_TOKEN=change_me
TZ=Europe/London

Optionally set GEMINI_MODEL to a comma-separated list of models in priority order (e.g. gemini-2.5-pro,gemini-2.5-flash). The bot tries them in order and skips models that have hit their rate limit for 8 hours.

3. Start all services

make up

4. Wait for automatic setup to complete

On first boot, a setup container starts automatically once qBittorrent and Jellyfin are healthy. It sets passwords, creates the Jellyfin admin account, adds the Downloads library, and wires up autorun. Watch progress with:

make logs

Usage

Send your bot any natural language request — it uses Gemini AI with access to qBittorrent and Jellyfin tools. You can also send magnet links or .torrent files directly.

Examples:

• "Show active downloads"
• "What's been added to Jellyfin recently?"
• "Update the metadata for Inception"

Useful Commands

make up           # start all services (runs first-time setup automatically)
make down         # stop all services
make logs         # tail logs for all services
make bot-logs     # tail bot logs only
make restart-bot  # restart the bot container
make upgrade      # pull latest images and recreate services

GPU Hardware Transcoding (Intel)

If the host has an Intel GPU (/dev/dri present), make up automatically enables VA-API hardware transcoding in Jellyfin — no manual configuration needed. The Makefile detects the GPU at startup and applies the overlay; on hosts without /dev/dri the stack runs normally.

Connecting via SMB

The downloads/ directory is shared as \\<host-ip>\downloads (read-only, no password required).

• Windows: open File Explorer → address bar → \\<host-ip>\downloads
• macOS: Finder → Go → Connect to Server → smb://<host-ip>/downloads