SummaryGram

A Dockerized Telegram bot that summarize last messages in a group

Features

• 🤖 Telegram bot interface
• ⚙️ Command-based interaction
• 📝 Summarize recent messages in group chats
• 🧠 AI-powered summarization
• 🔌 Customizable LLM integration (OpenAI, Ollama, Cloudflare AI, llama.cpp)
• 🎤 Voice message transcription using whisper.cpp or Cloudflare AI Whisper
• 🐳 Docker containerized for easy deployment

Environment Variables

Variable	Description	Default
TELEGRAM_BOT_TOKEN	Your Telegram bot token (required)	-
WHITELISTED_CHATS	Comma-separated list of allowed chat IDs (optional)	-
LLM_PROVIDER	Explicit LLM provider selection (optional). Valid options: openai, cloudflare, ollama, llama.cpp. If not set, auto-detects based on configured credentials (priority: OpenAI, Cloudflare, Ollama, llama.cpp).	-
OPENAI_API_KEY	Your OpenAI API key (optional, for OpenAI integration)	-
OPENAI_BASE_URL	Custom OpenAI API base URL (optional, for OpenAI-compatible APIs)	-
OPENAI_MODEL	OpenAI model to use (optional, for OpenAI integration)	gpt-4o-mini
OLLAMA_URI	URI for the Ollama server (optional)	http://localhost:11434
OLLAMA_MODEL	Model to use with Ollama (optional)	llama3.1
CLOUDFLARE_ACCOUNT_ID	Cloudflare account ID (optional, for Cloudflare AI LLM and STT)	-
CLOUDFLARE_AUTH_KEY	Cloudflare authorization key (optional, for Cloudflare AI LLM and STT)	-
CLOUDFLARE_MODEL	Cloudflare model name (optional, for Cloudflare AI)	-
LLAMA_CPP_MODEL_PATH	Path to your GGUF model file (optional, for local llama.cpp inference)	-
STT_PROVIDER	Explicit STT provider selection (optional). Valid options: whisper.cpp, cloudflare. If not set, auto-detects based on configured credentials (whisper.cpp prioritized if available).	-
WHISPER_CPP_MODEL_PATH	Path to your Whisper GGML model file (optional, for local voice transcription with whisper.cpp)	-
CRON_SCHEDULE	Cron schedule for automatic summaries, in cron syntax (optional). Set to never to disable.	59 23 * * *
REDIS_URL	URL for the Redis server (optional)	redis://localhost:6379
MSG_LENGTH_LIMIT	Minimum message length to trigger automatic summarization	1000

Setup

1. Create Telegram Bot

1. Open Telegram and search for @BotFather.
2. Send /newbot command and follow the instructions.
3. Choose a name and username for your bot.
4. Save the bot token provided by BotFather.

2. Configure Bot Privacy

1. Send /setprivacy to @BotFather.
2. Select your bot from the list.
3. Choose "Disable" to allow the bot to read group messages.
4. This is required for the bot to access and summarize group messages.

3. Configure Environment

1. For OpenAI: Get your API key from OpenAI Platform.
2. For Ollama: No API key needed, just ensure Ollama is running locally.
3. For Cloudflare AI: Get your account ID and auth key from Cloudflare Dashboard.
4. For llama.cpp: Download a GGUF model file (e.g., from Hugging Face) and set the path to it. This enables fully local inference without external API calls.
5. For Voice Transcription: Choose one of the following STT (Speech-to-Text) options:
   • whisper.cpp (local): Download a Whisper GGML model from whisper.cpp models and set the path. Available models: ggml-tiny.bin, ggml-base.bin, ggml-small.bin, ggml-medium.bin, ggml-large-v3.bin. Larger models are more accurate but require more memory (large-v3 needs ~4GB RAM).
   • Cloudflare AI Whisper: Uses the @cf/openai/whisper-large-v3-turbo model. Just configure CLOUDFLARE_ACCOUNT_ID and CLOUDFLARE_AUTH_KEY (same credentials used for LLM). No local model download required.
6. Create a .env file in the project root with your configuration:

   TELEGRAM_BOT_TOKEN=your_bot_token_here
   
   # LLM Configuration (choose one provider)
   # LLM_PROVIDER=llama.cpp  # Optional: force specific provider (openai, cloudflare, ollama, llama.cpp)
   LLAMA_CPP_MODEL_PATH=/path/to/models/your-model.gguf
   
   # STT Configuration for voice transcription (choose one provider)
   # STT_PROVIDER=whisper.cpp  # Optional: force specific provider (whisper.cpp, cloudflare)
   WHISPER_CPP_MODEL_PATH=/path/to/models/ggml-base.bin
   
   # Or use Cloudflare AI for both LLM and STT
   # CLOUDFLARE_ACCOUNT_ID=your_account_id_here
   # CLOUDFLARE_AUTH_KEY=your_auth_key_here
   # LLM_PROVIDER=cloudflare
   # STT_PROVIDER=cloudflare

Deployment

Start the Bot

docker compose -f docker/docker-compose.yml up -d

Stop the Bot

docker compose -f docker/docker-compose.yml down

Usage

1. Add the Bot to a Group

1. Add your bot to any Telegram group where you want to use it.
2. The bot will automatically start monitoring messages in the group.
3. If you want to restrict the bot to specific groups, add the group IDs to the WHITELISTED_CHATS environment variable, separated by commas, and restart the bot.

2. Use Summary Command

• Send /summary in the group chat.
• The bot will analyze all available messages in the chat and provide a concise summary. The message history is automatically cleared after 8 hours of inactivity.

Credits

SummaryGram is made with ♥ by derogab and it's released under the MIT license.

Contributors

Tip

If you like this project or directly benefit from it, please consider buying me a coffee:
🔗 bc1qd0qatgz8h62uvnr74utwncc6j5ckfz2v2g4lef
⚡️ derogab@sats.mobi
💶 Sponsor on GitHub

Stargazers over time