https://github.com/shablin/telegram-chat-automation
Implementation of Chat Automation bot mode that allows you to delegate chats management with a bot
https://github.com/shablin/telegram-chat-automation
chat-automation llm ollama secretary-bot telegram telegram-bot-ai-assistant telegrambot telegrambotapi
Last synced: 17 days ago
JSON representation
Implementation of Chat Automation bot mode that allows you to delegate chats management with a bot
- Host: GitHub
- URL: https://github.com/shablin/telegram-chat-automation
- Owner: shablin
- License: mit
- Created: 2026-05-31T10:52:49.000Z (29 days ago)
- Default Branch: main
- Last Pushed: 2026-05-31T13:59:38.000Z (29 days ago)
- Last Synced: 2026-05-31T15:23:43.592Z (29 days ago)
- Topics: chat-automation, llm, ollama, secretary-bot, telegram, telegram-bot-ai-assistant, telegrambot, telegrambotapi
- Language: Python
- Homepage:
- Size: 24.4 KB
- Stars: 1
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# telegram-chat-automation
A local AI-powered Telegram Business Bot that uses Ollama and a Large Language Model (LLM)
to automatically respond to messages through Telegram Business Connections.
The project is designed to be simple, self-hosted, extensible, and privacy-friendly.
All inference runs locally on your own machine without relying on external AI APIs.
## Features
- 🤖 Telegram [Secretary Bot](https://core.telegram.org/bots/features#secretary-bots) support (Chat Automation)
- 🧠 Local LLM inference via Ollama
- 📜 Custom system prompts using Markdown
- 🐳 Docker-based deployment
- 💬 Automatic replies in personal chats
- ⌨️ Typing indicator while generating responses
- 🔒 Fully self-hosted
- 🚀 Extensible architecture for future skills and tools
## How it works
## Requirements
- [Docker](https://docker.com) >=28.1.1
- [Docker Compose](https://docs.docker.com/compose/)
- [Python](https://python.org) >=3.11.2
- [aiogram](https://aiogram.dev/)
- [Ollama](https://ollama.com/)
## Installation
### 0. Create Telegram bot
Go to [BotFather bot](https://t.me/BotFather) and follow provided instructions
### 1. Clone the repository
```bash
git clone https://github.com/shablin/telegram-chat-automation.git
cd telegram-chat-automation
```
### 2. Configure environment variables
Create a `.env` file from `.env.example`:
```env
BOT_TOKEN=your_bot_token
OLLAMA_URL=http://ollama:11434
OLLAMA_MODEL=ollama_model_name
OLLAMA_TIMEOUT=model_response_timeout
```
### 3. Start Ollama
```bash
docker compose up -d ollama
```
### 4. Download a model
Example:
```bash
docker exec -it ollama_sevice ollama pull qwen2.5:7b
```
Verify:
```bash
docker exec -it ollama_sevice ollama list
```
### 5. Start the bot
```bash
docker compose up -d --build
```
## Customizing the bot
The bot behavior is defined in:
```text
bot/prompts/system.md
```
> [!IMPORTANT]
> The contents of this file are injected into every conversation as the system prompt.
## Supported Models
Any Ollama-compatible model can be used.
Recommended:
- qwen2.5:3b
- qwen2.5:7b
- qwen3:4b
- mistral:7b
- deepseek-r1
You can change the model in `.env`:
```env
OLLAMA_MODEL=qwen2.5:7b
```
## Development
View logs:
```bash
docker compose logs -f bot
```
Check Ollama:
```bash
docker compose logs -f ollama
```
Open a shell inside the bot container:
```bash
docker exec -it telegram_bot sh
```
Open a shell inside the Ollama container:
```bash
docker exec -it ollama_sevice sh
```
## Roadmap
- [ ] Conversation memory
- [ ] Skill system
- [ ] Tool calling
- [ ] Streaming responses
- [ ] Multi-model routing
- [ ] Vector search (RAG)
- [ ] User profiles
- [ ] Conversation analytics
- [ ] Plugin system
## Security
This project executes all AI inference locally.
No messages are sent to third-party AI providers unless you explicitly modify the source code.
Always review prompts, tools, and future integrations before deploying the agent in production environments.