Telegram Multi-Model AI Bot (Python)

A small, dependency-free Telegram bot that sends chat requests to an OpenAI-compatible API. It supports configurable models, bounded conversation history, long polling, retries, and Telegram-safe message splitting.

APIMART · multi-model AI API gateway

A runnable Telegram bot using one OpenAI-compatible API endpoint. Run a Telegram AI bot with one multi-model API key

Transparent disclosure: this repository is maintained by APIMART.

中文说明：README.zh-CN.md

Features

• Pure Python standard library; no framework dependency
• OpenAI-compatible /chat/completions
• /model shows the active model
• /reset clears the current chat history
• Per-chat history with a configurable message limit
• Retry with exponential backoff for transient API errors
• Splits replies to stay within Telegram's message limit

Run

Requirements: Python 3.10+

cp .env.example .env
# Set TELEGRAM_BOT_TOKEN and APIMART_API_KEY in your shell.
python bot.py

PowerShell example:

$env:TELEGRAM_BOT_TOKEN = "123456:telegram-token"
$env:APIMART_API_KEY = "your-api-key"
$env:AI_MODEL = "gpt-4o-mini"
python .\bot.py

The default API endpoint is https://api.apimart.ai/v1. Set AI_API_BASE_URL to use another OpenAI-compatible gateway.

Commands

• /start — usage help
• /model — show the configured model
• /reset — clear this chat's history
• Any other text — send it to the model

Test

python -m unittest discover -s tests -v

The tests do not call Telegram or an AI API.

Production checklist

• Store tokens in a secret manager; never commit them.
• Use a dedicated Telegram bot token.
• Add allow-lists or authentication if the bot is private.
• Add persistent storage if conversations must survive restarts.
• Review rate limits and model pricing before opening the bot publicly.

License

MIT