Telegram Chatbot

OpenAI-compatible and Gemini-native LLM queries for groups, channels and private chats on Telegram.

Setup

• Clone the repo into a working folder git clone https://github.com/dhjw/telegram-chatbot && cd telegram-chatbot
• Set up and enter a virtualenv (optional). python -m venv venv then source ./venv/bin/activate (Linux) or .\venv\Scripts\activate (Windows)
• Install requisites to the current environment pip install -r requirements.txt
• Get a bot token from @Botfather
• Get API keys from Google, OpenAI and xAI (you will need credits except for Google; note OpenAI's expire in a year and mini models are cheap so just get $5 to start)
• Copy config.example.json to config.json and configure it
• Search for your bot and add it to your groups/channels (click profile name > Add to group), or open a private chat
• When happy with the active groups, use BotFather's /setjoingroups command to prevent the bot being added to more
• Use the hidden /id command to find out the chat_id for each chat. Add them to allow_chat_ids in config.json in an array, e.g. [-12312312, 123123],
• When you restart the bot all other chats will be ignored
• Be careful not to break the strict JSON config file

Run

There's a few ways to do it.

• Activate the venv, as in Setup, and run python ./bot.py, or make it executable chmod +x ./bot.py and run it directly ./bot.py
• Run it from the venv without activating /path/to/venv/bin/python /path/to/bot.py
• If you're not using a venv, run it with the system python3 or make it executable and run it

Update

Clone the repo and apply changes, keeping your config.json and venv.

cd /path/to/parent
git clone https://github.com/dhjw/telegram-chatbot tmp
rm -rf tmp/.git
cp -rf tmp/. telegram-chatbot/
rm -rf tmp

Notes

• gemini-2.5-flash is noticeably faster than gemini-2.5-pro and has 250 instead of 100 free requests per day

TODO, maybe

• Gemini grounding + live search (big free-tier)
• Grok 3 live search (expensive)
• Media input support
• Treat replies as new requests automatically
• Image generation