Simple OpenAI text model Telegram bot

A simple Telegram bot wrapping OpenAI API text models (like ChatGPT), build with python-telegram-bot!

Just start a conversation and all messages in the chat will be used as inputs for ChatGPT.

Conversation/context is not stored permanently and will be removed when the bot is restarted.

Requirements

This bot was built with Python 3.11, python-telegram-bot and openai-python. Full list of Python requirements is in the requirements.txt file, you can use it to install all of them.

Configuration

Configuration is done through a .env file. You can copy example file .env.example as .env and fill required parameters.

cp .env.example .env

Telegram bot

Only required parameter is a bot token.

You can also restrict who can access the bot via ALLOWED_USERNAMES. You can specify multiple usernames delimited by space. If you don't want to restrict the bot at all you can remove this parameter or leave it empty.

BOT_TOKEN='<your secret bot token>'
ALLOWED_USERNAMES='myusername friendusername'

OpenAI API

One required parameter is API key.

OPENAI_TOKEN='<your secret API key>'

Through .env you can also configure other parameters:

• OPENAI_MODEL - which model to use (gpt-3.5-turbo is used by default)
• OPENAI_SYSTEM_MESSAGE - system message
• OPENAI_CONTEXT_LIMIT - how many messages will be kept in the context aside from prompt, all messages will be kept if empty
• OPENAI_INITIAL_MESSAGE - additional message added after system message to all conversations, can be empty for no additional messages

Note that gpt-3.5-turbo doesn't pay strong attention to system message, so changing it might not provide significant changes to responses. You can use OPENAI_INITIAL_MESSAGE to tweak initial behaviour of the model.

OPENAI_MODEL='gpt-3.5-turbo'
OPENAI_SYSTEM_MESSAGE='You are a helpful assistant.'
OPENAI_CONTEXT_LIMIT=1000
OPENAI_INITIAL_MESSAGE='You are a helpful assistant acting like 18th century butler,'

Commands

• /start - prints initial message returned from the model for just system message and optional initial message, doesn't impact conversation context
• /reset - resets current conversation and removes all context, other than system message
• /promptset - set new custom prompt, changes both system message and custom initial message (overwrites both OPENAI_SYSTEM_MESSAGE and OPENAI_INITIAL_MESSAGE)
• /promptreset - restore prompt to default
• /promptget - get custom prompt, won't respond with default one to avoid leaking configuration to users
• /promptremove - force-remove any prompts from the conversation, including one defined in configuration files
• /cancel - cancels ongoing operation, applies only to /promptset

Running the bot

You can run the bot from the source code directly, or in a Docker container.

From source code

1. Create a Telegram bot via BotFather
2. Create OpenAI API key
3. Install all packages from requirements.txt
4. Fill .env file
5. Run main.py file with Python

Docker

1. Create a Telegram bot via BotFather
2. Create OpenAI API key
3. Fill .env file
4. Run docker compose up -d --build in terminal

Note that .env file is used only for loading environment variables into Docker container through compose. The file itself isn't added to the container.