Simple OpenAI Telegram Bot

• When it gets a text message it passes it to ChatGPT and returns the answer.
• When it gets a voice note it passes it through Whisper and returns it as text.
• You can also send a URL to an audio file and it will be transcribed.
• If you reply to a message (no matter your reply) it returns a summary of the message you replied to.

Installation

• Configure your own .env file from the provided example .env_example:
  • CHATGPT_KEY You get it in your OpenAI account settings
  • TELEGRAM_KEY Create it using BotFather
  • MODELS Comma separated list of the Whisper models that will be avaible. This allows lo exclude larger models that require much more resources.
  • DEFAULT_MODEL The Whisper model in use when the app starts.
• Run on Docker compose.

Ussage

This is a Telegram Bot, you interact with it through Telegram chat.

• /start Shows welcome message.
• /ayuda Shows list of possible commands.
• /modelo Shows the Whisper model in use.
• /tiny, /base, /small, /medium and /large change the Whisper model.
• /clear Starts a new conversation with ChatGPT.
• Any other text message is sent to chatGPT and its answer is returned.
• Send any voice note and a transcription is sent back.
• Reply (with whatever text) to a message and a summary is returned.
• Send a URL to an audio file and you will get a transcription. Depending on the model in use, the lenght of the audio and the hardware capabilities, it can be better to set a scheduled message for late at night, because the bot could get busy for hours long.