🦙🗣️ Ollama Telegram Bot

Chat **and speak** to your LLM, using Telegram bot!
Feel free to contribute!

Shamelessly forked from Ollama-telegram.

I only added the voice part and couple of things such as preregistering users once they appear in the .env file at bot start.

Prerequisites

• Telegram-Bot Token

Installation (Non-Docker)

• Clone Repository

git clone https://github.com/xawos/owt

• Install requirements from requirements.txt

pip install -r requirements.txt

• Enter all values in .env.example as shown in the original repo for starters.

NB: My fork has 2 additional flags, WSPRMODEL and USE_CUDA, respectively set to base and True.

Both those options are meant for Whisper, list of models here.

• Rename .env.example -> .env after setting the above options

• Launch bot and wait until you see a line starting with INFO:aiogram.dispatcher:Run polling for bot

python3 run.py

If everything went well good job! You can now send voice messages and images to your Telegram bot!

It also works on Raspberry Pi, with Phi3 (without vision) it replies to a voice prompt in ~30s!

Credits

• Ollama-telegram (original bot by ruecat)
• Ollama
• Whisper
• FFmpeg

Libraries used

• Aiogram 3.x
• OpenAI-Whisper
• ffmpeg-python
• and more