Logger Bot

A Telegram bot that accepts voice messages, transcribes them, and parses workout data (exercise, reps, weight) for logging to a csv or database.

Architecture

flowchart LR
    user([User])
    telegram(Telegram Bot)
    server([Server])
    extractor([Speech2Text + parsing])
    logger([Logger])
    db([Database])

    user -- voice --> telegram
    telegram -- webhook --> server
    server -- audio --> extractor
    extractor -- text --> server
    server -- text --> logger
    logger --entry--> db

Loading

Requirements

• uv package manager
• A Telegram bot token (from @BotFather)

Setup your bot server using uv

1. Configure the bot token Create a bot using Telegram's @BotFather. Grab your token and create a .env file in the project root with it like this:

TELEGRAM_API_TOKEN=your_telegram_bot_token_here

2. Launch Server Make sure you have uv installed in your machine.

Run this on your terminal (at project root level):

./run_server.sh

That will set up the venv, install dependencies and launch the server that communicates with your telegram bot.

You can healthcheck the server typing /health on your telegram bot chat and see this:

Using the bot

The bot will listen for voice, messages and commands (start with /).

You can send a voice message describing your workout. The bot transcribes it, parses the result, and logs it into the database/csv file. The bot understands natural language, for example, you can say:

• "Press de banca, 10 repeticiones, 80 kilos, categoría pecho"
• "Sentadilla 5 reps 100 kg"
• "Shoulder press 8 repetitions 60 lbs"

Bot Commands

These are the commands

• /last <exercise>: Returns the most recent logged set for a given exercise, as a JSON code block, for example /last sentadilla.

• /sql <query>: Runs a SQL query against log.db and returns the results as a JSON code block. The connection is read-only, so only SELECT statements work, for example:

Launch the database view You can visualize your database and run queries against it using a UI. Run this command to launch the UI:

./launch_db_view.sh

This script:

• Imports your latest log.csv into a SQLite database called log.db (table: logs).
• Starts a local Datasette server.

Open your browser on http://127.0.0.1:8001 — you'll see your workouts as an interactive table with an SQL interface to interact with it.

GPU Support on the server

The server detects automatically available GPU on the machine and uses it to compute run the neural network computation (Speech2Text), which makes it much faster.

Docker

You can run the server inside docker using Dockerfile running:

docker build -f Dockerfile . -t bot
docker run -it bot