MFP Auto Bot

Telegram bot that auto-logs meals on MyFitnessPal by learning your eating patterns. Predicts recurring meals, tracks macros in real time, and suggests foods that fit your remaining budget.

Built for gym diet tracking — optimizes for hitting protein/carbs/fat targets.

How It Works

1. Connect your MFP account via /start
2. Run /setup to register your typical foods (with exact portions)
3. Use /today — the bot predicts your meals, you confirm with one tap
4. After each confirmation, see your macro progress vs goals
5. The bot learns: the more you use it, the better the predictions

Bot Commands

Command	What it does
/start	Step-by-step guide to connect your MFP account
/token <json>	Authenticate with MFP (token is deleted from chat immediately)
/setup	Register your typical foods per meal slot with portions
/today	Log today's meals — slot by slot with predictions
/tomorrow	Same, for tomorrow
/day Mon	Log meals for a specific day
/week	7-day wizard with stop/resume
/macros	Check remaining macros for today
/suggest	Foods from your history that fit remaining macros
/copy yesterday	Copy all meals from another day to today
/history	7-day macro adherence overview
/status	Slots filled/pending per day
/undo	Remove the latest food entry from MFP for today
/retry	Retry failed MFP syncs

Getting Started (User)

After starting the bot on Telegram:

1. Send /start — the bot explains how to get your MFP token
2. Log into myfitnesspal.com in your browser
3. Visit https://www.myfitnesspal.com/user/auth_token?refresh=true
4. Copy ALL the JSON text on that page
5. Send /token <paste here> to the bot
6. Run /setup to configure your typical foods with exact portions
7. Done! Use /today to start logging

The token expires after some time. When it does, the bot tells you to refresh it with /token.

Dev Setup

1. Create a Telegram bot via @BotFather
2. Generate a Fernet key:

   python -c "from cryptography.fernet import Fernet; print(Fernet.generate_key().decode())"

3. Copy .env.example to .env and fill in the values
4. Create venv and install:

   python -m venv .venv
   source .venv/Scripts/activate  # Windows Git Bash
   pip install -r requirements.txt

5. Run:

   python main.py

Testing

# Unit tests (always work, no credentials needed)
pytest --ignore=tests/test_integration_telegram.py --ignore=tests/test_integration_mfp.py

# Integration tests (need real tokens in .env)
pytest tests/test_integration_mfp.py -v -s
pytest tests/test_integration_telegram.py -v -s

# All tests
pytest -v

Integration tests require TELEGRAM_TEST_CHAT_ID and MFP_AUTH_JSON in .env (see .env.example).

Deployment

Deployed on Render with auto-deploy from GitHub.

1. Fork this repo
2. Create a Web Service on render.com
3. Set environment variables: TELEGRAM_BOT_TOKEN, ENCRYPTION_KEY
4. Deploy — render.yaml handles the rest

Architecture

main.py              Entry point, health check, handler registration
config.py            Env vars, meal slots, thresholds

bot/
  onboarding.py      /start, /token — step-by-step connection wizard
  setup.py           /setup — register foods with serving sizes per slot
  daily.py           /today, /tomorrow, /day — slot-by-slot with macro tracking
  week.py            /week — 7-day wizard with timeout and stop/resume
  suggest.py         /suggest — macro-aware food suggestions from history
  utility.py         /status, /undo, /retry, /macros, /copy, /history
  reminder.py        Daily 21:00 reminder for unfilled slots
  keyboards.py       Inline button builders (confirm, alternatives, serving sizes)
  messages.py        Message formatting (slots, macros, history)

engine/
  pattern_analyzer.py  Meal pattern extraction with temporal decay
  predictor.py         Daily predictions per slot with confidence levels
  learner.py           Weight updates on confirm/replace feedback

mfp/
  client.py           MFP API client (auth, diary, search, add entry, goals)
  scraper.py           History import (rate-limited)
  sync.py              Retry queue for failed writes

db/
  database.py          Async SQLite CRUD + Fernet encryption
  models.py            User, MealEntry, MealPattern, WeekProgress

Tech Stack

• Python 3.14, async throughout
• python-telegram-bot 22.7
• curl_cffi (Chrome TLS impersonation for MFP API)
• aiosqlite + SQLite
• pandas (pattern analysis)
• cryptography (Fernet encryption for stored tokens)