Telegram AI CalTracker Bot

AI-powered Telegram calorie and nutrition tracker running on Cloudflare Workers. The bot extracts meal details with Gemini, stores logs in PostgreSQL, and gives daily summaries.

What It Does

• Onboards users with name, timezone, and calorie goal
• Logs meals from free-text input using AI extraction
• Supports guided logging with /log and review buttons
• Saves detailed nutrient fields for each food item
• Shows grouped daily totals with /today

Bot Access

Use either option below:

1. Open bot link: Open Telegram Bot (or)
2. Scan QR code:

Place your QR image at docs/images/telegram-bot-qr.png so it renders on GitHub.

Commands

• /start - Start onboarding
• /log - Start guided logging (bot asks for meal text next)
• /log <meal text> - Quick log in one message
• /today - Show today's saved foods and totals

Tech Stack

• Cloudflare Workers + Wrangler
• TypeScript
• PostgreSQL (postgres runtime client)
• Prisma schema and migrations
• Gemini API

Project Structure

src/
	data/db.ts
	handlers/telegram-handler.ts
	services/ai.ts
	services/telegram.ts
	types/index.ts
	worker.ts
docs/
	backend-architecture-v2.md

Prerequisites

• Node.js 20+
• npm
• Cloudflare account
• PostgreSQL database
• Telegram bot token

Setup

npm install

Set Worker secrets:

wrangler secret put DATABASE_URL
wrangler secret put TELEGRAM_BOT_TOKEN
wrangler secret put TELEGRAM_WEBHOOK_SECRET

GEMINI_MODEL is configured via wrangler.toml. Gemini API keys are loaded from DB table gemini_keys.

Local Development

npm run cf:login
npm run cf:whoami
npm run cf:dev

Deploy

npm run cf:deploy

Endpoints

• GET /health - Health check
• POST /telegram-webhook - Telegram updates

Database and Prisma

npm run prisma:validate
npm run prisma:generate
npm run prisma:push

Type Check

npm run typecheck