I recently launched multiwel.mihaimangu.ro, a multi-user wellness tracker built for couples, families, or any small group who wants to keep an eye on what they eat, how they train, and how they sleep—without opening yet another app every time something needs to be logged.

The Problem With Wellness Apps

Most fitness and nutrition trackers are built for a single person. You log your own meals, your own workouts, your own sleep, in your own private silo. But wellness is often a shared effort—a couple trying to eat better together, a family keeping each other accountable, parents wanting visibility into what their kids ate that day.

I wanted something different: a shared space where everyone in a household can see what everyone else is doing, without any of them having to fight with a clunky logging UI.

The Core Idea: Log Through Conversation

Instead of building yet another form-heavy interface, Multiwel is designed around the way people already communicate: a Telegram group chat.

Everyone in the household joins the same Telegram group. When someone eats a meal, finishes a workout, or wakes up, they just... type it. No menus, no dropdowns, no searching a food database:

• "had two eggs and toast for breakfast"
• "just finished a 5k run, felt good"
• "went to bed at 11, slept great"

That's it. The message goes into the shared group, and everyone can see it in the conversation immediately—but it also gets parsed, structured, and stored automatically so it shows up on the web dashboard too.

Enter Hermes

The bridge between Telegram and the web app is a bot I built called Hermes. It sits in the Telegram group, reads every relevant message, and uses an AI model to interpret free-form text into structured data:

• Food → estimated calories and macronutrient breakdown
• Exercise → activity type, duration, distance, and intensity (running, strength training, padel, and more)
• Sleep → bedtime, wake time, and quality, grouped by the day you woke up

Hermes doesn't require any special syntax. People just talk naturally, and the bot figures out the intent. If it misreads something, users can correct it directly from the dashboard.

The Stack

Multiwel is built with a fairly simple, boring-in-a-good-way stack:

• Next.js — the web app and dashboard
• MongoDB — storing users, households, and all logged activity
• Hermes (Telegram bot) — the conversational interface and AI parsing layer connecting Telegram to the app

The web dashboard shows daily totals, per-activity summaries, and trends over time, while Telegram stays the primary input surface. Access is invite-only, since the whole point is that the people in a group can see each other's data—so a household signs up, invites the people they want to share with, and starts logging through the chat they already use every day.

Why This Approach Works

The biggest lesson from building this: the best interface is often the one people don't have to learn. Everyone already knows how to send a Telegram message. By meeting users where they already are instead of asking them to adopt a new habit, logging stops feeling like a chore and starts feeling like just... talking about your day.

#NextJS #MongoDB #Telegram #AI #WellnessTech