Verdi — AI Mediator for Everyday Conflicts (Telegram Bot)

Inspiration

Everyday arguments happen all the time — between friends, classmates, siblings, or even strangers online. Most of them aren't serious enough to go to a counselor, yet they leave people feeling misunderstood, unheard, or just angry.
We asked ourselves: What if there was a neutral third party who could calmly listen to both sides and help them reflect?
That’s what Verdi was created for — a Telegram bot that acts as an AI mediator, helping people understand each other better through structured, fair, and insightful analysis.

What it does

Verdi is a Telegram bot designed for two users to resolve a conflict. It:

Asks each person to describe their point of view
Uses Gemini AI to structure and compare their arguments
Highlights agreements, disagreements, and logical fallacies
Delivers individual responses to both users — offering clarity, balance, and, when possible, a fair verdict

Sometimes Verdi determines who’s more reasonable, other times it explains how both are misunderstanding each other. It’s not about winning — it’s about understanding (most of cases).

How we built it

We used the following technologies:

Python with aiogram for the Telegram bot logic
Gemini API (Google AI Studio) to analyze and process text from both users
MongoDB for optional session storage
Self-hosted backend on a personal server for full control and deployment flexibility

User sessions are handled asynchronously. Arguments are collected, structured, and sent to Gemini with a carefully tuned prompt template to generate fair, context-aware feedback.

Challenges we ran into

Prompt tuning: Getting Gemini to analyze arguments neutrally and individually for each user was tricky. We had to engineer prompts that balanced structure with emotional sensitivity.
Input imbalance: One user often writes more or better than the other. Making the bot fairly evaluate both, regardless of length or tone, took iteration.
Bot UX in Telegram: Designing a user flow for two people in a single chat with a bot required a mix of creativity and Telegram-specific logic.

Accomplishments that we're proud of

We created a fully functional AI-powered mediator inside a familiar app — Telegram — with zero setup required for users.
Verdi doesn’t just answer — it listens, compares, and gives emotionally intelligent feedback.
Early testers said it helped avoid unnecessary fights and gave them new ways to express their thoughts.

What we learned

How to integrate AI models like Gemini into conversational interfaces
Prompt engineering for empathy, fairness, and structure
Handling multiple users' input in real-time within one bot
That AI can’t solve every argument — but it can absolutely help people talk better

What's next for Verdi — AI Mediator for Everyday Conflicts (Telegram Bot)

Web interface for wider access and analytics (early concept already has been made)
Integration with group chats for mediating minor disputes in real time

We believe Verdi has the potential to become a go-to companion for healthy communication, starting with everyday conflicts.