ClashChat active debate on Economy topic showing AI and user exchange with Debate Stats panel

ClashChat

March 2025Team of 2 — led design and development

A full-featured AI-powered debate training app built with Flutter (Dart) and Cloud Firestore, letting users debate an LLM opponent — powered by GROQ — live, across Casual, Ranked, and Guided Learning modes. Users pick from 8 topic categories or a custom input and set their stance through animated segmented controls.

In Learning Mode, the system prompt returns structured JSON carrying both the opponent's counter-arguments and real-time coaching tips. In Ranked Mode, difficulty is matched to the user's competitive tier over a fixed 10-minute timer. At the end of a match, an automated scoring judge evaluates the full transcript across logic, rhetoric, and difficulty to award a 10–100 logic score, qualitative feedback, and Elo-style rank points.

The client itself carries a 5-tier rank ladder (Newcomer through Grandmaster) rendered as custom Canvas-painted badges, local 24-hour session validation, full dark/light theme support, animated score arcs, and a filterable debate history with complete conversation replay.

Built with

  • Flutter
  • Dart
  • Firebase Auth
  • Firestore
  • Firebase Hosting
  • Groq API
  • Cloudflare Workers
  • Hive
  • Provider

Screenshots

All work