The first build of the app that became Wivme. Pick a topic, get a short podcast-style lesson generated for it, listen, and keep track of what you covered.
Built July 2025. Kept here as it was, under the original "Wisme" spelling. Not maintained.
- Generates the lesson. A topic goes in, an LLM writes a short spoken lesson for it.
- Reads it aloud. Two voice coaches with different personalities, through text to speech.
- Tracks progress. Streaks, achievements and a leaderboard over what you have listened to.
- Works offline. Lessons download so they play without a connection.
The Flutter project is in wisme_app/, with the source under wisme_app/lib/ split into UI, services, repositories, models, providers and a design_system. The development notes from the build are the markdown files alongside it.
Flutter and Dart, an LLM for lesson generation, TTS for the audio. Runs on Android, iOS and web.
cd wisme_app
cp .env.example .env # add your own API keys
flutter pub get
flutter runNo keys are committed. .env.example lists the ones you need.
Phase 1 was there to answer one question: is a generated audio lesson actually worth listening to. It was, but listening alone did not make anything stick, which is the gap the later work went after. That became Wivme: the same audio-first idea, plus the part this version had no answer for, a spaced-repetition engine that decides when a concept has faded and brings it back.
Archived prototype from July 2025. It is here as a record of where the product started, not as something to run in production.