Your AI-Powered Fitness Companion
A Flutter-based mobile app backed by Firebase and integrated with Gemini LLM agents to generate personalized workout plans, track user progress, and provide exercise recommendations.
- Chat with Gemini Agent: Use Gemini’s GPT‑style models via Firebase AI to generate custom workouts, respond to health queries, and deliver exercise guidance.
- Workout Plan Generator: Create personalized exercise routines based on user goals, preferences, and history using LLM-powered function calling.
- Session & Streak Tracking: Monitor completed workouts, maintain progress records, and track streaks/rewards.
- Video Demonstrations: Integrate Cloudflare-hosted stock videos to show proper form for exercises.
| Layer | Technology |
|---|---|
| Frontend | Flutter (Dart) |
| Backend & AI | Firebase (Firestore & Authentication, Firebase AI + Gemini |
| Chat Agent | Gemini LLM with Function Calling |
| Media Storage | Cloudflare for workout video hosting |
Flutter App (UI & State)
↔ Firebase Auth + Firestore (User data + chat history)
↔ Firebase AI Agent (wraps Gemini models)
└─ Function calls generate workout plans based on inputs + context
Media assets (videos) served via Cloudflare
- Chat UI displays LLM-driven conversations
- Function calling patterns allow structured requests/responses
- Minimal exercise database (~45+ items), stored locally and in Firestore
-
Clone the repository
git clone https://github.com/soamn/exerai.git cd exerai -
Install dependencies
flutter pub get
-
Set up Firebase
- Add
google-services.json(Android) orGoogleService-Info.plist(iOS) - Enable Firebase Authentication, Firestore, and Firebase GenAI
- Add
-
Run the app
flutter run
- Flutter learning curve: UI setup and state management require experience
- Limited exercise database: Currently around 45 exercises included
- Prompt engineering: Output format inconsistency can occur with LLMs
- Built a conversational UI powered by Gemini LLM
- Implemented structured function calling for workout generation
- Created a tracking system for sessions, streaks, and history
- 🧍♀️ Pose detection for form correction (ML Kit / Posenet)
- 🎧 Voice-based assistant and workout narration
- 📊 Personal analytics dashboard with workout history
- 🥗 AI-powered diet/nutrition recommendations
- 🏋️ More exercises, tags, categories, and difficulty levels
- Flutter — Cross-platform UI framework
- Firebase — Backend services (Auth, Firestore, AI)
- Gemini — LLM for workout planning
- Cloudflare — Static media hosting for exercise videos
- Aman Negi — Developer, Gemini Agent Integration, Firebase Setup, UI/UX