MindAura brings supportive conversation, mood awareness, mindful breathing, movement, therapeutic sound, personal-growth reading, and cognitive play into one focused experience. Aurora responds to common emotional needs with distinct, practical guidance while keeping clear boundaries around crisis support, diagnosis, and abusive language.
| 7 wellness spaces | 30 mind games | Voice-enabled | 3D companion | Private analytics |
|---|---|---|---|---|
| One calm home | Play and practise | Speak and listen | Emotion-aware motion | Session-isolated data |
Important
MindAura is an educational wellness-support project. It is not a replacement for professional medical advice, diagnosis, therapy, or emergency care.
| Space | What you can do |
|---|---|
| Aurora Companion | Have natural, intent-aware conversations with voice input and spoken replies |
| Mood Dashboard | Log daily mood, follow streaks, and review private session activity |
| Meditation Studio | Use guided breathing, calm timers, and mindful sessions |
| Move & Yoga | Follow structured exercises and camera-assisted posture experiences |
| Sound Therapy | Explore relaxing playlists, playback controls, and sleep timers |
| Growth Library | Read curated personal-growth material and save favourites |
| Mind Games | Play 30 exercises across memory, focus, language, speed, math, and logic |
flowchart LR
CHECK[Check in] --> TALK[Talk with Aurora]
TALK --> RESET[Breathe or move]
RESET --> GROW[Read or train your mind]
GROW --> REFLECT[Review your progress]
REFLECT --> CHECK
The game library includes Memory Match, 2048, Wordle, Sudoku, Simon Says, Typing Speed, Trivia, Pattern Recall, Mental Math, Word Hunt, Grid Navigator, Logic Gates, Reflex Trainer, 2-Back Focus, Mindful Recall, Odd One Out, and more.
- Crisis language receives immediate, region-relevant support guidance.
- English and Hindi/Hinglish abusive language—including common evasive spellings—is rejected.
- Rejected abusive content is not stored in conversation history.
- Wellness analytics are isolated by session rather than shared across users.
- Aurora does not diagnose conditions or present itself as a therapist.
flowchart LR
A[React + Vite] --> B[FastAPI]
B --> C[(SQLite)]
A --> D[Three.js Avatar]
A --> E[MediaPipe Vision]
B --> F[Safety + Intent Engine]
F --> B
| Layer | Technology |
|---|---|
| Interface | React 18, Vite, CSS, Lucide React |
| Interactive media | Three.js, MediaPipe Tasks Vision, Web Speech APIs |
| API | Python, FastAPI, Pydantic, Uvicorn |
| Storage | SQLite |
| Production | Vercel frontend with configurable backend URL |
Aurora uses a safety-first intent engine to recognise what kind of support a message needs before selecting a response. It provides dedicated conversational paths for motivation, anxiety, overwhelm, grounding, sadness, anger, loneliness, sleep, gratitude, and open sharing. Multiple responses per intent keep the experience natural, while strict moderation and crisis routing run first.
Message → Safety check → Intent detection → Support strategy → Natural response
│
├── Crisis language → Immediate support guidance
└── Abuse detected → Reject without storing the message
cd mental-health-backend
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
uvicorn app:app --reloadThe API is available at http://127.0.0.1:8000; interactive documentation is
available at http://127.0.0.1:8000/docs.
cd FRONTEND
npm install
npm run devOpen the local address printed by Vite.
| Variable | Purpose |
|---|---|
VITE_API_URL |
Public URL of the deployed FastAPI service |
AURORA_SECRET_KEY |
Strong private signing secret for authentication tokens |
AURORA_DB_PATH |
Optional SQLite database location |
CORS_ORIGINS |
Comma-separated trusted frontend origins |
Never commit production secrets or local .env files.
Every release is checked for:
- Production frontend compilation and route chunking
- Backend import and API health
- Authentication and token validation
- Exact prompt responses for motivation, anxiety, calming, overwhelm, and sharing
- Abuse spelling variations and rejected-message storage safety
- Crisis-response priority
- Per-session analytics isolation
- Responsive keyboard- and touch-friendly controls
AI-COMPANION/
├── FRONTEND/
│ ├── public/
│ ├── src/
│ │ ├── components/
│ │ ├── pages/
│ │ └── services/
│ └── package.json
├── mental-health-backend/
│ ├── app.py
│ ├── main.py
│ └── requirements.txt
└── README.md
Thoughtful improvements are welcome. Please open an issue for substantial changes and keep accessibility, privacy, emotional safety, and mobile usability at the centre of every contribution.
- Fork the repository and create a focused branch.
- Keep changes accessible and responsive.
- Run the frontend production build and backend checks.
- Open a pull request explaining the user-facing improvement.
Built with care by Khushi Soni
Pause. Breathe. Begin again.