Silent Guardian is a privacy-first mental health support web app that helps users express their feelings safely using AI — without compromising personal data.
The core idea is simple: your emotions stay on your device.
Mental health support often comes with a trade-off:
- 🧠 Helpful AI
- 🔒 But poor privacy
Silent Guardian aims to break this trade-off by offering:
- Local / user-controlled AI usage
- No forced login
- No data stored on servers by default
-
🎙️ Voice or Text Input
Express emotions through short recordings or text. -
🧠 AI-Based Emotional Understanding
Speech-to-text conversion with emotion or stress pattern detection using keywords and tone. -
🔐 Privacy-First Design
No mandatory authentication, no cloud storage by default, and support for user-provided AI API keys. -
💬 Supportive AI Responses
Calm, reassuring replies with grounding or breathing suggestions. -
📊 Local History (Optional)
Conversations can be stored locally with full user control over deletion.
- Frontend: Next.js (App Router)
- Language: TypeScript
- Styling: Tailwind CSS
- AI (Optional): Google Gemini API, OpenAI API
- Speech: Browser Speech-to-Text (Web Speech API)
The application is currently configured to prioritize Gemma 3 models (Instruction Tuned) for high-quality, efficient responses.
Model Priority:
gemma-3-12b-it(High Intelligence)gemma-3-27b-it(Complex Reasoning)gemma-3-4b-it(Fast/Lightweight)
This configuration can be modified in app/api/ask-gemini/route.ts.
Clone the repository, install dependencies, start the development server, and open the app locally:
git clone https://github.com/AyanKumar766/Silent-Guardian.git
cd Silent-Guardian
npm install
npm run devSilent Guardian is not a medical or diagnostic tool. It provides emotional support only and is not a substitute for professional therapy.
If you are in crisis, please contact a licensed mental health professional or local emergency services.
This is an early-stage project.
Feedback, ideas, and contributions are welcome.