CipherDev - Privacy-First AI Chat

Run powerful AI models entirely in your browser. No servers, no API keys, no data transmission.

CipherDev is a fully client-side AI chat application that runs Large Language Models using WebGPU or WASM. Every conversation stays on your device. IBM Bob certified for zero data transmission.

✨ Features

🔒 100% Private

• All LLM inference runs locally in your browser
• No backend servers for chat processing
• No API keys required
• No telemetry or analytics
• IBM Bob Privacy Audit certified

⚡ Hardware Accelerated

• WebGPU support for blazing-fast inference
• Automatic WASM fallback for compatibility
• Smart device tier detection
• Optimized model recommendations

🤖 Multiple Models

• TinyLlama 1.1B (650MB) - Fast, lightweight
• Gemma 2 2B IT (1.4GB) - Balanced performance
• Llama 3.2 3B (1.9GB) - High quality
• Phi-3.5-mini (2.2GB) - Advanced reasoning

🎙️ Live Meeting Transcription (NEW!)

• Real-time speech-to-text with Whisper AI
• Voice Activity Detection (VAD) for optimal quality
• AI-powered meeting summaries
• Export to Markdown or JSON
• 100% local processing - no audio leaves your device

🏥 Health Risk Assessment

• Real-time health/disaster risk evaluation
• Temperature, heart rate, and condition analysis
• Watsonx AI integration (optional)
• Privacy-preserving local processing

🚀 Quick Start

Prerequisites

• Node.js 18+ and npm
• Chrome 113+ or Edge 113+ (for WebGPU)
• 4GB+ RAM recommended

Installation

# Clone the repository
git clone https://github.com/yourusername/chipherdev.git
cd chipherdev

# Install dependencies
npm install

# Start development server
npm run dev

Open https://devcipher.vercel.app/ in your browser.

Build for Production

npm run build
npm start

**Note about Build Warnings**: During production build, you may see warnings about error pages (404/500). These are non-fatal warnings from Next.js static generation and do not affect the application functionality. All main pages build successfully.

📁 Project Structure

chipherdev/
├── app/                      # Next.js 14 App Router
│   ├── (app)/               # App routes with layout
│   │   ├── chat/           # Chat interface
│   │   ├── models/         # Model selection
│   │   ├── audit/          # Privacy audit
│   │   └── settings/       # Settings page
│   ├── api/                # API routes
│   │   └── check-risk/     # Health risk assessment
│   ├── layout.tsx          # Root layout
│   ├── page.tsx            # Landing page
│   └── globals.css         # Global styles
├── components/              # React components
│   ├── ui/                 # UI primitives
│   ├── layout/             # Layout components
│   └── check-risk/         # Risk assessment UI
├── features/                # Feature modules
│   ├── hardware/           # Device detection
│   ├── llm/                # LLM engines
│   ├── audit/              # Privacy audit
│   └── conversation/       # Export utilities
├── store/                   # Zustand state management
│   ├── slices/             # State slices
│   └── useAppStore.ts      # Combined store
├── lib/                     # Utilities
├── bob_sessions/            # Privacy proof screenshots
└── public/                  # Static assets

🛠️ Technology Stack

Layer	Technology
Framework	Next.js 14 (App Router)
Language	TypeScript 5 (strict mode)
UI	React 18, Tailwind CSS 3.4
LLM Engine	@mlc-ai/web-llm (WebGPU)
Fallback	@xenova/transformers (WASM)
State	Zustand 5
Icons	Lucide React

🎯 How It Works

1. Hardware Detection

CipherDev automatically detects your device capabilities:

• WebGPU availability
• GPU name and memory
• RAM and CPU cores
• Device tier classification (High/Mid/Low/Minimal)

2. Model Loading

Models are downloaded from HuggingFace and cached locally:

• Quantized models (4-bit) for efficiency
• Progressive loading with status updates
• IndexedDB caching for instant reloads

3. Local Inference

All chat processing happens in your browser:

• WebGPU acceleration when available
• WASM fallback for universal support
• Streaming responses for real-time feedback
• No data ever sent to external servers

4. Privacy Audit

IBM Bob verifies zero data transmission:

• Network request analysis
• Storage inspection
• System verification
• Visual proof via screenshots

🔐 Privacy Guarantees

What CipherDev DOES:

✅ Run AI models locally in your browser
✅ Download model weights from HuggingFace
✅ Store models in browser cache (IndexedDB)
✅ Export conversations as local files

What CipherDev DOES NOT:

❌ Send your messages to any server
❌ Collect analytics or telemetry
❌ Require API keys or accounts
❌ Track your usage
❌ Share data with third parties

📊 Performance

Device Tier	GPU	RAM	Recommended Model	Speed
High	WebGPU, 2GB+ VRAM	8GB+	Llama 3.2 3B	~30 tokens/s
Mid	WebGPU	4GB+	Gemma 2 2B	~20 tokens/s
Low	WebGPU	4GB+	TinyLlama 1.1B	~15 tokens/s
Minimal	WASM only	2GB+	TinyLlama 1.1B	~5 tokens/s

🏥 Health Risk Assessment API

CipherDev includes a health/disaster risk assessment feature:

POST /api/check-risk
{
  "age": 35,
  "location": "Mumbai",
  "healthCondition": "diabetic",
  "temperature": 38.5,
  "humidity": 80,
  "heartRate": 95
}

Response:
{
  "risk": "medium",
  "action": "medical_kit",
  "reason": "Elevated temperature with pre-existing condition"
}

Optional Watsonx Integration

Set environment variables to use IBM Watsonx AI:

WATSONX_API_KEY=your_api_key
WATSONX_PROJECT_ID=your_project_id
WATSONX_URL=https://us-south.ml.cloud.ibm.com

#### How to Get Watsonx API Credentials

**Step-by-step guide:**

1. **Sign up for IBM Cloud** (if you don't have an account):
   - Visit [https://cloud.ibm.com/registration](https://cloud.ibm.com/registration)
   - Create a free IBM Cloud account

2. **Create a Watsonx.ai instance**:
   - Go to [IBM Cloud Catalog](https://cloud.ibm.com/catalog)
   - Search for "watsonx.ai"
   - Click on "watsonx.ai" service
   - Select your region (e.g., Dallas, Frankfurt, Tokyo)
   - Choose a pricing plan (Lite plan available for free)
   - Click "Create"

3. **Get your API Key**:
   - Go to [IBM Cloud API Keys](https://cloud.ibm.com/iam/apikeys)
   - Click "Create an IBM Cloud API key"
   - Give it a name (e.g., "CipherDev Watsonx Key")
   - Click "Create"
   - **Important**: Copy and save the API key immediately (you won't be able to see it again)

4. **Get your Project ID**:
   - Go to [Watsonx Projects](https://dataplatform.cloud.ibm.com/projects)
   - Create a new project or select an existing one
   - Click on the "Manage" tab
   - Copy the "Project ID" from the project details

5. **Find your Watsonx URL**:
   - Based on your region:
     - **US South (Dallas)**: `https://us-south.ml.cloud.ibm.com`
     - **EU (Frankfurt)**: `https://eu-de.ml.cloud.ibm.com`
     - **JP (Tokyo)**: `https://jp-tok.ml.cloud.ibm.com`

#### Configure Your Environment

Create a `.env.local` file in your project root:

```bash
# Copy from example
cp .env.local.example .env.local

Edit .env.local and add your credentials:

WATSONX_API_KEY=your_actual_api_key_here
WATSONX_PROJECT_ID=your_actual_project_id_here
WATSONX_URL=https://us-south.ml.cloud.ibm.com

Then restart the development server:

npm run dev

Note: Without these credentials, the app will use a simple rule-based risk assessment system (no external API calls).

## 🧪 Development

### Run Tests
```bash
npm test

Lint Code

npm run lint

Type Check

npm run type-check

📸 IBM Bob Privacy Audit

CipherDev is certified by IBM Bob for zero data transmission. See bob_sessions/ for proof screenshots:

1. Landing Page - Feature showcase
2. Hardware Detection - Device capabilities
3. Model Loading - Download progress
4. Chat Session - Live conversation
5. Audit Page - Privacy certification
6. Network DevTools - Zero external requests

🤝 Contributing

Contributions are welcome! Please read our contributing guidelines before submitting PRs.

🙏 Acknowledgments

• @mlc-ai/web-llm - WebGPU inference engine
• @xenova/transformers - WASM fallback
• HuggingFace - Model hosting
• IBM Bob - Privacy audit certification

📞 Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions

---

Made with ❤️ and chaos also with IBM Bob 🤖🔒

CipherDev - AI that respects your privacy

MIT licence