GramSathi AI 🌿 — Empowering Rural India

A highly scalable, multi-lingual, and completely Voice-First intelligent assistant designed exclusively for the agricultural, healthcare, governance, and educational sectors of rural India.

---

🚀 Key Features

• 🎙️ 100% Voice-First Interface: Zero-typing UI. Simply press the microphone, speak naturally (in Hindi, Marathi, or English), and the AI instantly responds with contextual voice audio.
• 🛡️ Deep Domain Isolation: The AI physically refuses to answer off-topic queries, guaranteeing that village users receive highly accurate, actionable information explicitly mapped to their requested sector (Agriculture, Health, Edu, Govt).
• 🌐 Dual-Brain Fallback API: Runs securely on Google Gemini 1.5 Flash. Should the primary cloud fail, the backend instantly intercepts the error and routes the exact context to Sarvam AI (sarvam-30b). Zero downtime.
• 📱 Progressive Web App (PWA): Easily installable to any local home screen directly from the browser window (Add to Home Screen), with completely off-grid offline local-database caching.
• 🔒 Zero-PII Secure Engine: The API explicitly deletes and aggressively scrubs sensitive PII (Aadhaar Data, Phone Numbers) using Regular Expressions before any payload hits cloud endpoints.
• 🎨 Premium Glassmorphism: Features a highly optimized 100dvh CSS layout rendering absolute Frosted Glass panels against an infinite natural-moving gradient, maximizing visual engagement globally without external CSS/JS bloatware.

---

🛠️ Technology Stack

Frontend (React + Vite)

• Core: React 18, Vite (HMR)
• Styling: Pure Modern CSS-in-JS Variables, Dynamic Glassmorphic Viewports
• Build/Offline: Vite-Plugin-PWA (Progressive Web Application standard)
• Audio Pipeline: window.SpeechRecognition (STT Input)

Backend (Python + FastAPI)

• Core: Python 3.10+, FastAPI (ASGI Asynchronous Routing), Uvicorn
• Database: sqlite3 natively handling session recall and caching mechanisms
• Inference APIs: Google Gemini SDK, sarvamai (Fallback pipeline)
• Audio Pipeline: gTTS (Response Output)

---

⚙️ Installation & Setup

1. Clone the Repository:

   git clone https://github.com/your-username/gramsathi-ai.git
   cd gramsathi-ai

2. Configure Environment Variables: Open the root .env file and strictly insert your API keys:

   GEMINI_API_KEY=your_gemini_key_here
   SARVAM_API_KEY=your_sarvam_key_here
   VITE_OLLAMA_URL=http://localhost:11434

3. Initialize the Python Backend: Open a terminal into the /backend folder.

   cd backend
   pip install fastapi uvicorn google-genai sarvamai Pydantic gTTS python-dotenv
   python main.py

   The server dynamically creates gramsathi.db memory instances and spins up on http://localhost:8000.

4. Initialize the React Frontend: Open a separate terminal in the root project folder.

   npm install --legacy-peer-deps
   npm run dev

   The blazing frontend will instantly boot on http://localhost:3002.

---

📱 PWA Deployment Instructions

If you want to install GramSathi natively to your mobile device:

1. Since browsers tightly lock Progressive Web Apps to secure contexts, use a secure tunnel mapping to your laptop (e.g., ngrok http 3002).
2. Open the secure https link on Chrome/Safari.
3. Open the Settings Panel (⚙️) in the application, and hit the blue 🚀 Install App button!

---

👥 Contributors & Documentation

• Presentations? See the included gramsathi_ppt_content.md for immediate deck copy-paste.
• Code architecture breakdown? Read gramsathi_architecture.md for a strict line-by-line methodology of the Dual-Brain Python stack and CSS grids.

Engineered with precision for a connected, smarter future.