PranaAI — Your Personalized Ayurvedic Concierge Agent

An Intelligent Multi-Agent Ayurvedic System Leveraging Google Agent Development Kit (ADK) and Gemini 2.5 Flash for Safe, Personalised Wellness Guidance.

---

1. Overview

PranaAI is a state-of-the-art traditional wellness assistant designed to bring the age-old wisdom of Ayurveda into the modern digital era. It utilizes the Google Agent Development Kit (ADK) and the Gemini 2.5 Flash model to analyze a user's symptoms, lifestyle, age, and duration of concerns to calculate their dominant Dosha constitution (Vata, Pitta, Kapha, or a combination).

By referencing a curated Ayurvedic knowledge database and using context-aware memory, PranaAI provides highly tailored, safe, and actionable daily routines, dietary recommendations, and traditional herbal suggestions, accompanied by a dynamic PDF report generation engine.

---

2. Multi-Agent Architecture

PranaAI is powered by a sequential multi-agent pipeline built on top of the Google ADK. Each agent executes a highly specialized role:

[User Message]
      │
      ▼
┌──────────────┐
│ Intake Agent │ ──► Collects symptoms, lifestyle, age range, and duration
└──────────────┘
      │
      ▼
┌─────────────────┐
│ Prakriti Expert │ ──► Analyzes intake summary and determines Dosha breakdown
└─────────────────┘
      │
      ▼
┌─────────────────┐
│ Knowledge Agent │ ──► Connects to local MCP servers to fetch validated guidance
└─────────────────┘
      │
      ▼
┌──────────────────────┐
│ Recommendation Agent │ ──► Synthesizes data into dietary and routine tips
└──────────────────────┘
      │
      ▼
┌──────────────┐
│ Safety Guard │ ──► Runs a final validation check for medical red flags
└──────────────┘
      │
      ▼
[Final Clean Response]

• Intake Agent: Empathically collects user data (symptom, lifestyle details, age range, duration) one question at a time and packages it into a structured JSON payload once complete.
• Prakriti Expert: Interprets the intake summary and performs keyword-based or semantic mapping to calculate the percentage breakdown of Vata, Pitta, and Kapha.
• Knowledge Agent: Connects dynamically to local Model Context Protocol (MCP) servers (including the Seasonal Awareness Server) to retrieve traditional characteristics and seasonal guidance.
• Recommendation Agent: Formulates highly personalized advice (such as morning checklists and customized food evaluations) based on the user's history and current constitution.
• Safety Guard: Serves as a final check block to scan for clinical red flags (e.g. severe chest pain, shortness of breath) and instantly escalates to a medical disclaimer and suggestion to consult a doctor.

---

3. Key Features

• Follow-up Q&A Memory: Supports continuous follow-up conversations using context memory. Users can ask specific queries (e.g. "how to use basil seeds without losing weight"), and the agent provides direct, actionable answers rather than repeating general dosha evaluations.
• Safety Guardrails (Red Flag Detection): Scans user messages for critical symptoms and immediately suspends suggestions, returning a prominent warning block to see a doctor.
• PDF Report Generation: Once the intake is complete and the Dosha state is generated, a clean, nature-themed "Download Weekly Wellness Report" button appears. Users can click this to dynamically download a beautifully styled PDF report containing their personalized Ayurvedic profile, clinical reasoning, and database guidelines.

---

4. Tech Stack

• Frontend: React, Vite, CSS3 (with nature-themed design tokens and hover animations)
• Backend: Flask, Flask-CORS, Python 3.12, ReportLab (for PDF generation)
• AI Engine: Gemini 2.5 Flash, Google Agent Development Kit (ADK)
• Model Context Protocol (MCP): Stdio MCP Server for seasonal and health knowledge lookups
• Deployment Platform: Vercel (for frontend static hosting and backend python serverless functions)

---

5. Live Demo

🌐 Live Demo Link: [Live Demo Link Here]

---

6. Local Setup Instructions

Follow these simple steps to run the frontend and backend servers locally:

Prerequisites

• Python 3.10+
• Node.js 18+

Step 1: Clone the Repository

git clone https://github.com/udisha1/ayurcare-agent.git
cd ayurcare-agent

Step 2: Configure Environment Variables

Create a .env file in the root directory:

GOOGLE_API_KEY=your_gemini_api_key_here

Step 3: Install Backend Dependencies & Run

1. Create and activate a Python virtual environment:

   python -m venv venv
   # On Windows:
   .\venv\Scripts\activate
   # On macOS/Linux:
   source venv/bin/activate

2. Install the required libraries:

   pip install -r requirements.txt

3. Run the Flask server:

   python api/agent.py

   The backend will start on http://127.0.0.1:5000.

Step 4: Install Frontend Dependencies & Run

1. Open a new terminal window/tab and navigate to the frontend/ folder:

   cd frontend

2. Install Node packages:

   npm install

3. Start the Vite development server:

   npm run dev

   The frontend will start on http://localhost:5173/.

Step 5: Test the Integration

Open your browser and navigate to http://localhost:5173/ to interact with your local instance of PranaAI.