Empowering Indian agriculture with AI, real-time data, and a beautiful user experience.
FarmGenius is an advanced, multi-agent agricultural assistant designed for Indian farmers, agri-entrepreneurs, and researchers. It combines AI-driven chat, real-time yield prediction, market price forecasting, and disease diagnosis in a unified, modern platform.
- AI Chat Assistant: Nationwide agricultural expertise powered by Gemini LLMs
- Market Price Prediction: Real-time and historical mandi prices, time series forecasting
- Yield Prediction: Weather-aware, map-based, ML-driven yield estimates
- Disease Diagnosis: Crop image upload and instant disease detection
- Professional UI/UX: Responsive, modern design
- Multi-Agent Chat: Conversational assistant with intent routing and context memory
- Market Data: AgMarkNet scraping, static fallback, and forecasting
- Yield Prediction: Google Maps farm selection, OpenWeather integration, ML models
- Disease Detection: Deep learning or Gemini Vision analysis of crop images
- Nationwide Coverage: All major crops, markets, and Indian states
- Modern UI: Shareable conversations, streaming responses, mobile-first
Backend:
- Python, FastAPI
- Multi-agent system (Gemini LLM integration)
- Web scraping (AgMarkNet)
- ML models: yield, disease
- Caching, static data fallback
Frontend:
- React (Vite)
- Modular feature structure
- Google Maps JS API, OpenWeather API
- Modern CSS Modules, context/hooks
404/
├── backend/
│ ├── app/
│ │ ├── api/ # FastAPI endpoints (market, yield, disease, chat)
│ │ ├── core/ # Multi-agent logic, AI services, config
│ │ ├── data/ # Static/fallback data
│ │ ├── models/ # Data models
│ │ ├── services/ # Scrapers, ML services
│ │ └── main.py # FastAPI app entry point
│ ├── requirements.txt
│ └── .env
├── frontend/
│ ├── src/
│ │ ├── features/
│ │ │ ├── chat/ # Chat UI, avatars, animation
│ │ │ ├── market/ # Market price prediction UI
│ │ │ ├── yield/ # Yield prediction with map/weather
│ │ ├── components/ # Shared components
│ │ ├── services/ # API calls
│ │ └── theme/ # Theme (dark/light mode)
│ ├── public/
│ ├── package.json
│ └── .env
└── README.md
flowchart TD
subgraph Frontend [React Frontend]
ChatUI[Chat Assistant UI]
MarketUI[Market Predictor UI]
YieldUI[Yield Predictor UI]
DiseaseUI[Disease Diagnosis UI]
MapComp[Google Maps Integration]
WeatherComp[Weather API Integration]
end
subgraph Backend [FastAPI Backend]
APIMain[main.py]
MarketEP[Market Endpoints]
YieldEP[Yield Endpoints]
DiseaseEP[Disease Endpoints]
ChatEP[Chat Endpoints]
Agents[Multi-Agent System]
Scraper[AgMarkNet Scraper]
YieldML[Yield ML Model]
DiseaseML[Disease ML Model]
StaticData[Static Market Data]
end
subgraph External
GoogleMaps[Google Maps API]
OpenWeather[OpenWeather API]
AgMarkNet[AgMarkNet Portal]
Gemini[Gemini LLM]
EarthEngine[Earth Engine API]
end
ChatUI -->|REST| ChatEP
MarketUI -->|REST| MarketEP
YieldUI -->|REST| YieldEP
DiseaseUI -->|REST| DiseaseEP
MapComp --> GoogleMaps
WeatherComp --> OpenWeather
MarketEP --> Scraper
MarketEP --> StaticData
YieldEP --> YieldML
YieldEP --> WeatherComp
YieldEP --> EarthEngine
DiseaseEP --> DiseaseML
ChatEP --> Agents
Agents --> Gemini
Scraper --> AgMarkNet
- Header: Logo, navigation (Chat, Market, Yield, Disease), dark/light toggle
- Main: Quick links to features, latest market/yield highlights, user tips
- Left: Conversation history, agent avatars
- Center: Chat window (Markdown, streaming), input box, send button
- Right: Contextual tips, share conversation
- Top: Select commodity, state, market
- Main: Price chart (historical & predicted), market info, refresh button
- Side: Data source info, last updated
- Map Panel: Google Maps with farm selection
- Form: Crop, season, acreage, weather (auto-filled)
- Output: Predicted yield, recommendations
- Upload: Image upload box
- Result: Detected disease, advice, treatment suggestions
cd backend
python -m venv venv
source venv/bin/activate # or venv\Scripts\activate on Windows
pip install -r requirements.txt
cp .env.example .env # Add your API keys
uvicorn app.main:app --reloadcd frontend
npm install
cp .env.example .env # Add your API keys
npm run dev- Open
http://localhost:5173for the frontend. - Interact with chat, market, yield, and disease features.
- Backend runs on
http://localhost:8000by default.
- AgMarkNet, OpenWeather, Google Maps, Gemini LLM
- Built for hackathons and real-world agricultural impact
For questions or demo requests, open an issue or contact the maintainers.
FarmGenius uses a multi-agent architecture on the backend. Instead of a single AI module, there are multiple specialized agents:
- Yield Agent: Handles crop yield predictions.
- Disease Agent: Analyzes plant images for disease detection.
- Market Agent: Scrapes and predicts market prices.
- Chat Agent: Provides conversational AI support for agricultural queries.
Why Multi-Agent?
- Separation of Concerns: Each agent is focused and easier to maintain/upgrade.
- Extensibility: New agents (e.g., weather, pest, soil) can be added easily.
- Parallelism: Agents can work independently or together, improving performance and reliability.
This protocol allows agents to share context and state for smarter responses:
- Context Sharing: Agents exchange info (e.g., user location, crop type, previous queries) to provide holistic answers.
- Stateful Interactions: Enables the system to remember past interactions and tailor recommendations.
- Example: If a user asks for a yield prediction and then asks about disease risk for the same crop, the Disease Agent can access info from the Yield Agent’s context.
- Disease Detection: Users upload crop images. The backend runs a vision model (deep learning or Gemini Vision API) to analyze the image, returning the detected disease, confidence score, and actionable tips.
- Pipeline: Image → Preprocessing → AI Model → Analysis → Response
- Gemini 1.5 Pro: The chat agent uses Google’s Gemini model for natural, context-aware conversations.
- Coverage: Answers agri questions for all of India.
- Markdown & Rich Formatting: Responses are formatted for clarity (bullet points, code blocks, etc.).
- Streaming & Sharing: The chat UI supports streaming responses and lets users share conversations.
- Intent Routing: The chat agent can delegate queries to domain agents (e.g., yield, market) and aggregate their responses.
- Context Awareness: Maintains conversation context for multi-turn dialogues.
- AgMarkNet Scraper: Fetches live mandi prices for major crops across India.
- Nationwide Coverage: Supports all major agricultural markets.
- Caching: Avoids excessive requests by caching recent data.
- Integration: Market data is used for both chat responses and the market predictor UI.
- Google Maps Integration: Users can select their farm location for precise predictions.
- Weather & Soil Data: Uses APIs and default coefficients for region, season, and crop type.
- Recommendations: Provides actionable tips along with yield estimates.
- FastAPI Backend: Modular, async, and production-ready.
- React + Tailwind Frontend: Beautiful, mobile-first UI with animated transitions.
- REST API: Clean, versioned endpoints for all features.
- Environment Variables: Secure management of API keys for Gemini, Google Maps, and OpenWeather.
In summary:
- User interacts with the frontend (uploads image, asks question, selects location).
- Frontend calls the backend API (RESTful endpoints).
- Backend routes the request to the appropriate agent(s) using the multi-agent system.
- Agents process the request, possibly sharing context or fetching external data.
- Response is aggregated and sent back to the frontend.
- Frontend displays the result with rich UI/UX.
- Yield Prediction: Predict crop yields using AI, Google Maps for location, and weather/soil data.
- Disease Detection: Upload plant images to detect diseases using AI models.
- Market Price Prediction: Get real-time prices for crops across India via AgMarkNet scraping.
- AI Chat Assistant: Ask agricultural questions and get contextual answers (powered by Gemini).
- Modern UI: Responsive, mobile-first React + Tailwind interface.
- Multi-agent Backend: Extensible system for integrating more AI/ML services.
- Frontend: React, Tailwind CSS, Vite, React Router, Axios
- Backend: FastAPI, Python, Gemini API, OpenWeather, Google Maps, AgMarkNet Scraper
- Communication: REST API (
/api/v1/...) - Multi-Agent System: Context protocol for extensibility
[ React + Tailwind ] <--> [ FastAPI Backend ] <--> [ AI/ML APIs, Scrapers, DB ]
- Python 3.10+
- Node.js 18+
- npm 9+
cd backend
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env # Set API keys for Gemini, Google Maps, OpenWeather
uvicorn app.main:app --reloadcd frontend
npm install
cp .env.example .env # Set VITE_GOOGLE_MAPS_API_KEY
npm run dev- Yield Predictor: Enter crop, area, state, and soil/weather info. Select location on map. Get instant yield prediction and recommendations.
- Disease Detector: Upload or drag/drop a plant image. Get AI-powered disease analysis and management tips.
- Market Prices: View real-time mandi prices for major crops, powered by live scraping.
- AI Chat: Ask any agri question and get a contextual, India-specific answer.
POST /api/v1/yield/predict— Predict crop yieldPOST /api/v1/disease/detect— Detect plant disease from imageGET /api/v1/market/prices— Get current mandi pricesPOST /api/v1/chat/ask— AI chat assistant
- Backend: FastAPI, Python, Gemini API, OpenWeather, Google Maps, AgMarkNet Web Scraper
- Frontend: React, Tailwind CSS, Vite, React Router, React Icons, Axios
- Testing: Pytest (backend), ESLint (frontend)
- User enters crop, area, state, soil, and weather data.
- Optionally selects exact farm location via Google Maps.
- Backend uses crop coefficients, weather, and soil factors to predict yield.
- Returns yield per hectare and total production, plus actionable recommendations.
- User uploads a plant image.
- AI model analyzes the image and returns disease diagnosis and management advice.
- Backend scrapes AgMarkNet for real-time mandi prices across India.
- Data is cached and aggregated for fast, reliable access.
- Seamlessly integrated into the platform with a modern, responsive chat UI (matching the overall design system).
- Powered by Gemini 1.5 Pro for natural, conversational, and context-aware responses.
- Supports Markdown formatting, streaming replies, and sharing of conversations.
- Understands context from other features (e.g., yield, market, disease) for truly smart, multi-turn assistance.
- Designed specifically for Indian agriculture: covers all major crops, regions, and agri practices.
- Backend is built to allow easy addition of more AI/ML agents and features.
- Built for Hackathons by 404
- Special thanks to: FastAPI, React, Google, OpenAI, AgMarkNet, and the open-source community.
MIT