AgriSense 🌾

Smart farming assistant for Indian farmers

AgriSense is a full-stack AI farming platform built for Indian agriculture workflows. It combines a multilingual React frontend with a Node/Express backend to help farmers get fast, practical assistance on crops, weather, disease detection, soil health, finance, and government schemes.

✨ Highlights

• 🤖 AI chat assistance in English, Hindi, and Punjabi
• 🎙️ Voice-first chat with microphone recording, transcription, and spoken responses
• 🌤️ Live location-based weather with provider failover
• 🌿 Plant disease detection from uploaded images
• 🪴 Soil image analysis with crop suggestions
• 📈 Crop advisory and financial guidance
• 🏛️ Government scheme discovery
• 🚜 Farmer profile, farm details, and crop history management

🧭 Project Snapshot

Area	Details
Frontend	React 19, Vite, Tailwind CSS, React Router
Backend	Node.js, Express, Mongoose, Multer
AI Services	Groq LLMs, Whisper transcription, optional Python RAG service
Languages	English, Hindi, Punjabi
Deployments	Vercel frontend, Render backend
Public Frontend	agri-sense-lime.vercel.app
Public Backend	agrisense-backend-h3a6.onrender.com

🗂️ Table of Contents

• Repo Structure
• Tech Stack
• Key Features
• Environment Setup
• Local Development
• Verification Status
• Deployment Notes
• Why Issues Like This Happen
• Precautions To Prevent Future Breakage
• Known Gaps
• Suggested Next Checks

Repo Structure

.
|-- backend/
|   |-- controllers/
|   |-- data/
|   |-- middleware/
|   |-- models/
|   |-- python-scripts/
|   |-- routes/
|   |-- services/
|   |-- server.js
|   `-- requirements.txt
|-- frontend/
|   |-- public/
|   |-- src/
|   |-- package.json
|   `-- vite.config.js
|-- assets/
`-- uploads/

Tech Stack

• Frontend: React 19, Vite, Tailwind CSS, React Router, Framer Motion
• Backend: Node.js, Express, Mongoose, Multer, JWT
• AI: Groq LLMs, optional Python RAG service with ChromaDB
• Data: MongoDB
• Deployment: Vercel frontend, Render backend

Key Features

Core Experience

• Multilingual landing page and AI assistant
• Login/signup with local auth and Google sign-in
• Voice input with backend transcription and voice output playback
• Location-aware weather and scheme discovery
• Farm profile capture for more personalized recommendations

Frontend

• Login/signup with local auth and Google sign-in
• Multilingual UI and chat flow
• Voice-enabled chat with microphone recording and TTS playback
• Weather card with location prompt and fallback weather providers
• Scheme discovery with state filtering
• Soil and plant analysis upload flows
• Crop advisory and financial guidance modals
• Farm profile and crop history management

Backend

• POST /api/auth/signup
• POST /api/auth/login
• POST /api/auth/google
• GET /api/auth/me
• PUT /api/auth/profile
• POST /api/auth/crop-history
• DELETE /api/auth/crop-history/:id
• GET /api/weather
• GET /api/schemes
• POST /api/chat
• GET /api/tts
• POST /api/analyze-soil
• POST /api/analyze-plant
• POST /api/crop-advice
• POST /api/financial-guidance
• POST /api/location-schemes
• GET /api/health

Environment Setup

Backend

Create backend/.env using backend/.env.example as the template.

Important variables:

• PORT or APP_PORT
• MONGODB_URI
• JWT_SECRET
• GROQ_API_KEY
• TOMORROW_API_KEY
• WEATHERAPI_KEY
• OPENWEATHERMAP_KEY
• GOOGLE_CLIENT_ID
• GOOGLE_CLOUD_VISION_KEY
• FRONTEND_ORIGINS
• CHAT_MODE
• RAG_SERVICE_URL

Notes:

• Weather now prefers TOMORROW_API_KEY, then falls back through WEATHERAPI_KEY, OPENWEATHERMAP_KEY, Open-Meteo, met.no, and wttr.in.
• AI chat, soil analysis, plant analysis, crop advice, and financial guidance still require a valid GROQ_API_KEY.
• If CHAT_MODE=pure_rag, the backend can auto-start the Python RAG service when needed.

Frontend

Create frontend/.env using frontend/.env.example.

Important variables:

• VITE_API_URL
• VITE_GOOGLE_CLIENT_ID

If VITE_API_URL is missing, the app now auto-selects:

• http://localhost:4001/api for local development
• https://agrisense-backend-h3a6.onrender.com/api for production hosts

Local Development

1. Install dependencies

cd backend
npm install

cd ../frontend
npm install

2. Start backend

cd backend
npm run dev

Default local API:

http://localhost:4001

3. Start frontend

cd frontend
npm run dev

Default frontend URL:

http://localhost:5173

4. Optional RAG setup

🚀 What Makes This Repo Useful

• It supports text, image, and voice-based farmer interactions in one product.
• It is designed for Indian agriculture use cases instead of generic chatbot behavior.
• It has layered fallback behavior for weather, auth recovery, and voice playback.
• It is deployable as separate frontend and backend services with clear health-check targets.

See backend/RAG_SETUP.md.

Verification Status

The following checks were run in this workspace on June 3, 2026:

• Frontend production build: passed
• Backend JavaScript syntax check: passed
• Local backend health endpoint: passed
• Local weather endpoint: passed
• Local chat endpoint: passed
• Local plant analysis endpoint: passed
• Deployed backend health endpoint: passed
• Deployed backend chat endpoint: passed
• Deployed backend plant analysis endpoint: passed

Issues found during verification:

• Frontend lint currently fails with existing ESLint and prop-types issues.
• Deployed backend weather was failing because WEATHERAPI_KEY was not configured in production.
• The voice assistant could silently fail when the browser had not granted microphone access before speech recognition started.
• The repo had no root README.md.

Deployment Notes

Current deployed frontend:

• https://agri-sense-lime.vercel.app

Current backend used by the deployed frontend:

• https://agrisense-backend-h3a6.onrender.com/api

Recommended uptime monitor targets:

• Frontend home: https://agri-sense-lime.vercel.app
• Backend health: https://agrisense-backend-h3a6.onrender.com/api/health
• Backend root: https://agrisense-backend-h3a6.onrender.com

If Vercel and Render are connected to GitHub auto-deploy, pushing to main should trigger a redeploy.

Why Issues Like This Happen

The main failure modes in this project are operational, not just code bugs:

1. Frontend builds can point to the wrong backend if VITE_API_URL is left on a local URL or overridden incorrectly.
2. Protected features depend on a valid JWT token; if the token expires but the UI still thinks the user is logged in, feature requests fail with 401.
3. Weather depended on a third-party key-based API, so missing quota, missing key, or rate limiting could blank the weather card.
4. Browser speech recognition can fail without a visible error if microphone permission is blocked or if start() is called before the browser has granted access.
5. AI features depend on external APIs and can fail if provider credentials are missing, exhausted, or rate-limited.
6. Render free or low-traffic services can cold-start or lag behind Vercel deploy timing.

Precautions To Prevent Future Breakage

1. Keep a working VITE_API_URL in Vercel environment settings and verify it after each deploy.
2. Keep GROQ_API_KEY, TOMORROW_API_KEY, WEATHERAPI_KEY, OPENWEATHERMAP_KEY, GOOGLE_CLIENT_ID, MONGODB_URI, and JWT_SECRET configured in Render.
3. Use an uptime monitor on both the frontend home page and backend health endpoint.
4. After every push, verify auth, weather, chat, and one upload-based endpoint from production.
5. Avoid relying on only one external weather provider; keep a keyed primary provider and multi-provider fallbacks, which this repo now does.
6. Test microphone permission and one spoken query after each frontend deploy, especially on Chrome mobile and desktop.
7. Expire bad sessions cleanly by forcing logout on 401 responses instead of silently bouncing users around the app.
8. Keep deploy-time environment variables documented and versioned in .env.example, but never commit real secrets.

Known Gaps

• No automated backend test suite is present.
• No automated frontend integration or e2e suite is present.
• Frontend lint errors should still be cleaned up in a follow-up pass.
• Several advanced features depend on external service credentials and uptime.

Suggested Next Checks

After deployment, verify:

1. Weather loads on the public homepage.
2. Login/signup works.
3. Chat responds after login.
4. Plant and soil uploads return analysis.
5. Crop advisory and financial guidance respond.