How prasooncodes/harvesthub works
Project Type and Purpose
• HarvestHub is a full-stack agricultural recommendation and community forum platform.
• It consists of three major parts:
a. A React.js SPA (client/) for farmers to view recommendations, browse posts, and interact.
b. A Node.js/Express.js REST API backend (server/) implementing MVC patterns, user auth, forum endpoints, and a proxy to ML.
c. A Flask-based ML microservice (ML/) that loads trained crop-recommendation models and exposes prediction endpoints.
File Structure Analysis
• Top-level folders:
• client/ – React project with src/, public/, i18n, components, pages, util for localization, API wrappers, and UI.
• server/ – Express app with index.js entry, config/db.js, Models/, Controllers/, Routes/, Middlewares/, util for JWT.
• ML/ – Python service with app.py, pretrained models (*.pkl), dataset, notebooks.
• Key patterns:
• Backend uses MVC: Models (Mongoose schemas), Controllers (business logic), Routes (HTTP layer), Middlewares (auth).
• ML is decoupled as a microservice (Flask) that the backend can call over HTTP.
• Config & Deployment:
• client/vite.config.js, vercel.json for frontend hosting.
• server uses package.json scripts.
• ML uses pip-installed dependencies.
README Insights
• Confirms MERN + Flask + ML ensemble.
• Notes use of JWT/Bcrypt for security.
• Describes how to start each service locally on different ports (Node, React dev server, Flask).
a. Main Components
1. User’s Browser / React SPA
2. Node.js/Express REST API
3. Flask ML Service
4. MongoDB Database
5. (Optional) Hosting/CI – Vercel for client, Heroku/AWS/GCP for server & ML
6. Third-party libs: JWT, Bcrypt, Mongoose, Scikit-Learn, Pandas
b. Relationships & Data Flow
1. Browser → REST API: HTTP(S) calls for auth, forum, crop data.
2. REST API → MongoDB: Mongoose ORM queries for Users, Posts, Comments, Crops.
3. REST API → Flask ML: HTTP POST with soil params → JSON response with crop recommendation.
4. ML Service → (uses) preloaded model files (minmaxscaler.pkl, model.pkl, standscaler.pkl)
5. Browser ← REST API responses render in UI components.
c. Architectural Patterns
• Frontend: Component-based React, client-side routing, i18n.
• Backend: MVC, layered security (AuthMiddleware), token-based stateless sessions.
• Microservice: Flask service solely for ML inference.
• Data persistence: NoSQL MongoDB, using Mongoose ODM.
d. Key Technologies
• React (Vite), React Router, Context or direct API calls (api.js).
• Express.js, Node.js, JWT & Bcrypt for auth, Mongoose for data models.
• Flask, Scikit-Learn, pickled models for ML.
Detailed Elements to Capture
• Clear labels: “React SPA (port 3000)”, “Express API (port 5000)”, “Flask ML (port 5001)”.
• Distinguish internal modules vs. external dependencies (MongoDB Atlas, third-party APIs).
• Callouts for security: Bcrypt hash & JWT verify in AuthController/Middleware.
• Internationalization subsystem in React (i18n.js and locales).