README.md

---
title: SIH Crop Yield Prediction API
emoji: 🌾
colorFrom: green
colorTo: blue
sdk: docker
pinned: false
license: mit
---

# 🌾 SIH Crop Yield Prediction API

A FastAPI-based machine learning application that predicts crop yields using Random Forest models, deployed on Hugging Face Spaces.

## 🚀 Quick Start - API Usage

### Making Predictions

**POST /predict** - Main prediction endpoint:

```bash
curl -X POST https://your-space-url/predict \
  -H "Content-Type: application/json" \
  -d '{
    "year": 2024,
    "state": "Punjab",
    "crop": "Rice",
    "season": "Kharif", 
    "area": 10.0,
    "production": 25.0,
    "rainfall": 1200,
    "fertilizer": 75,
    "pesticide": 8
  }'
```

**Response:**
```json
{
  "model": "Random Forest",
  "predicted_yield": "2017.7 kg/hectare", 
  "total_expected_production": "20.18 tons",
  "assessment": "Good yield expected"
}
```

### Other Endpoints

- **GET /** - API information
- **GET /health** - Health check
- **GET /docs** - Interactive documentation 
- **GET /available-options** - Supported crops, states, seasons

---

# 📁 Development Documentation

This repository contains a complete pipeline to combine datasets, train multiple models for crop yield prediction, and serve predictions via a FastAPI web service.

## Project Structure

```
.
├── src/                    # Main application code
│   ├── app.py             # FastAPI web service
│   ├── crop_yield_ml_pipeline.py  # ML training pipeline
│   ├── crop_yield_predictor.py    # Prediction utilities
│   └── test_api.py        # API testing utilities
├── scripts/               # Utility scripts
│   ├── combine_datasets.py      # Data combining script
│   ├── analyze_combined_data.py  # Data analysis script
│   ├── results_summary.py       # Results summary script
│   └── test_models.py           # Model evaluation script
├── data/                  # Datasets and CSV files
│   ├── crop_yield.csv          # Historical crop yield data
│   ├── combined_crop_data.csv  # Generated merged dataset
│   └── sample_batch.csv        # Sample data for testing
├── models/                # Trained models and preprocessors
│   ├── preprocessor.pkl        # Data preprocessor
│   ├── random_forest_model.pkl # Random Forest model
│   ├── xgboost_model.json      # XGBoost model
│   └── pytorch_model.pth       # PyTorch model
├── docs/                  # Documentation
│   ├── API_DOCUMENTATION.md    # API usage guide
│   ├── DEPLOYMENT_GUIDE.md     # Deployment instructions
│   ├── CLI_USAGE.md           # CLI usage guide
│   └── PROJECT_SUMMARY.md     # Project overview
├── tests/                 # Test files
│   ├── test_predictor.py      # Predictor tests
│   └── test_server.py         # Server tests
├── deployment/            # Deployment configurations
│   ├── docker-compose.yml     # Docker Compose setup
│   └── render.yaml           # Render deployment config
├── logs/                  # Application logs
├── artifacts/             # Generated plots and visualizations
└── README.md              # This file
```

## Requirements

We recommend using conda.

- Python 3.9
- pandas, numpy, scikit-learn
- xgboost (CPU), seaborn, matplotlib
- pytorch with CUDA (for GPU training)

Example installation (Linux, CUDA 11.8):

```bash
conda install -y pandas numpy scikit-learn xgboost seaborn matplotlib
conda install -y pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia
```

## Usage

### Data Pipeline

1) Combine datasets

```bash
python scripts/combine_datasets.py
```

2) Explore data

```bash
python scripts/analyze_combined_data.py
```

3) Train models

```bash
python src/crop_yield_ml_pipeline.py
```

4) Test models (on a sample) and generate plots

```bash
python scripts/test_models.py
```

5) Show summary

```bash
python scripts/results_summary.py
```

### API Service

1) Start the FastAPI server locally:

```bash
python src/app.py
```

2) Access the API documentation at: `http://localhost:8000/docs`

3) Test the API:

```bash
python src/test_api.py
```

### Deployment

✅ **RAILWAY DEPLOYMENT STATUS: READY**

All critical deployment issues have been resolved:
- **Model Loading**: ✅ Intelligent fallback system implemented  
- **Port Configuration**: ✅ Dynamic PORT variable support
- **API Endpoints**: ✅ All working correctly (/health, /predict, etc.)
- **Error Handling**: ✅ Graceful failure recovery

**See [DEPLOYMENT_FIX_GUIDE.md](DEPLOYMENT_FIX_GUIDE.md) for complete deployment instructions.**

The application is configured for deployment on Railway:

1) The `start.py` file handles Railway's dynamic PORT configuration
2) The `Dockerfile` builds a container with all dependencies  
3) The `railway.json` file configures Railway deployment settings
4) Access the deployed API at: https://sih-2-production.up.railway.app/

## Notes

- By default, .gitignore excludes large artifacts in `data/*.csv`, `artifacts/*.png`, and `logs/*.log`. Remove those rules if you want to commit them.
- Model files in `models/` are included for deployment (required for Railway)
- XGBoost runs with CPU (hist). PyTorch uses CUDA if available.
- The FastAPI service is production-ready and deployed on Railway
- Use the `/docs` endpoint for interactive API testing