🌡️ Minima AI | Backend Engine

The computational heart of Minima AI, providing real-time LSTM (Long Short-Term Memory) inference for high-precision weather forecasting.

• Live API: https://weather-prediction-backend-x4pv.onrender.com
• Production Frontend: https://minimaai.vercel.app/
• Frontend: https://github.com/Thisarakavishka/Weather-Prediction-Frontend

---

🚀 Overview

The backend is built with FastAPI using a modular service-oriented architecture. It serves a serialized TensorFlow LSTM model that has been optimized for time-series forecasting.

🏗️ Architecture

• Routes: Handles HTTP requests and data validation using Pydantic.
• Services: Manages the AI lifecycle, including sequence preprocessing, feature engineering, and inference.
• Models: Stores the .h5 model files and .pkl scalers.

🛣️ API Endpoints

Method	Endpoint	Description
GET	/health	Returns system status, model readiness, and accuracy metrics.
POST	/predict	Processes 14-day history to return a 24-hour forecast + Confidence scores.

---

🛠️ Local Setup

1. Clone the repository:

   git clone <your-backend-repo-url>
   cd Weather-Prediction-Backend
   

2. Setup Virtual Environment:

   python3 -m venv venv
   source venv/bin/activate
   

3. Install Dependencies:

   pip install -r requirements.txt
   

4. Run the Server:

   uvicorn main:app --reload
   

🧠 Model Specifications

• Type: Bidirectional LSTM (Deep Learning)

• Input Shape: (1, 1, 4) representing [Samples, Timesteps, Features]

• Features: MinTemp_Lag1, MaxTemp_Lag1, Temp_Rolling_7d, Month

• Mean Absolute Error (MAE): 8.62°F