👶 Stuntify API: Intelligent Inference Engine

📌 Overview

Stuntify API is a specialized backend service designed to bridge the gap between raw health data and actionable stunting diagnostics.

It isn't just a script; it's the intelligent engine working behind the scenes. While the frontend handles the looks, this API handles the thinking. Unlike simple model wrappers, this project implements a complete MLOps Inference Pipeline. It intelligently orchestrates distinct serialized artifacts (Encoders, Scalers, Models) to ensure that every prediction request undergoes the exact same rigorous preprocessing as the training phase.

✨ Key Features

🧠 Multi-Artifact Orchestration (The "Invisible" Brain)

The system acts as a centralized inference unit. It doesn't just "guess"; it reconstructs the mathematical environment by synchronizing 4 key artifacts:

• Gender Encoder: Translates human categories (Laki-laki) into machine vectors.
• Standard Scaler: Normalizes anthropometric data (Age, Height, Weight) to the model's expected distribution.
• Classifier Model: The core logic engine trained for high accuracy.
• Target Decoder: Translates the math result back to a human-readable label (e.g., Severely Stunted).

🛡️ Defensive Architecture

• Schema Validation: Strictly enforces JSON structure to prevent garbage-in-garbage-out.
• Smart Error Handling: Catches edge cases (missing keys, wrong types) and returns meaningful HTTP 400 errors instead of crashing the server.
• Cross-Origin Ready: Fully configured with Flask-CORS to serve requests from any device—be it a Web App, Mobile Phone, or IoT Device.

🛠️ Tech Stack

• Core: Python 3.9+
• Service: Flask (Microframework)
• Computation: NumPy, Scikit-Learn, Joblib
• Security: Flask-CORS (Cross-Origin Resource Sharing enabled)

🚀 The Inference Pipeline (How It Works)

Most tutorials just load a model. This project ensures data consistency through a strict lifecycle for every request:

1. Ingestion: The API receives raw JSON data: {"jenis_kelamin", "umur", "tinggi", "berat"}.
2. Context Reconstruction: It loads the synchronized artifacts (Encoder, Scaler, Classifier, Decoder).
3. Processing: The raw data flows through the pipeline:

   Input Validated -> Gender Encoded -> Metrics Scaled -> Predicted -> Label Decoded

4. Response: The final output is packaged as clean JSON for the client.

🔌 Integration Guide (API Contract)

To talk to Stuntify, send a POST request to /predict.

1. The Request (What you send):

{
    "jenis_kelamin": "Laki-laki",
    "umur": 24,
    "tinggi": 85.5,
    "berat": 10.1
}

2. The Response (What you get):

{
    "prediction": "Severely Stunted"
}

📦 Installation & Usage

1. Clone the Repository

   git clone https://github.com/viochris/API-Stuntify.git
   cd API-Stuntify

2. Install Dependencies

   pip install -r requirements.txt

3. Run the Server Execute the main script directly:

   python api_predict.py

   Output: Running on http://127.0.0.1:5000

4. Consume Endpoint Send the POST request described in the Integration Guide above to http://127.0.0.1:5000/predict.

---

Author: Silvio Christian, Joe "Code that speaks JSON, Logic that saves lives."