app.py

from fastapi import FastAPI
from pydantic import BaseModel
import pandas as pd
import joblib

# ==========================
# Load Model & Encoders
# ==========================

model = joblib.load("xgboost_model.pkl")

le_area = joblib.load("area_encoder.pkl")
le_dept = joblib.load("dept_encoder.pkl")
le_cat = joblib.load("category_encoder.pkl")

# ==========================
# FastAPI App
# ==========================

app = FastAPI(
    title="FFCL Consumable Prediction API",
    description="Predict Turnaround Consumable Quantity",
    version="1.0"
)

# ==========================
# Input Schema
# ==========================

class PredictionInput(BaseModel):
    unit_price: float
    y2017: float
    y2019: float
    y2021: float
    area: str
    dept: str
    category: str

# ==========================
# Home
# ==========================

@app.get("/")
def home():
    return {
        "message": "FFCL Quantity Prediction API is Running"
    }

# ==========================
# Prediction API
# ==========================

@app.post("/predict")
def predict(data: PredictionInput):

    avg_history = (
        data.y2017 +
        data.y2019 +
        data.y2021
    ) / 3

    trend = data.y2021 - data.y2019

    consumption_count = sum([
        data.y2017 > 0,
        data.y2019 > 0,
        data.y2021 > 0
    ])

    area_encoded = le_area.transform([data.area])[0]
    dept_encoded = le_dept.transform([data.dept])[0]
    cat_encoded = le_cat.transform([data.category])[0]

    input_df = pd.DataFrame({

        "Unit Price":[data.unit_price],

        2017:[data.y2017],

        2019:[data.y2019],

        2021:[data.y2021],

        "Avg_History":[avg_history],

        "Trend":[trend],

        "Consumption_Count":[consumption_count],

        "Area_Encoded":[area_encoded],

        "Dept_Encoded":[dept_encoded],

        "Category_Encoded":[cat_encoded]

    })

    prediction = model.predict(input_df)[0]

    return {
        "Predicted Qty": round(float(prediction),2)
    }