Update app.py

app.py

@@ -1,90 +1,25 @@
-import gradio as gr
 import joblib
-import pandas as pd
-import numpy as np
+from skl2onnx import convert_sklearn
+from skl2onnx.common.data_types import FloatTensorType
 
-ARTIFACT_PATH = "dynamic_pricing_artifact.joblib"
+# Load trained artifact
+artifact = joblib.load("dynamic_pricing_artifact.joblib")
 
-artifact = joblib.load(ARTIFACT_PATH)
+model = artifact["model"]          # trained ML model
+features = artifact["features"]    # feature list
+n_features = len(features)
 
-model = artifact["model"]
-FEATURES = artifact["features"]
-FIXED_FARE = artifact["fixed_fare"]
-RATE_PER_KM = artifact["rate_per_km"]
+print("Number of features:", n_features)
 
-def predict_dynamic_price(
-    zone_id, demand, supply, driver_availability,
-    weather_factor, event_factor, temperature, traffic_index,
-    distance_km, duration_min,
-    hour, day_of_week, is_weekend, month,
-    is_holiday, is_festival
-):
-    # 1️⃣ Base price (distance based)
-    base_price = FIXED_FARE + distance_km * RATE_PER_KM
-
-    # 2️⃣ Build feature row
-    row = {f: 0.0 for f in FEATURES}
-
-    row.update({
-        "zone_id": zone_id,
-        "hour": hour,
-        "day_of_week": day_of_week,
-        "is_weekend": is_weekend,
-        "month": month,
-        "is_holiday": is_holiday,
-        "is_festival": is_festival,
-        "demand": demand,
-        "supply": supply,
-        "driver_availability": driver_availability,
-        "weather_factor": weather_factor,
-        "event_factor": event_factor,
-        "temperature": temperature,
-        "traffic_index": traffic_index,
-        "distance_km": distance_km,
-        "duration_min": duration_min,
-        "base_fare": base_price
-    })
-
-    # 3️⃣ Demand–Supply ratio
-    row["demand_supply_ratio"] = np.clip(demand / (supply + 1), 0, 20)
-
-    df_row = pd.DataFrame([[row[f] for f in FEATURES]], columns=FEATURES)
-
-    # 4️⃣ Predict surge
-    surge = float(model.predict(df_row)[0])
-    surge = np.clip(surge, 1.0, 2.5)   # Rapido/Ola realistic cap
-
-    final_price = base_price * surge
+# Convert to ONNX
+onnx_model = convert_sklearn(
+    model,
+    initial_types=[("input", FloatTensorType([None, n_features]))]
+)
 
-    return round(base_price, 2), round(surge, 3), round(final_price, 2)
+# Save ONNX model
+with open("dynamic_pricing_model.onnx", "wb") as f:
+    f.write(onnx_model.SerializeToString())
 
-demo = gr.Interface(
-    fn=predict_dynamic_price,
-    inputs=[
-        gr.Number(label="Zone ID", value=1),
-        gr.Number(label="Demand", value=150),
-        gr.Number(label="Supply", value=80),
-        gr.Number(label="Driver Availability", value=60),
-        gr.Number(label="Weather Factor", value=1.0),
-        gr.Number(label="Event Factor", value=1.0),
-        gr.Number(label="Temperature (°C)", value=30),
-        gr.Number(label="Traffic Index", value=0.5),
-        gr.Number(label="Distance (km)", value=10),
-        gr.Number(label="Duration (min)", value=20),
-        gr.Number(label="Hour", value=18),
-        gr.Number(label="Day of Week", value=4),
-        gr.Number(label="Is Weekend (0/1)", value=0),
-        gr.Number(label="Month", value=11),
-        gr.Number(label="Is Holiday (0/1)", value=0),
-        gr.Number(label="Is Festival (0/1)", value=0),
-    ],
-    outputs=[
-        gr.Number(label="Base Price"),
-        gr.Number(label="Surge Factor"),
-        gr.Number(label="Final Price"),
-    ],
-    title="Dynamic Pricing Model",
-    description="Realistic dynamic pricing similar to Rapido/Ola"
-)
+print("✅ ONNX model saved as dynamic_pricing_model.onnx")
 
-demo.launch()