Update app.py

app.py

@@ -1,33 +1,55 @@
 import gradio as gr
 import pickle
 import pandas as pd
+import numpy as np
 from huggingface_hub import hf_hub_download
 
-# Load your model from your existing model repository
-repo_id = "YOUR_USERNAME/random-forest-model"  # CHANGE THIS
-model_path = hf_hub_download(repo_id=repo_id, filename="random_forest_model.pkl")
+# Download model from MODEL repository (not stored in Space)
+repo_id = "DevNumb/random-forest-model-reduced"  # Your MODEL repo
 
+print("Downloading model (1.64 GB)...")
+model_path = hf_hub_download(
+    repo_id=repo_id,
+    filename="random_forest_model_reduced.pkl",
+    local_dir="/tmp/model"  # Downloads to Space's 50 GB disk
+)
+
+print("Loading model...")
 with open(model_path, 'rb') as f:
     model = pickle.load(f)
+print("✅ Model ready!")
+
+# Feature names from your JSON
+FEATURE_NAMES = [
+    'OA_TEMP', 'OA_TEMP_WB', 'Hour', 'Weekday', 'Month',
+    'CHL_STA_1', 'CHL_STA_2', 'CHL_STA_3',
+    'CHL_COMP_SPD_CTRL_1', 'CHL_COMP_SPD_CTRL_2', 'CHL_COMP_SPD_CTRL_3',
+    'CT_FAN_SPD_CTRL_1', 'CT_FAN_SPD_CTRL_2', 'CT_FAN_SPD_CTRL_3',
+    'CHL_CD_FLOW_1', 'CHL_CD_FLOW_2', 'CHL_CD_FLOW_3',
+    'CWL_SEC_LOAD'
+]
+
+def predict(*args):
+    input_data = pd.DataFrame([list(args)], columns=FEATURE_NAMES)
+    prediction = model.predict(input_data)[0]
+    
+    try:
+        probabilities = model.predict_proba(input_data)[0]
+        confidence = np.max(probabilities)
+        result = f"**Prediction:** {prediction:.2f}\n\n**Confidence:** {confidence:.2%}"
+    except:
+        result = f"**Prediction:** {prediction}"
+    
+    return result
 
-# Prediction function - adjust feature names as needed
-def predict(feature1, feature2, feature3):
-    df = pd.DataFrame([[feature1, feature2, feature3]], 
-                     columns=['feature1', 'feature2', 'feature3'])
-    pred = model.predict(df)[0]
-    prob = model.predict_proba(df)[0].max()
-    return f"Prediction: {pred} (Confidence: {prob:.2f})"
+# Create input fields for all 18 features
+inputs = [gr.Number(label=name) for name in FEATURE_NAMES]
 
-# Create the API endpoint
 demo = gr.Interface(
     fn=predict,
-    inputs=[
-        gr.Number(label="Feature 1"),
-        gr.Number(label="Feature 2"), 
-        gr.Number(label="Feature 3")
-    ],
-    outputs=gr.Textbox(label="Result"),
-    title="Random Forest API"
+    inputs=inputs,
+    outputs=gr.Markdown(label="Result"),
+    title="HVAC Chiller System Prediction API"
 )
 
 demo.launch()