Update app.py

app.py

@@ -1,193 +1,186 @@
-import os
-import json
-import uvicorn
-import pandas as pd
-import numpy as np
-from fastapi import FastAPI, HTTPException
-from fastapi.middleware.cors import CORSMiddleware
-from pydantic import BaseModel, ConfigDict
-from catboost import CatBoostClassifier
-from typing import Dict, Any
-
-# ==========================================
-# 1. SETUP & CONFIGURATION
-# ==========================================
-
-app = FastAPI(
-    title="PPD Risk Assessment API",
-    description="AI-powered screening tool for Postpartum Depression Risk (Top 20 Features)",
-    version="1.0.0"
-)
-
-# Enable CORS (Allows your frontend/website to talk to this API)
-app.add_middleware(
-    CORSMiddleware,
-    allow_origins=["*"],  # In production, replace "*" with your frontend URL
-    allow_credentials=True,
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
-
-ARTIFACTS_DIR = "artifacts"
-
-# ==========================================
-# 2. LOAD ARTIFACTS (The Brain)
-# ==========================================
-
-print("⏳ Loading AI Models and Config...")
-
-try:
-    # A. Load Model
-    model_path = os.path.join(ARTIFACTS_DIR, "catboost_model_top20.cbm")
-    if not os.path.exists(model_path):
-        raise FileNotFoundError(f"Model not found at {model_path}")
-    
-    model = CatBoostClassifier()
-    model.load_model(model_path)
-    print("✅ Model Loaded.")
-
-    # B. Load Metadata (Thresholds & Feature List)
-    meta_path = os.path.join(ARTIFACTS_DIR, "catboost_metadata.json")
-    with open(meta_path, "r") as f:
-        metadata = json.load(f)
-    
-    TOP_FEATURES = metadata["top_features"]
-    
-    # --- THRESHOLD CONFIGURATION ---
-    # Originally 0.3, but we updated to 0.5 to reduce False Positives 
-    # based on your testing (Patient with 43% risk should be Low Risk).
-    THRESHOLD = 0.3 
-    
-    print(f"✅ Metadata Loaded. Threshold set to: {THRESHOLD}")
-
-    # C. Load UI Schema (For Frontend Dynamic Forms)
-    ui_path = os.path.join(ARTIFACTS_DIR, "model_ui_schema.json")
-    with open(ui_path, "r") as f:
-        ui_schema = json.load(f)
-    print("✅ UI Schema Loaded.")
-
-except Exception as e:
-    print(f"❌ CRITICAL ERROR LOADING ARTIFACTS: {e}")
-    raise e
-
-
-# ==========================================
-# 3. DATA VALIDATION (Pydantic)
-# ==========================================
-
-class PatientData(BaseModel):
-    data: Dict[str, Any]
-
-    # Updated for Pydantic V2 (No warnings)
-    model_config = ConfigDict(
-        json_schema_extra={
-            "example": {
-                "data": {
-                    "Need for Support": "High",
-                    "Recieved Support": "Low",
-                    "Abuse": "Yes",
-                    "Disease before pregnancy": "None",
-                    "Occupation before latest pregnancy": "Housewife",
-                    "Pregnancy plan": "Unplanned",
-                    "Relationship with husband": "Bad",
-                    "Major changes or losses during pregnancy": "Yes",
-                    "Relationship with the in-laws": "Bad",
-                    "Birth compliancy": "No",
-                    "Relationship between father and newborn": "Bad",
-                    "Education Level": "Secondary",
-                    "Family type": "Nuclear",
-                    "Diseases during pregnancy": "Yes",
-                    "Trust and share feelings": "No",
-                    "Relationship with the newborn": "Average",
-                    "Occupation After Your Latest Childbirth": "Unemployed",
-                    "Age": 24,
-                    "Addiction": "No",
-                    "Husband's education level": "Secondary"
-                }
-            }
-        }
-    )
-
-# ==========================================
-# 4. HELPER FUNCTIONS
-# ==========================================
-
-def preprocess_input(raw_data: dict) -> pd.DataFrame:
-    """
-    Cleans input dictionary: lowercases strings, handles missing cols, sorts cols.
-    """
-    clean_data = {}
-    
-    # 1. Lowercase string inputs to match model training
-    for k, v in raw_data.items():
-        if isinstance(v, str):
-            clean_data[k] = v.lower()
-        else:
-            clean_data[k] = v
-            
-    # 2. Create DataFrame
-    df = pd.DataFrame([clean_data])
-    
-    # 3. Ensure all Top 20 columns exist (fill missing with 'unknown')
-    # This prevents crashing if the user leaves a field blank
-    for col in TOP_FEATURES:
-        if col not in df.columns:
-            df[col] = "unknown"
-            
-    # 4. Reorder columns to match exactly what the model expects
-    df = df[TOP_FEATURES]
-    
-    return df
-
-# ==========================================
-# 5. API ENDPOINTS
-# ==========================================
-
-@app.get("/")
-def home():
-    """Health check endpoint."""
-    return {"status": "online", "model": "CatBoost Top20", "threshold": THRESHOLD}
-
-@app.get("/config")
-def get_ui_config():
-    """
-    Returns the UI Schema (Dropdown options, Labels).
-    Your Frontend (React/Streamlit) should call this to build the form automatically.
-    """
-    return ui_schema
-
-@app.post("/predict")
-def predict_risk(payload: PatientData):
-    """
-    Main prediction endpoint.
-    Accepts patient data -> Returns Risk Status & Probability.
-    """
-    try:
-        # 1. Preprocess Data
-        input_df = preprocess_input(payload.data)
-        
-        # 2. Get Probability of Risk (Class 1)
-        # [0][1] index gets the probability of "Positive" (Risk)
-        risk_prob = model.predict_proba(input_df)[0][1]
-        
-        # 3. Apply Threshold Logic
-        is_high_risk = bool(risk_prob >= THRESHOLD)
-        
-        return {
-            "prediction": "HIGH RISK" if is_high_risk else "LOW RISK",
-            "risk_probability": round(float(risk_prob), 4),
-            "threshold_used": THRESHOLD,
-            "flag": 1 if is_high_risk else 0,
-            "clinical_note": "Refer to specialist" if is_high_risk else "Standard monitoring"
-        }
-        
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=str(e))
-
-# ==========================================
-# 6. RUNNER
-# ==========================================
-if __name__ == "__main__":
-    # Updated to 0.0.0.0 and 7860 for Docker/Hugging Face compatibility
-    print("🚀 Server starting...")
-    uvicorn.run(app, host="0.0.0.0", port=7860)
+import os
+import json
+import uvicorn
+import pandas as pd
+import numpy as np
+
+from fastapi import FastAPI, HTTPException
+from fastapi.middleware.cors import CORSMiddleware
+from pydantic import BaseModel, ConfigDict
+from catboost import CatBoostClassifier
+from typing import Dict, Any
+
+# ==========================================
+# 1. APP SETUP
+# ==========================================
+
+app = FastAPI(
+    title="PPD Risk Assessment API",
+    description="AI-powered screening tool for Postpartum Depression Risk (Top 20 Features)",
+    version="1.0.0"
+)
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+# ==========================================
+# 2. PATH CONFIG (HF SAFE)
+# ==========================================
+
+BASE_DIR = os.path.dirname(os.path.abspath(__file__))
+ARTIFACTS_DIR = os.path.join(BASE_DIR, "artifact_final")
+
+MODEL_FILE = "catboost_model_top20.cbm"
+META_FILE = "catboost_metadata.json"
+UI_FILE = "model_ui_schema.json"
+
+THRESHOLD = 0.3  # calibrated threshold
+
+# ==========================================
+# 3. LOAD ARTIFACTS (FAIL FAST)
+# ==========================================
+
+print("⏳ Loading AI Models and Config...")
+
+try:
+    print("📁 Artifacts directory:", ARTIFACTS_DIR)
+    print("📄 Files found:", os.listdir(ARTIFACTS_DIR))
+
+    # --- Load Model ---
+    model_path = os.path.join(ARTIFACTS_DIR, MODEL_FILE)
+    if not os.path.exists(model_path):
+        raise FileNotFoundError(f"Model not found at {model_path}")
+
+    model = CatBoostClassifier()
+    model.load_model(model_path)
+    print("✅ CatBoost model loaded")
+
+    # --- Load Metadata ---
+    meta_path = os.path.join(ARTIFACTS_DIR, META_FILE)
+    with open(meta_path, "r") as f:
+        metadata = json.load(f)
+
+    TOP_FEATURES = metadata["top_features"]
+    print(f"✅ Metadata loaded ({len(TOP_FEATURES)} features)")
+
+    # --- Load UI Schema ---
+    ui_path = os.path.join(ARTIFACTS_DIR, UI_FILE)
+    with open(ui_path, "r") as f:
+        ui_schema = json.load(f)
+
+    print("✅ UI schema loaded")
+    print(f"🚦 Threshold set to {THRESHOLD}")
+
+except Exception as e:
+    print("❌ CRITICAL ERROR LOADING ARTIFACTS")
+    raise e
+
+# ==========================================
+# 4. REQUEST SCHEMA
+# ==========================================
+
+class PatientData(BaseModel):
+    data: Dict[str, Any]
+
+    model_config = ConfigDict(
+        json_schema_extra={
+            "example": {
+                "data": {
+                    "Need for Support": "High",
+                    "Recieved Support": "Low",
+                    "Abuse": "Yes",
+                    "Disease before pregnancy": "None",
+                    "Occupation before latest pregnancy": "Housewife",
+                    "Pregnancy plan": "Unplanned",
+                    "Relationship with husband": "Bad",
+                    "Major changes or losses during pregnancy": "Yes",
+                    "Relationship with the in-laws": "Bad",
+                    "Birth compliancy": "No",
+                    "Relationship between father and newborn": "Bad",
+                    "Education Level": "Secondary",
+                    "Family type": "Nuclear",
+                    "Diseases during pregnancy": "Yes",
+                    "Trust and share feelings": "No",
+                    "Relationship with the newborn": "Average",
+                    "Occupation After Your Latest Childbirth": "Unemployed",
+                    "Age": 24,
+                    "Addiction": "No",
+                    "Husband's education level": "Secondary"
+                }
+            }
+        }
+    )
+
+# ==========================================
+# 5. PREPROCESSING
+# ==========================================
+
+def preprocess_input(raw_data: dict) -> pd.DataFrame:
+    clean_data = {}
+
+    for k, v in raw_data.items():
+        if isinstance(v, str):
+            clean_data[k] = v.lower()
+        else:
+            clean_data[k] = v
+
+    df = pd.DataFrame([clean_data])
+
+    # ensure all features exist
+    for col in TOP_FEATURES:
+        if col not in df.columns:
+            df[col] = "unknown"
+
+    df = df[TOP_FEATURES]
+    return df
+
+# ==========================================
+# 6. API ENDPOINTS
+# ==========================================
+
+@app.get("/")
+def health():
+    return {
+        "status": "online",
+        "model": "CatBoost Top-20",
+        "threshold": THRESHOLD
+    }
+
+@app.get("/config")
+def get_ui_config():
+    return ui_schema
+
+@app.post("/predict")
+def predict(payload: PatientData):
+    try:
+        input_df = preprocess_input(payload.data)
+        risk_prob = model.predict_proba(input_df)[0][1]
+        is_high_risk = risk_prob >= THRESHOLD
+
+        return {
+            "prediction": "HIGH RISK" if is_high_risk else "LOW RISK",
+            "risk_probability": round(float(risk_prob), 4),
+            "threshold_used": THRESHOLD,
+            "flag": int(is_high_risk),
+            "clinical_note": (
+                "Refer to specialist"
+                if is_high_risk
+                else "Standard monitoring"
+            )
+        }
+
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+
+# ==========================================
+# 7. RUNNER (HF / DOCKER)
+# ==========================================
+
+if __name__ == "__main__":
+    print("🚀 Starting server...")
+    uvicorn.run(app, host="0.0.0.0", port=7860)