app.py

import os
import json
import uvicorn
import pandas as pd
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
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],  # Replace "*" with your frontend URL in production
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# ==========================================
# 2. ARTIFACT PATH SETUP
# ==========================================

BASE_DIR = os.path.dirname(os.path.abspath(__file__))
ARTIFACTS_DIR = os.path.join(BASE_DIR, "artifacts_final")  # Hugging Face compatible path

print("ARTIFACTS DIR:", ARTIFACTS_DIR)
print("EXISTS:", os.path.exists(ARTIFACTS_DIR))

# ==========================================
# 3. LOAD ARTIFACTS
# ==========================================

print(" Loading AI Models and Config...")

try:
    # A. Load CatBoost 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
    meta_path = os.path.join(ARTIFACTS_DIR, "catboost_metadata.json")
    if not os.path.exists(meta_path):
        raise FileNotFoundError(f"Metadata not found at {meta_path}")

    with open(meta_path, "r") as f:
        metadata = json.load(f)

    TOP_FEATURES = metadata.get("features_used", [])
    THRESHOLD = metadata.get("thresholds", {}).get("optimal_balanced", 0.5)

    print(f" Metadata Loaded. Threshold set to: {THRESHOLD}")

    # C. Load UI Schema
    ui_path = os.path.join(ARTIFACTS_DIR, "model_ui_schema.json")
    if not os.path.exists(ui_path):
        raise FileNotFoundError(f"UI schema not found at {ui_path}")

    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

# ==========================================
# 4. DATA VALIDATION
# ==========================================

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. HELPER FUNCTION
# ==========================================

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])
    
    # Fill missing features
    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 home():
    return {"status": "online", "model": "CatBoost Top20", "threshold": THRESHOLD}

@app.get("/config")
def get_ui_config():
    return ui_schema

@app.post("/predict")
def predict_risk(payload: PatientData):
    try:
        input_df = preprocess_input(payload.data)
        risk_prob = model.predict_proba(input_df)[0][1]
        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))

# ==========================================
# 7. RUNNER
# ==========================================

if __name__ == "__main__":
    print(" Server starting...")
    uvicorn.run(app, host="0.0.0.0", port=7860)