main.py

import logging
from contextlib import asynccontextmanager

import joblib
import pandas as pd
from fastapi import FastAPI, HTTPException
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel, Field, field_validator # Updated import
from typing import Optional # Added for optional fields

# --- Configuration ---
QNR_MODEL_FILENAME = "asd_classifier_model.pkl" # Make sure this matches your uploaded file
CLASS_NAMES_QNR = ["Non-Autistic", "Autistic"] # 0 maps to No ASD, 1 maps to ASD

# Setup logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

ml_models = {} # Dictionary to hold loaded models

# --- Model Loading Logic ---
@asynccontextmanager
async def lifespan(app: FastAPI):
    # Load the ML model during startup
    logger.info(f"Attempting to load questionnaire model: {QNR_MODEL_FILENAME}")
    try:
        ml_models['questionnaire_classifier'] = joblib.load(QNR_MODEL_FILENAME)
        logger.info("Questionnaire model loaded successfully.")
    except Exception as e:
        logger.error(f"Error loading questionnaire model '{QNR_MODEL_FILENAME}': {e}")
        ml_models['questionnaire_classifier'] = None # Indicate loading failure
    yield
    # Clean up the ML models and release the resources
    ml_models.clear()
    logger.info("Cleaned up models.")

# --- FastAPI App ---
app = FastAPI(lifespan=lifespan)

# --- CORS Middleware ---
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"], # Allows all origins
    allow_credentials=True,
    allow_methods=["*"], # Allows all methods
    allow_headers=["*"], # Allows all headers
)

# --- Input Data Schema (Pydantic Model) ---
# Ensure field names match EXACTLY what the model expects
# Added Field defaults and type hints based on your notebook/previous code
class QuestionnaireData(BaseModel):
    A1_Score: int = Field(..., ge=0, le=1)
    A2_Score: int = Field(..., ge=0, le=1)
    A3_Score: int = Field(..., ge=0, le=1)
    A4_Score: int = Field(..., ge=0, le=1)
    A5_Score: int = Field(..., ge=0, le=1)
    A6_Score: int = Field(..., ge=0, le=1)
    A7_Score: int = Field(..., ge=0, le=1)
    A8_Score: int = Field(..., ge=0, le=1)
    A9_Score: int = Field(..., ge=0, le=1)
    A10_Score: int = Field(..., ge=0, le=1)
    age: Optional[float] = Field(25.0, gt=0, le=120) # Made optional with default
    gender: str = Field("m", pattern="^(m|f)$") # Allow only m or f
    ethnicity: str = Field("White-European")
    jaundice: str = Field("no", pattern="^(yes|no)$") # Allow only yes or no
    # Corrected field name typo from 'contry_of_res' to 'country_of_res' if needed
    contry_of_res: str = Field("United States") # Keep original if model expects typo
    used_app_before: str = Field("no", pattern="^(yes|no)$") # Allow only yes or no
    result: Optional[float] = Field(0.0) # This might be recalculated or ignored if it was the target
    age_desc: str = Field("18 and more")
    relation: str = Field("Self")

    # Pydantic v2 validator
    @field_validator('age')
    def check_age(cls, v):
        if v is None:
            return 25.0 # Return default if None is explicitly passed
        if not (0 < v <= 120):
            raise ValueError('Age must be between 0 and 120')
        return v

    # You might add more validators for other fields if needed

# --- Prediction Endpoint (Questionnaire) ---
@app.post("/predict_questionnaire/")
async def predict_questionnaire(data: QuestionnaireData):
    """Receives questionnaire data, preprocesses using loaded pipeline, returns prediction."""
    if ml_models.get('questionnaire_classifier') is None:
         logger.error("Questionnaire model is not loaded.")
         raise HTTPException(status_code=500, detail="Questionnaire model could not be loaded")

    try:
        # Convert Pydantic model to dictionary, then to DataFrame
        input_data = data.model_dump() # Use model_dump() in Pydantic v2
        logger.info(f"Received data: {input_data}")
        input_df = pd.DataFrame([input_data])

        # Recalculate 'result' based on A_Scores if needed by the model pipeline
        # (Assuming 'result' column in training was sum of A*_Score)
        a_scores = [f"A{i}_Score" for i in range(1, 11)]
        input_df['result'] = input_df[a_scores].sum(axis=1)
        logger.info(f"Recalculated result score: {input_df['result'].iloc[0]}")


        # Predict using the loaded pipeline (handles preprocessing)
        prediction = ml_models['questionnaire_classifier'].predict(input_df)
        predicted_class_index = int(prediction[0])

        # Get probability (optional)
        # probability = ml_models['questionnaire_classifier'].predict_proba(input_df)[0]
        # prob_asd = float(probability[1]) # Probability of class 1 (ASD)

        predicted_class_name = CLASS_NAMES_QNR[predicted_class_index]
        logger.info(f"Prediction successful: {predicted_class_name}")

        return {"prediction": predicted_class_name}
        # If returning probability: return {"prediction": predicted_class_name, "probability_asd": prob_asd}

    except ValueError as ve:
         # Catch potential validation errors not caught by Pydantic (e.g., during predict)
         logger.error(f"Value error during prediction: {ve}")
         raise HTTPException(status_code=422, detail=f"Invalid input data: {ve}")
    except Exception as e:
        logger.error(f"Error during questionnaire prediction: {e}", exc_info=True)
        raise HTTPException(status_code=500, detail=f"Prediction error: {e}")

# --- Root Endpoint (Optional) ---
@app.get("/")
async def root():
    return {"message": "Autism Questionnaire Classification API"}