import joblib
import pandas as pd
import numpy as np
from fastapi import FastAPI
from pydantic import BaseModel
from typing import List, Optional
from huggingface_hub import hf_hub_download
from sklearn.impute import SimpleImputer

# --- Constants ---
REPO_ID = "DP1110/mlp-accessibility-model"
MODEL_FILENAME = 'mlp_regressor_model.joblib'
IMPUTER_FILENAME = 'simple_imputer.joblib'

FEATURE_COLUMNS = ['% ASF (Euclidean)', '% Built-Up Area', '% ASF (Network)', '% ASF from Bus Stops ', '% ASF from Bus Stops', '% ASF (Network) ']

# --- Load Model and Imputer ---
loaded_mlp_model = None
loaded_imputer = None

try:
    model_path = hf_hub_download(repo_id=REPO_ID, filename=MODEL_FILENAME)
    imputer_path = hf_hub_download(repo_id=REPO_ID, filename=IMPUTER_FILENAME)

    loaded_mlp_model = joblib.load(model_path)
    loaded_imputer = joblib.load(imputer_path)
    print("Model and imputer loaded successfully!")
except Exception as e:
    print(f"Error loading model or imputer: {e}")

# --- FastAPI Application ---
app = FastAPI()

# --- Pydantic Input Data Model ---
class InputData(BaseModel):
    perc_ASF_Euclidean: Optional[float] = None # Example: 0.5
    perc_Built_Up_Area: Optional[float] = None # Example: 0.5
    perc_ASF_Network: Optional[float] = None # Example: 0.5
    perc_ASF_from_Bus_Stops_: Optional[float] = None # Example: 0.5
    perc_ASF_from_Bus_Stops: Optional[float] = None # Example: 0.5
    perc_ASF_Network_: Optional[float] = None # Example: 0.5

# --- Prediction Endpoint ---
@app.post("/predict")
async def predict(data: InputData):
    if loaded_mlp_model is None or loaded_imputer is None:
        return {"error": "Model or imputer not loaded."}

    # Convert input data to pandas DataFrame
    input_dict = data.dict()

    # Reconstruct input_row with original feature names for imputer/model input
    input_row = {}
    for col in FEATURE_COLUMNS:
        sanitized_col = col.replace(' ', '_').replace('-', '_').replace('%', 'perc').replace('(', '').replace(')', '')
        input_row[col] = input_dict[sanitized_col]

    input_df = pd.DataFrame([input_row])

    # Ensure column order matches training features
    input_df = input_df[FEATURE_COLUMNS]

    # Impute missing values
    input_imputed = loaded_imputer.transform(input_df)
    input_imputed_df = pd.DataFrame(input_imputed, columns=FEATURE_COLUMNS)

    # Make prediction
    prediction = loaded_mlp_model.predict(input_imputed_df)[0]

    return {"predicted_overall_accessibility_score": prediction}

# --- Health Check Endpoint ---
@app.get("/health")
async def health_check():
    return {"status": "ok", "model_loaded": loaded_mlp_model is not None, "imputer_loaded": loaded_imputer is not None}

if __name__ == '__main__':
    import uvicorn
    # Run the FastAPI application using Uvicorn
    uvicorn.run(app, host="0.0.0.0", port=8000)