"""
FASTAPI + GRADIO SERVING APPLICATION - Production-Ready ML Model Serving
========================================================================

This application provides a complete serving solution for the PaySim model
with both programmatic API access and a user-friendly web interface.

Architecture:
- FastAPI: High-performance REST API with automatic OpenAPI documentation
- Gradio: User-friendly web UI for manual testing and demonstrations
- Pydantic: Data validation and automatic API documentation
"""
from fastapi import FastAPI, HTTPException
from pydantic import BaseModel
from typing import Literal
import gradio as gr

from src.app.inference import predict

app = FastAPI(
    title="PaySim Prediction API",
    description="ML API for predicting Transaction Fraud",
    version="1.0.0"
)

@app.get("/health")
def health_check():
    """
    Health check endpoint for monitoring and load balancer health checks.
    """
    return {"status": "ok"}



class TransactionData(BaseModel):

    """
    This is the schema for PaySim prediction
    
    This schema defines the exact features required for prediction
    Here goes only the columns that are consumed not even the engineered ones"""

    step: int #Represents a unit of time in the real world, with 1 step equating to 1 hour. 
            #The total simulation spans 744 steps, equivalent to 30 days
    type:str #Transaction type
    amount:float #The transaction amount
    oldbalanceOrg: float #Sender balance before transaction
    newbalanceOrig: float #Sender new balance after transactions
    oldbalanceDest: float #receiver old balance
    newbalanceDest: float #receiver new balance


@app.post("/predict")
def predict_endpoint(data: TransactionData):
    try:
        return predict(data.model_dump())   # model_dump() -> plain dict
    except ValueError as e:
        raise HTTPException(status_code=400, detail=str(e))






# ---------------GRADIO--------------------


def gradio_interface(step,type,amount,oldbalanceOrg,
                     newbalanceOrig,oldbalanceDest,newbalanceDest):
    """"This function:
    1. Takes individual form inputs from Gradio UI
    2. Constructs the data dictionary matching the API schema
    3. Calls the same inference pipeline used by the API
    4. Returns user-friendly prediction string"""

    #construct data dictionary matching the schema
    data = {
        "step":int(step),
        "type":type,
        "amount":float(amount),
        "oldbalanceOrg":float(oldbalanceOrg),
        "newbalanceOrig":float(newbalanceOrig),
        "oldbalanceDest": float(oldbalanceDest),
        "newbalanceDest":float(newbalanceDest),
    }

    #Call same inference pipeline as API endpoint
    result = predict(data)
    prob = result["fraud_probability"]
    label = "Fraud" if result["is_fraud"] else "Not Fraud"
    return f"{label} — fraud probability: {prob * 100:.2f}%"



# === GRADIO UI CONFIGURATION ===
demo = gr.Interface(
        fn = gradio_interface,
        inputs=[
            gr.Number(label="Step", value=0, minimum=0, maximum=743),
            gr.Dropdown(["CASH_OUT", "TRANSFER"], label="Transaction Type", value="TRANSFER"),
            gr.Number(label="Amount (EU)", value=0, minimum=0),
            gr.Number(label="Sender's Old Balance"),
            gr.Number(label="Sender's New Balance"),
            gr.Number(label="Receiver's Old Balance"),
            gr.Number(label="Receiver's New Balance"),
        ],


        outputs=gr.Textbox(label= "Fraud Prediction",lines=1),
        title= "Pay Sim transaction Fraud Prediction",
        description="""
                    Predict if the transaction is a fraud or not using Machine Learning

                    Fill in the transaction's details below to get a prediction. The
                    model is using a simple MLP on historical data.""",

        flagging_mode="never",   # no writable flagged/ dir needed (HF runs non-root)
        theme = gr.themes.Soft()
        )


app = gr.mount_gradio_app(
    app,
    demo,
    path="/"          # Gradio UI is the landing page; API still at /predict and /docs
)