app.py

import gradio as gr
from transformers import pipeline
from fastapi import FastAPI, Form
from fastapi.responses import Response
import uvicorn

# Load model
model_id = "ST-THOMAS-OF-AQUINAS/SCAM"
pipe = pipeline("text-classification", model=model_id)

# Label map (adjust based on your model)
label_map = {0: "author1", 1: "author2"}

def predict(text: str):
    if not text.strip():
        return "⚠️ No input text"

    results = pipe(text)
    label = results[0]["label"]
    score = results[0]["score"]

    if label.startswith("LABEL_"):
        idx = int(label.split("_")[1])
        label = label_map.get(idx, label)

    return f"Prediction: {label}\nConfidence: {round(score * 100, 2)}%"


# --- Gradio Interface (for testing manually) ---
iface = gr.Interface(
    fn=predict,
    inputs=gr.Textbox(label="Enter WhatsApp Message"),
    outputs=gr.Textbox(label="Prediction"),
    title="📲 WhatsApp Scam Detector",
    description="Paste a WhatsApp message and the model will predict its author."
)

# --- FastAPI App (for Twilio) ---
app = FastAPI()

# Twilio will POST with form data (Body=message text)
@app.post("/predict")
async def predict_api(Body: str = Form(...)):
    reply_text = predict(Body)

    # Twilio expects TwiML XML
    twiml_response = f"""<?xml version="1.0" encoding="UTF-8"?>
<Response>
    <Message>{reply_text}</Message>
</Response>"""

    return Response(content=twiml_response, media_type="application/xml")

# Mount Gradio UI inside FastAPI
app = gr.mount_gradio_app(app, iface, path="/")

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