app.py

import gradio as gr
from src.inference import predict_ticket  # Uses the fixed inference.py with nltk fix

def predict_interface(ticket_text):
    try:
        result = predict_ticket(ticket_text)
        issue = result.get("issue_type", "Unknown")
        urgency = result.get("urgency_level", "Unknown")
        entities = result.get("entities", {})

        # Format entity output
        lines = []
        for key in ["products", "dates", "complaints"]:
            vals = entities.get(key, [])
            lines.append(f"{key.capitalize()}: {', '.join(vals) if vals else 'None'}")
        entities_str = "\n".join(lines)

        return issue, urgency, entities_str

    except Exception as e:
        return f"Prediction error: {str(e)}", "Prediction error", "Prediction error"

# Build the Gradio interface
iface = gr.Interface(
    fn=predict_interface,
    inputs=gr.Textbox(
        label="📝 Customer Support Ticket",
        lines=6,
        placeholder=(
            "Describe your issue clearly.\n"
            "Example: 'I returned the washing machine on 10th May but no refund received.'"
        )
    ),
    outputs=[
        gr.Textbox(label="📌 Predicted Issue Type"),
        gr.Textbox(label="⏱️ Predicted Urgency Level"),
        gr.Textbox(label="🧠 Extracted Entities"),
    ],
    title="📬 Customer Support Ticket Analyzer",
    description=(
        "Paste a customer support ticket. This tool uses machine learning to predict:\n\n"
        "- 📌 Issue Type (e.g., Late Delivery, Refund)\n"
        "- ⏱️ Urgency Level (Low / Medium / High)\n"
        "- 🧠 Extracted Entities (Products, Dates, Complaints)"
    ),
    allow_flagging="never"
)

if __name__ == "__main__":
    iface.launch()