# app.py
# Telecom Customer Complaint Classification and Routing App
# Using Hugging Face Transformers + Gradio

from transformers import pipeline
import gradio as gr

# Load zero-shot classification model
classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")

# Complaint categories
categories = [
    "Network Issue",
    "Billing Issue",
    "SIM Issue",
    "Recharge Issue",
    "Device Issue"
]

# Mapping categories to routing teams
routing_teams = {
    "Network Issue": "Network Operations",
    "Billing Issue": "Billing Support",
    "SIM Issue": "SIM Support",
    "Recharge Issue": "Payments Team",
    "Device Issue": "Technical Support"
}

# Function to classify complaint
def classify_complaint(complaint_text):
    if not complaint_text.strip():
        return "No input provided", "0.0%", "N/A"
    
    result = classifier(complaint_text, candidate_labels=categories)
    top_category = result['labels'][0]
    confidence_score = result['scores'][0]
    suggested_team = routing_teams.get(top_category, "General Support")
    confidence_percent = f"{confidence_score*100:.2f}%"
    
    return top_category, confidence_percent, suggested_team

# Build Gradio UI
with gr.Blocks() as demo:
    gr.Markdown("## 📞 Telecom Customer Complaint Classification and Routing")
    gr.Markdown("Enter a customer complaint below to get category, confidence, and routing team.")
    
    with gr.Row():
        complaint_input = gr.Textbox(label="Customer Complaint", placeholder="Type your complaint here...", lines=4)
        submit_btn = gr.Button("Submit")
    
    with gr.Row():
        category_output = gr.Textbox(label="Predicted Category")
        confidence_output = gr.Textbox(label="Confidence Score")
        team_output = gr.Textbox(label="Suggested Routing Team")
    
    submit_btn.click(
        classify_complaint,
        inputs=complaint_input,
        outputs=[category_output, confidence_output, team_output]
    )

# Launch the app
if __name__ == "__main__":
    demo.launch()