app.py

import os
import gradio as gr
import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer
from huggingface_hub import snapshot_download

# Student Information
My_info = "Student ID: 6319250G, Name: Aung Hlaing Tun"

# Define Hugging Face Model Repo
MODEL_REPO_ID = "ZAM-ITI-110/Distil_Bert_V3"

# Load Model & Tokenizer from Hugging Face
def load_model(repo_id):
    """Download and load the model and tokenizer."""
    cache_dir = "/home/user/app/hf_models"
    os.makedirs(cache_dir, exist_ok=True)
    download_dir = snapshot_download(repo_id, cache_dir=cache_dir, local_files_only=False)
    model = AutoModelForSequenceClassification.from_pretrained(download_dir)
    tokenizer = AutoTokenizer.from_pretrained(download_dir)
    return model, tokenizer

# Load Model
model, tokenizer = load_model(MODEL_REPO_ID)
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model.to(device)
model.eval()

# Prediction Function (Single Ticket)
def predict_team_and_email(text):
    """Predict team and email for a single ticket description."""
    if not text.strip():  # Return empty if no input
        return "", ""
    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True).to(device)
    with torch.no_grad():
        logits = model(**inputs).logits
    pred = torch.argmax(logits, dim=-1).item()
    
    label_mapping = {
        0: "Code Review Team", 1: "Functional Team", 2: "Infrastructure Team",
        3: "Performance Team", 4: "Security Team"
    }
    email_mapping = {
        0: "[email protected]", 1: "[email protected]",
        2: "[email protected]", 3: "[email protected]",
        4: "[email protected]"
    }
    
    return label_mapping.get(pred, "Unknown"), email_mapping.get(pred, "Unknown")

# Send Ticket Function (Simulation)
def send_tickets(*args):
    """Simulate sending tickets based on predictions."""
    tickets = []
    for i, (text, team, email) in enumerate(zip(args[::2], args[1::2], args[2::2]), 1):
        if text.strip() and team and email:
            tickets.append(f"Ticket {i}: '{text}' -> {team} ({email})")
    if tickets:
        return "\n".join(tickets) + "\n\nSent successfully!"
    return "No tickets to send."

# Clear Function
def clear_all():
    """Clear all inputs and outputs."""
    return [""] * 19  # 6 tickets x (input, team, email) + 1 sent_output = 19 fields

# Gradio UI Setup
with gr.Blocks(title="AI Ticket Classifier") as interface:
    gr.Markdown("📩 **Development of an AI Ticket Classifier Model Using DistilBERT**")
    gr.Markdown(f"*{My_info}*")
    gr.Markdown(
        """
        **🔍 About this App**
        - Predicts the appropriate **team** and **email** for up to 6 ticket descriptions.
        - Click 'Predict' for each ticket, then 'Send Tickets' to process.
        """
    )

    # Ticket Entry Section
    with gr.Column():
        gr.Markdown("### Enter Ticket Descriptions")
        inputs = []
        outputs = []
        buttons = []
        for i in range(6):
            with gr.Row():
                ticket_input = gr.Textbox(lines=2, placeholder=f"Ticket {i+1} description...", label=f"Ticket {i+1}")
                team_output = gr.Textbox(label="Predicted Team", interactive=False)
                email_output = gr.Textbox(label="Team Email", interactive=False)
                predict_btn = gr.Button(f"Predict {i+1}")
                inputs.append(ticket_input)
                outputs.extend([team_output, email_output])
                buttons.append(predict_btn)

    # Action Buttons
    with gr.Row():
        send_btn = gr.Button("Send Tickets")
        clear_btn = gr.Button("Clear")

    # Output for Sent Tickets
    sent_output = gr.Textbox(label="Sent Tickets", interactive=False)

    # Event Handlers for Predict Buttons
    for i, btn in enumerate(buttons):
        btn.click(
            fn=predict_team_and_email,
            inputs=inputs[i],
            outputs=[outputs[i*2], outputs[i*2 + 1]]  # Team and email for this ticket
        )

    # Send and Clear Handlers
    send_btn.click(
        fn=send_tickets,
        inputs=inputs + outputs,
        outputs=sent_output
    )
    clear_btn.click(
        fn=clear_all,
        inputs=None,
        outputs=inputs + outputs + [sent_output]
    )

# Launch the interface
interface.launch()