app.py

import gradio as gr
from transformers import pipeline
from pptx import Presentation
import re
import json

# Load the classification and summarization pipelines
classifier = pipeline("text-classification", model="Ahmed235/roberta_classification", tokenizer="Ahmed235/roberta_classification")
summarizer = pipeline("summarization", model="Falconsai/text_summarization")

# Cache for model weights
classification_model_loaded = False
summarization_model_loaded = False

def load_models():
    global classifier, summarizer, classification_model_loaded, summarization_model_loaded
    if not classification_model_loaded:
        classifier = pipeline("text-classification", model="Ahmed235/roberta_classification", tokenizer="Ahmed235/roberta_classification")
        classification_model_loaded = True
    if not summarization_model_loaded:
        summarizer = pipeline("summarization", model="Falconsai/text_summarization")
        summarization_model_loaded = True

# Extract text from PowerPoint
def extract_text_from_pptx(file_path):
    try:
        presentation = Presentation(file_path)
        text = []
        for slide in presentation.slides:
            for shape in slide.shapes:
                if hasattr(shape, "text"):
                    text.append(shape.text)
        return "\n".join(text)
    except Exception as e:
        print(f"Error extracting text from PowerPoint: {e}")
        return ""

# Limit text length
def limit_text_length(text, max_length=512):
    return text[:max_length]

# Predict content from PowerPoint
def predict_pptx_content(file_path):
    try:
        load_models()  # Load models if not loaded already
        extracted_text = extract_text_from_pptx(file_path)
        cleaned_text = re.sub(r'\s+', ' ', extracted_text)
        limited_text = limit_text_length(cleaned_text)
        result = classifier(limited_text)
        predicted_label = result[0]['label']
        predicted_probability = result[0]['score']
        summary = summarizer(cleaned_text, max_length=1000, min_length=30, do_sample=False)[0]['summary_text']
        output = {
            "predicted_label": predicted_label,
            "evaluation": predicted_probability,
            "summary": summary
        }
        return json.dumps(output, indent=3)
    except Exception as e:
        print(f"Error predicting content from PowerPoint: {e}")
        return {"error": str(e)}

# Gradio interface
iface = gr.Interface(
    fn=predict_pptx_content,
    inputs=gr.File(type="filepath", label="Upload PowerPoint (.pptx) file"),
    outputs=gr.Textbox("output"),
    live=False,
    title="<h1 style='color: lightgreen; text-align: center;'>HackTalk Analyzer</h1>",
)

# Deploy the Gradio interface
iface.launch(share=True)