app.py

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import requests
import PyPDF2
import gradio as gr

# Replace with your Hugging Face API token
api_token = "YOUR_HUGGING_FACE_TOKEN"
mistral_model_id = "NousResearch/Hermes-2-Pro-Mistral-7B"  # Choose appropriate model version

tokenizer = AutoTokenizer.from_pretrained(mistral_model_id)
model = AutoModelForSeq2SeqLM.from_pretrained(mistral_model_id)


def extract_paragraphs(pdf_file):
    pattern = "IIT GATE "  # Adjust the pattern as needed
    pdf_file = open(pdf_file, "rb")
    pdf_reader = PyPDF2.PdfReader(pdf_file)
    num_pages = len(pdf_reader.pages)
    text = ""

    for i in range(num_pages):
        page = pdf_reader.pages[i]
        text += page.extract_text()

    pdf_file.close()
    words = text.split()
    paragraphs = []
    paragraph = ""
    count = 0

    for word in words:
        paragraph += word + " "
        count += 1
        if count == 200 or word == words[-1]:
            paragraphs.append(paragraph)
            count = 0
            paragraph = ""

    return paragraphs


def Generate_mcq_from_pdf(pdf_file):
    paragraphs = extract_paragraphs(pdf_file)
    for para in paragraphs:
        template = """Generate only one MCQ question based on text \
        that is delimited by triple backticks \
        with {pattern} pattern. \
        text: `{text}` \
        """
        prompt = template.format(pattern="IIT GATE", text=para)
        inputs = tokenizer(prompt, return_tensors="pt")

        headers = {"Authorization": f"Bearer {api_token}"}
        url = f"https://api-inference.huggingface.co/models/{mistral_model_id}"
        response = requests.post(url, headers=headers, json=inputs)
        response.raise_for_status()  # Raise an error if request fails

        output_ids = response.json()["generated_ids"]
        output_text = tokenizer.batch_decode(output_ids, skip_special_tokens=True)[0]

        output_file = "questions.txt"
        with open(output_file, "w") as f:
            f.write(output_text)

    return output_text, output_file


app = gr.Interface(
    fn=Generate_mcq_from_pdf,
    inputs=gr.File(type="filepath", file_types=["pdf"]),
    outputs=[gr.Textbox(label="Questions"), gr.File(label="Output File")],
)
app.launch()