app.py

import gradio as gr
import google.generativeai as genai
import markdown
from docx import Document
from bs4 import BeautifulSoup
import shutil
import os
import PyPDF2  # PDF işleme için subprocess yerine Python kütüphanesi kullanın

# Setup your API key
def setup_api_key():
    google_api_key = os.getenv("GOOGLE_API_KEY")
    genai.configure(api_key=google_api_key)

def upload_file(file_path):
    print(f"Uploading file...")
    text_file = genai.upload_file(path=file_path)
    print(f"Completed upload: {text_file.uri}")
    return text_file

def to_markdown(text):
    text = text.replace('•', '  *')
    return markdown.markdown(text)

def build_model(text_file):
    generation_config = {
        "temperature": 0.2,
        "top_p": 0.95,
        "top_k": 64,
        "max_output_tokens": 8192,
        "response_mime_type": "text/plain",
    }

    model = genai.GenerativeModel(
        model_name="gemini-1.5-flash",
        generation_config=generation_config,
        system_instruction="""Answer the questions based on the uploaded file.
        If there is no related info in the file just reply 'I don't know.' """,
    )

    chat_session = model.start_chat(history=[])

    response = chat_session.send_message(["Summarize the doc in one sentence", text_file])
    return chat_session

def chat(chat_session, prompt):
    response = chat_session.send_message(prompt)
    return response.text

def generate_report(chat_session, questions):
    report_text = ""
    report_text += f"\n## QUESTIONS & ANSWERS\n"
    for question in questions:
        report_text += f"\n## {question}\n"
        answer = chat(chat_session, question)
        report_text += f"\n{answer}\n"
    return report_text

def convert_markdown_to_html(report_text):
    html_text = markdown.markdown(report_text)
    return html_text

def add_html_to_word(html_text, doc):
    soup = BeautifulSoup(html_text, 'html.parser')
    for element in soup:
        if element.name == 'h1':
            doc.add_heading(element.get_text(), level=1)
        elif element.name == 'h2':
            doc.add_heading(element.get_text(), level=2)
        elif element.name == 'h3':
            doc.add_heading(element.get_text(), level=3)
        elif element.name == 'h4':
            doc.add_heading(element.get_text(), level=4)
        elif element.name == 'h5':
            doc.add_heading(element.get_text(), level=5)
        elif element.name == 'h6':
            doc.add_heading(element.get_text(), level=6)
        elif element.name == 'p':
            doc.add_paragraph(element.get_text())
        elif element.name == 'ul':
            for li in element.find_all('li'):
                doc.add_paragraph(li.get_text(), style='List Bullet')
        elif element.name == 'ol':
            for li in element.find_all('li'):
                doc.add_paragraph(li.get_text(), style='List Number')
        elif element.name:
            doc.add_paragraph(element.get_text())  # For any other tags

def extract_text_from_pdf(pdf_path):
    """PDF dosyasından metin çıkarmak için PyPDF2 kullanır"""
    text = ""
    with open(pdf_path, 'rb') as file:
        pdf_reader = PyPDF2.PdfReader(file)
        for page_num in range(len(pdf_reader.pages)):
            text += pdf_reader.pages[page_num].extract_text() + "\n"
    return text

def process_pdf(pdf_file, user_questions):
    file_name = pdf_file.split('/')[-1]
    saved_file_path = f"/tmp/{file_name}"
    shutil.copyfile(pdf_file, saved_file_path)
    
    # PDF'den doğrudan metin çıkar
    text = extract_text_from_pdf(saved_file_path)
    
    # Çıkarılan metni bir dosyaya yaz
    with open("/tmp/text_file.txt", "w", encoding="utf-8") as f:
        f.write(text)
    
    text_file = upload_file("/tmp/text_file.txt")
    chat_session = build_model(text_file)

    questions = user_questions.strip().split('\n')
    report_text = generate_report(chat_session, questions)

    doc = Document()
    html_text = convert_markdown_to_html(report_text)
    add_html_to_word(html_text, doc)

    doc_name = file_name.replace(".pdf", ".docx")
    doc_name = "Report_" + doc_name
    doc.save(f"/tmp/{doc_name}")

    return html_text, f"/tmp/{doc_name}"

questions = [
    "Makalenin yazarları kimlerdir?",
    "Hangi modeller kullanılmıştır?",
    "Kaç referans vardır?",
    "Hangi yılda yayınlanmıştır?"
]

questions_str = "\n".join(questions)

iface = gr.Interface(
    fn=process_pdf,
    inputs=[
        gr.File(label="Upload PDF", type="filepath"),
        gr.TextArea(label="Enter Questions", placeholder="Type your questions here, one per line.", value=questions_str)
    ],
    outputs=[
        gr.HTML(label="HTML Formatted Report"),
        gr.File(label="DOCX File Output", type="binary")
    ],
    title="Pdflerinizden kısa rapor oluşturma arac @YED",
    description="Sorularınızı sormak ve cevap almak için PDF'inizi yükleyin."
)

setup_api_key()
# Hugging Face Spaces için önerilen launch konfigürasyonu
iface.launch(server_name="0.0.0.0", server_port=7860)