app.py

import spaces
import gradio as gr
import fitz  # PyMuPDF
import os
import tempfile
import ocrmypdf

def extract_text_markdown(doc):
    markdown_output = ""
    image_dir = "extracted_images"
    os.makedirs(image_dir, exist_ok=True)
    image_counter = 0

    for page in doc:
        blocks = page.get_text("dict")["blocks"]
        elements = []

        # Extraer imágenes y guardar para asignar link
        image_list = page.get_images(full=True)
        xref_to_placeholder = {}

        for img in image_list:
            xref = img[0]
            pix = fitz.Pixmap(doc, xref)
            img_path = os.path.join(image_dir, f"imagen_{image_counter}.png")

            if pix.n > 4:
                pix = fitz.Pixmap(fitz.csRGB, pix)
            pix.save(img_path)
            pix = None

            xref_to_placeholder[xref] = f"![imagen]()"
            image_counter += 1

        for b in blocks:
            y = b["bbox"][1]
            if b["type"] == 0:  # Texto
                paragraph = ""
                for line in b["lines"]:
                    line_text = " ".join([span["text"].strip() for span in line["spans"]])
                    paragraph += line_text + " "
                paragraph = paragraph.strip()
                if paragraph:
                    elements.append((y, paragraph))
            elif b["type"] == 1:  # Imagen
                xref = b.get("image")
                elements.append((y, "![imagen]()"))

        elements.sort(key=lambda x: x[0])

        for _, content in elements:
            markdown_output += content + "\n\n"

    return markdown_output.strip()

@spaces.GPU
def convert(pdf_file):
    original_doc = fitz.open(pdf_file)
    plain_text = "\n".join([page.get_text() for page in original_doc])

    # Si es imagen escaneada sin texto, aplicamos OCR
    if len(plain_text.strip()) < 100:
        ocr_temp_path = tempfile.NamedTemporaryFile(suffix=".pdf", delete=False).name
        ocrmypdf.ocr(pdf_file, ocr_temp_path, force_ocr=True)
        doc = fitz.open(ocr_temp_path)
    else:
        doc = original_doc

    markdown = extract_text_markdown(doc)
    metadata = {}  # Puedes agregar metadatos si quieres
    return markdown, metadata

gr.Interface(
    fn=convert,
    inputs=[gr.File(label="Upload PDF", type="filepath")],
    outputs=[gr.Text(label="Markdown crudo"), gr.JSON(label="Metadata")],
).launch()