app.py

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

def extract_text_markdown(doc):
    markdown_output = ""
    image_counter = 1

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

        for b in blocks:
            y = b["bbox"][1]
            if b["type"] == 0:  # Texto
                for line in b["lines"]:
                    line_y = line["bbox"][1]
                    spans = line["spans"]

                    if len(spans) > 1:
                        line_text = " | ".join(span["text"].strip() for span in spans)
                    else:
                        line_text = " ".join(span["text"].strip() for span in spans)

                    if line_text:
                        elements.append((line_y, line_text))
            elif b["type"] == 1:  # Imagen
                elements.append((y, f"[imagen_{image_counter}]()"))
                image_counter += 1

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

        previous_y = None
        for y, content in elements:
            if previous_y is not None and abs(y - previous_y) > 10:
                markdown_output += "\n"
            markdown_output += content + "\n"
            previous_y = y

        markdown_output += "\n---\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])

    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 = {}  # Añade metadatos si quieres
    return markdown, metadata

# Gradio Interface
with gr.Blocks(title="Extractor PDF a Markdown") as demo:
    gr.Markdown("### PDF → Markdown con imágenes como enlaces y botón de copiar")

    pdf_input = gr.File(label="Sube tu PDF", type="filepath")
    markdown_output = gr.Textbox(label="Markdown generado", lines=25, elem_id="md_output")
    metadata_output = gr.JSON(label="Metadata")

    with gr.Row():
        convert_btn = gr.Button("Convertir PDF")
        copy_btn = gr.HTML("""
            <button onclick="navigator.clipboard.writeText(document.getElementById('md_output').value)">📋 Copiar Markdown</button>
        """)

    convert_btn.click(fn=convert, inputs=pdf_input, outputs=[markdown_output, metadata_output])

demo.launch()