Create app.py

app.py

@@ -0,0 +1,47 @@
+import os
+import tempfile
+
+import gradio as gr
+from magic_pdf.data.data_reader_writer import (
+    FileBasedDataReader,
+    FileBasedDataWriter
+)
+from magic_pdf.data.dataset import PymuDocDataset
+from magic_pdf.model.doc_analyze_by_custom_model import doc_analyze
+from magic_pdf.config.enums import SupportedPdfParseMethod
+
+def pdf_to_markdown(uploaded_pdf):
+    # Save the uploaded file to a temp location
+    pdf_path = uploaded_pdf.name
+    # Prepare output dirs
+    img_dir = os.path.join(tempfile.gettempdir(), "magic_pdf_images")
+    os.makedirs(img_dir, exist_ok=True)
+
+    # Read bytes
+    reader = FileBasedDataReader("")
+    pdf_bytes = reader.read(pdf_path)
+
+    # Build dataset & run inference
+    ds = PymuDocDataset(pdf_bytes)
+    if ds.classify() == SupportedPdfParseMethod.OCR:
+        infer = ds.apply(doc_analyze, ocr=True)
+        pipe = infer.pipe_ocr_mode(FileBasedDataWriter(img_dir))
+    else:
+        infer = ds.apply(doc_analyze, ocr=False)
+        pipe = infer.pipe_txt_mode(FileBasedDataWriter(img_dir))
+
+    # Get Markdown string
+    md = pipe.get_markdown(os.path.basename(img_dir))
+    return md
+
+# Gradio UI
+iface = gr.Interface(
+    fn=pdf_to_markdown,
+    inputs=gr.File(file_types=[".pdf"], label="Upload PDF"),
+    outputs=gr.Textbox(lines=20, label="Markdown"),
+    title="PDF → Markdown Converter",
+    description="Powered by Magic-PDF"
+)
+
+if __name__ == "__main__":
+    iface.launch()