app.py

import spaces
import gradio as gr
from marker.markdown_extractor import MarkdownExtractorConfig, MarkdownExtractor
from pdf-extractor.pdf_extractor import PDFExtractorConfig, PDFExtractor
from gemini.gemini_extractor import GeminiExtractorConfig, GeminiExtractor
from openai.oai_extractor import OAIExtractorConfig, OAIExtractor
from indexify_extractor_sdk import Content

markdown_extractor = MarkdownExtractor()
pdf_extractor = PDFExtractor()
gemini_extractor = GeminiExtractor()
oai_extractor = OAIExtractor()

@spaces.GPU
def use_marker(pdf_filepath):
	if pdf_filepath is None:
		raise gr.Error("Please provide some input PDF: upload an PDF file")
	
	with open(pdf_filepath, "rb") as f:
		pdf_data = f.read()
		
	content = Content(content_type="application/pdf", data=pdf_data)
	config = MarkdownExtractorConfig(batch_multiplier=2)
	
	result = markdown_extractor.extract(content, config)
	return result

@spaces.GPU
def use_pdf_extractor(pdf_filepath):
	if pdf_filepath is None:
		raise gr.Error("Please provide some input PDF: upload an PDF file")
	
	with open(pdf_filepath, "rb") as f:
		pdf_data = f.read()
		
	content = Content(content_type="application/pdf", data=pdf_data)
    config = PDFExtractorConfig(output_types=["text", "table"])
	
	result = pdf_extractor.extract(content, config)
	return result

@spaces.GPU
def use_gemini(pdf_filepath, key):
	if pdf_filepath is None:
		raise gr.Error("Please provide some input PDF: upload an PDF file")
	
	with open(pdf_filepath, "rb") as f:
		pdf_data = f.read()
		
	content = Content(content_type="application/pdf", data=pdf_data)
	config = GeminiExtractorConfig(prompt="Extract all text from the document.", model_name="gemini-1.5-flash", key=key)
	
	result = gemini_extractor.extract(content, config)
	return result

@spaces.GPU
def use_openai(pdf_filepath, key):
	if pdf_filepath is None:
		raise gr.Error("Please provide some input PDF: upload an PDF file")
	
	with open(pdf_filepath, "rb") as f:
		pdf_data = f.read()
		
	content = Content(content_type="application/pdf", data=pdf_data)
	config = OAIExtractorConfig(prompt="Extract all text from the document.", model_name="gpt-4o", key=key)
	
	result = oai_extractor.extract(content, config)
	return result

with gr.Blocks(title="PDF data extraction with Marker & Indexify") as marker_demo:
	gr.HTML("<h1 style='text-align: center'>PDF data extraction with Marker & <a href='https://getindexify.ai/'>Indexify</a></h1>")
	gr.HTML("<p style='text-align: center'>Indexify is a scalable realtime and continuous indexing and structured extraction engine for unstructured data to build generative AI applications</p>")
	gr.HTML("<h3 style='text-align: center'>If you like this demo, please ⭐ Star us on <a href='https://github.com/tensorlakeai/indexify' target='_blank'>GitHub</a>!</h3>")
	gr.HTML("<h4 style='text-align: center'>Here's an example notebook that demonstrates how to build a continous <a href='https://github.com/tensorlakeai/indexify/blob/main/docs/docs/examples/SEC_10_K_docs.ipynb' target='_blank'>extraction pipleine</a> with Indexify</h4>")

	with gr.Row():
		with gr.Column():
			gr.HTML(
				"<p><b>Step 1:</b> Upload a PDF file from local storage.</p>"

				"<p style='color: #A0A0A0;'>Use this demo for single PDF file only. "
				"You can extract from PDF files continuously and try various other extractors locally with "
				"<a href='https://getindexify.ai/'>Indexify</a>.</p>"
			)

			pdf_file = gr.File(type="filepath")

		with gr.Column():
			gr.HTML("<p><b>Step 2:</b> Run the extractor.</p>")

			go_button = gr.Button(
				value="Run extractor",
				variant="primary",
			)

			model_output_text_box = gr.Textbox(
				label="Extractor Output",
				elem_id="model_output_text_box",
			)

	with gr.Row():

		gr.HTML(
			"<p style='text-align: center'>"
				"Developed with 🫶 by <a href='https://getindexify.ai/' target='_blank'>Indexify</a> | "
				"a <a href='https://www.tensorlake.ai/' target='_blank'>Tensorlake</a> product"
			"</p>"
		)

	go_button.click(
		fn=use_marker, 
		inputs = [pdf_file],
		outputs = [model_output_text_box]
	)

demo = gr.TabbedInterface([marker_demo, pdf_demo, gemini_demo, openai_demo], ["Marker Extractor", "PDF Extractor", "Gemini Extractor", "OpenAI Extractor"], theme=gr.themes.Soft())

demo.queue()
demo.launch()