import spaces import gradio as gr from marker.markdown_extractor import MarkdownExtractorConfig, MarkdownExtractor from pdf.pdf_extractor import PDFExtractorConfig, PDFExtractor from gemini.gemini_extractor import GeminiExtractorConfig, GeminiExtractor from oai.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 with gr.Blocks(title="PDF data extraction with Marker & Indexify") as marker_demo: gr.HTML("

PDF data extraction with Marker & Indexify

") gr.HTML("

Indexify is a scalable realtime and continuous indexing and structured extraction engine for unstructured data to build generative AI applications

") gr.HTML("

If you like this demo, please ⭐ Star us on GitHub!

") gr.HTML("

Here's an example notebook that demonstrates how to build a continous extraction pipleine with Indexify

") with gr.Row(): with gr.Column(): gr.HTML( "

Step 1: Upload a PDF file from local storage.

" "

Use this demo for single PDF file only. " "You can extract from PDF files continuously and try various other extractors locally with " "Indexify.

" ) pdf_file = gr.File(type="filepath") with gr.Column(): gr.HTML("

Step 2: Run the extractor.

") 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( "

" "Developed with 🫶 by Indexify | " "a Tensorlake product" "

" ) go_button.click( fn=use_marker, inputs = [pdf_file], outputs = [model_output_text_box] ) @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 with gr.Blocks(title="PDF data extraction with PDF Extractor & Indexify") as pdf_demo: gr.HTML("

PDF data extraction with PDF Extractor & Indexify

") gr.HTML("

Indexify is a scalable realtime and continuous indexing and structured extraction engine for unstructured data to build generative AI applications

") gr.HTML("

If you like this demo, please ⭐ Star us on GitHub!

") gr.HTML("

Here's an example notebook that demonstrates how to build a continous extraction pipleine with Indexify

") with gr.Row(): with gr.Column(): gr.HTML( "

Step 1: Upload a PDF file from local storage.

" "

Use this demo for single PDF file only. " "You can extract from PDF files continuously and try various other extractors locally with " "Indexify.

" ) pdf_file = gr.File(type="filepath") with gr.Column(): gr.HTML("

Step 2: Run the extractor.

") 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( "

" "Developed with 🫶 by Indexify | " "a Tensorlake product" "

" ) go_button.click( fn=use_pdf_extractor, inputs = [pdf_file], outputs = [model_output_text_box] ) 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 with gr.Blocks(title="PDF data extraction with Gemini & Indexify") as gemini_demo: gr.HTML("

PDF data extraction with Gemini & Indexify

") gr.HTML("

Indexify is a scalable realtime and continuous indexing and structured extraction engine for unstructured data to build generative AI applications

") gr.HTML("

If you like this demo, please ⭐ Star us on GitHub!

") gr.HTML("

Here's an example notebook that demonstrates how to build a continous extraction pipleine with Indexify

") with gr.Row(): with gr.Column(): gr.HTML( "

Step 1: Upload a PDF file from local storage.

" "

Use this demo for single PDF file only. " "You can extract from PDF files continuously and try various other extractors locally with " "Indexify.

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

Step 2: Enter your API key.

") key = gr.Textbox( info="Please enter your GEMINI_API_KEY", label="Key:" ) with gr.Column(): gr.HTML("

Step 3: Run the extractor.

") 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( "

" "Developed with 🫶 by Indexify | " "a Tensorlake product" "

" ) go_button.click( fn=use_gemini, inputs = [pdf_file, key], outputs = [model_output_text_box] ) 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 OpenAI & Indexify") as openai_demo: gr.HTML("

PDF data extraction with OpenAI & Indexify

") gr.HTML("

Indexify is a scalable realtime and continuous indexing and structured extraction engine for unstructured data to build generative AI applications

") gr.HTML("

If you like this demo, please ⭐ Star us on GitHub!

") gr.HTML("

Here's an example notebook that demonstrates how to build a continous extraction pipleine with Indexify

") with gr.Row(): with gr.Column(): gr.HTML( "

Step 1: Upload a PDF file from local storage.

" "

Use this demo for single PDF file only. " "You can extract from PDF files continuously and try various other extractors locally with " "Indexify.

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

Step 2: Enter your API key.

") key = gr.Textbox( info="Please enter your OPENAI_API_KEY", label="Key:" ) with gr.Column(): gr.HTML("

Step 3: Run the extractor.

") 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( "

" "Developed with 🫶 by Indexify | " "a Tensorlake product" "

" ) go_button.click( fn=use_openai, inputs = [pdf_file, key], 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()