Using VLLM (like GPT-4o) to parse PDF into markdown.

Our approach is very simple (only 293 lines of code), but can almost perfectly parse typography, math formulas, tables, pictures, charts, etc.

Average cost per page: $0.013

This package use GeneralAgent lib to interact with OpenAI API.

1. Use the PyMuPDF library to parse the PDF to find all non-text areas and mark them, for example:

2. Use a large visual model (such as GPT-4o) to parse and get a markdown file.

See examples/attention_is_all_you_need/output.md for PDF examples/attention_is_all_you_need.pdf.

pip install gptpdf

from gptpdf import parse_pdf
api_key = 'Your OpenAI API Key'
content, image_paths = parse_pdf(pdf_path, api_key=api_key)
print(content)

See more in test/test.py

parse_pdf(pdf_path, output_dir='./', api_key=None, base_url=None, model='gpt-4o', verbose=False)

parse pdf file to markdown file, and return markdown content and all image paths.

• pdf_path: pdf file path

• output_dir: output directory. store all images and markdown file

• api_key: OpenAI API Key (optional). If not provided, Use OPENAI_API_KEY environment variable.

• base_url: OpenAI Base URL. (optional). If not provided, Use OPENAI_BASE_URL environment variable.

• model: OpenAI Vison LLM Model, default is 'gpt-4o'. You also can use qwen-vl-max

• verbose: verbose mode

• gpt_worker: gpt parse worker number. default is 1. If your machine performance is good, you can increase it appropriately to improve parsing speed.