Comment by phenomen

I currently use https://github.com/firecrawl/anydoc in my PDF pipelines. In most cases it performs well. I'll test your lib to compare.