Firecrawl Open-Source PDF Classifier: 200 Docs in 0.47s!
Firecrawl pdf-inspector: open-source PDF classifier & extractor. 200 docs in 0.47s. 14K stars. Text, tables, Markdown.
Just discovered that Firecrawl has also open-sourced a PDF intelligent classification and text extraction library: pdf-inspector, which has already gained over 14,000 stars.
Simply put, pdf-inspector is the “traffic cop” of the PDF world—it first figures out what kind of vehicle is coming, then directs it to the corresponding lane.
Firecrawl has turned its attention to PDF processing scenarios with a very pragmatic approach: instead of blindly throwing everything into OCR, first determine what type of PDF it is, then decide how to handle it.
Specifically, pdf-inspector can do four things:
Intelligent classification: Determine whether the PDF is text-based, scanned, image-based, or mixed
Position-aware extraction: Not only extract text, but also retain coordinate and font information
Markdown conversion: Output clean Markdown that properly handles headings, lists, tables, and more
On-demand routing: Precisely identify which specific pages need OCR, rather than processing everything
The most impressive part is the performance data—classification takes only 10-50ms, text-based PDFs are processed locally in under 200ms, and a benchmark of 200 PDFs took just 0.47 seconds.
That’s right—less than half a second for two hundred files.
