OSS
Firecrawl's pdf-inspector classifies scanned versus text PDFs so pipelines can skip OCR entirely
firecrawl/pdf-inspector gained 589 stars today to reach 18,260, second on the Rust board. Rather than extracting text, it decides whether a PDF needs OCR at all, which is the routing decision that dominates cost in document pipelines. Its August 17 v1.15.0 was titled 'Selective OCR for scanned and mixed PDFs' and v1.14.2 four days earlier was 'Hardened parsing for pathological PDFs,' so the recent work is about surviving adversarial input rather than adding features.
↳ Follow the thread