947-Page Test: pdf-inspector Classifies in 0.65s but Extraction Would Take 36 Hours
Ubunta · x · 2026-09-04
The author benchmarked pdf-inspector against a pdfplumber/pdfminer + PyMuPDF pipeline on a 947-page clinical listing:
- classifypdf shines: classified all 947 pages in 0.65s, correctly detecting a text-based PDF — great for OCR routing and document inspection.
- Markdown extraction is very slow: 275s for just two pages, projecting to roughly 36 hours for 947 pages.
- No cell-level provenance: Markdown loses page/table/row/column and bounding-box info, and merges adjacent cells — unusable for evidence-grade citations in regulated documents.
- Existing pipeline: after optimization, full extraction dropped from 55 minutes to 6.2 minutes while preserving cell geometry and row context.
Verdict: solid for PDF classification and OCR routing, but the custom pipeline stays for evidence-grade table extraction.
More from coding & agent
- Claude Code v2.1.251 now saves effort levels per model via /effort or settings. — EricBuess · 2026-09-05
- Subagents Log Where Docs Fail: Docker Stuck Points and Install Gaps — lucasmeijer · 2026-09-05
- Dev Uses Subagents as Fake Users to Test His Coding Agent's Install Flow — lucasmeijer · 2026-09-05
- ChatGPT Android app hides the Codex menu but works via manually paired remote sessions — ezyang · 2026-09-05
- 9 lessons from running a write-capable MCP server against 7 real APIs — Ecstatic-Hurry-635 · 2026-09-05
- Reversing Super Mario Kart without source: decompilation is verifiable, and AI thrives at it — gandamu_ml · 2026-09-05