PDFToText Pack
V3 deterministic PDF text-extraction capability pack. By-value connector composition (service_kind=cli, no separate ingredient): the local Poppler pdftotext CLI exposes readiness, layout-preserving extraction, default reading-order extraction, fixed first-page preview extraction, bounding-box XHTML layout extraction, and TSV word-position extraction for a source PDF. Bundles a direct text-extraction recipe plus a preview-first PDF notes workflow that summarizes the extracted text by file ref. Fixed-function and reproducible: extraction happens locally before any model sees the PDF. Write-tier but approval=never: pdftotext writes only to an engine-managed throwaway output dir and each extracted artifact is captured as result.file_ref; downstream ai-* steps consume text through Gateway-gated file refs. The source can be a data.file ref or a trusted local path. This pack exposes no password args, arbitrary flags, stdin mode, caller-controlled output path, shell wrapper, or network egress. Requires the pdftotext binary on PATH (Debian/Ubuntu: apt install poppler-utils; macOS: brew install poppler).
What this pack installs
- Extract PDF text with PDFToText Recipe
- Summarize PDF with PDFToText Recipe
- pdftotext Ingredient
Trust & control
What installing this whole pack would let it do. Recued grants these permissions at install — review them there before approving.