CLI Reference

The CLI is intentionally thin. Command behavior should stay close to the Python API and should emit files and summaries that can be consumed by automation.

Image Digitization

Calibrated image-crop digitization:

figrecover digitize-image --help

PDF Rendering

Render selected one-based PDF pages to image files:

figrecover pdf render report.pdf --out-dir tmp/pages --pages 1,3-5 --dpi 200

The command prints a JSON summary with rendered page records, including source PDF path, page number, DPI, image dimensions, output path, and renderer metadata.

Figure Manifests

List a JSONL figure candidate manifest:

figrecover figures list tmp/manifests/figures.jsonl --json

Crop candidates from source page images:

figrecover figures crop tmp/manifests/figures.jsonl \
   --out-dir tmp/crops \
   --out-manifest tmp/manifests/cropped-figures.jsonl

Manifest entries preserve candidate provenance. Source PDFs, rendered pages, crops, and private-document manifests should remain under ignored local output directories unless explicitly sanitized for tracking.

Review

Generate visual and tabular review artifacts from digitization JSON results:

figrecover review bundle tmp/results/fig-001.json --out-dir tmp/review

Summarize review status counts and low-confidence entries:

figrecover review summarize tmp/review/review.jsonl --json

Export accepted or manually corrected tables only:

figrecover review export-accepted tmp/review/review.jsonl \
   --out-dir tmp/accepted-tables

Corpus

Initialize a corpus output root and config:

figrecover corpus init tmp/corpus \
   --pdf report.pdf \
   --config-path tmp/corpus-config.json \
   --dpi 200

Run a configured corpus:

figrecover corpus run tmp/corpus-config.json

Summarize a run manifest:

figrecover corpus summarize tmp/corpus/manifests/run-manifest.json

Export accepted review tables:

figrecover corpus export tmp/review/review.jsonl \
   --out-dir tmp/corpus/tables/accepted

Integration Exports

FEMIC, FHOPS, and generic modelling exports are currently Python API workflows rather than CLI commands. See Integration Exports for the public export contract.

Limitations

The CLI does not infer arbitrary chart calibration from a PDF. It exposes deterministic preparation, extraction, review, and corpus commands. Keep private PDFs, crops, overlays, manifests, and recovered tables under ignored output directories such as tmp/ unless explicitly sanitized.