Forms

The form catalogue and the pipeline that produces it: a blank arrives by download, scan or email, its permission bits come off, its fields are read if it already has them or drawn if it does not, and the answers go into a separate file so the blank stays reusable.

Loading form catalog into IndexedDB…

0Editions

0Publishers

0Mapped fields

0Instances

0Extractions

0Linked files

0Property links

The three kinds a PDF can be

This is the decision the whole pipeline turns on. Getting it wrong is not a degraded result, drawing over a form that already has widgets puts two on every blank, and a reader honours whichever it reaches first, so half of what is typed is discarded on save.

Already fillable Read the widgets. Do NOT draw. Drawing over an existing form puts two widgets on every blank. A reader honours whichever it reaches first, so half of what is typed is discarded on save -- by a file that looks correct in the reader that made it.

Flat vector + text Rebuild geometry, then draw. On a C.A.R. form a printed checkbox is four independent stroked segments, not a rectangle, so a detector looking for small squares finds exactly zero. Boxes are reassembled by pairing horizontal edges and confirming with verticals.

Image only OCR or VLM first. Detection refuses. Geometry over a bare raster finds scanner artefacts, not fields -- and an empty field map is indistinguishable downstream from a form that genuinely has none.

Pipeline

1 Acquire _TEMPLATE.pdf download, scan drop, or an email attachment promoted into the naming
2 Unprotect _TEMPLATE-SECURITY.json strip owner-password permission bits -- a C.A.R. template forbids modify_form
3 Transcribe _TEMPLATE.md verbatim text, so the blank is searchable before it is fillable
4 Detect _FORM-DATA.json read existing widgets, rebuild flat geometry, or refuse a bare scan
5 Draw _FORM.pdf write widgets with pikepdf, read them back with pypdf
6 Verify _FORM-PLACEMENT.pdf the round trip proves the PDF matches the map, not that the map is on the blanks
7 Answer _FILL-DATA.json every value carries a source; nothing is inferred by default
8 Fill _FILL.pdf a SEPARATE file -- _FORM.pdf stays a reusable blank
9 Register mc_form.db form / form_field / form_instance rows and the drive links

No corpus sweep yet, the board says how to run one.

Editions with a field map

A mapped edition is a reusable asset: the map belongs to the edition, so every future transaction on the same revision gets a fillable form for free. The answers never travel with it.


No rows

No rows

By publisher


No rows

No rows

By media

How the blank arrives decides the route detection takes.


No rows

No rows

Recently touched instances


No rows

No rows