Skip to content
Documentation

What it reads.

Thirteen formats, what each one gives up, and what happens to a file it cannot read confidently.

Quotes as they arrive, each read from its own file

File formats

Thirteen formats in all, counting each supported DWG release separately. OCR runs on every one that needs it.

FormatRead asCaveat
DWGTrue vector geometry and layersR2004 and later. Older releases are named and refused, never read as an empty drawing.
DXFTrue vector geometry and layersThe most portable exchange format; preferred if DWG export is awkward.
IFCModel geometry and element typesUsed for the 3D view and element-linked findings.
Vector PDFGeometry, text and layers where presentLayer filtering needs at least two layers; fewer shows a clean empty state.
Scanned PDFOCR over a rasterised sheetAt low resolution, recognition misreads exactly the characters that matter — 5/8 as 5/B, 90 as 9O. Low-confidence readings are marked “verify” and kept out of the total.
PNG / JPG / TIFFOCR over the imageSame caveat as scanned PDF. Scale has to be set or found on the sheet.
XLSX / CSVSchedules and rate tablesColumn mapping is confirmed on import.

Why the DWG matters

A PDF export of a drawing has already thrown away the layer structure and, often, the true geometry. Give the program the DWG the architect sent and it measures the drawn element instead of inferring it from a picture of one. Everything still works on a PDF — it just works better on the original.

Multi-sheet sets

Sheets that are parts of one plan — same level, same discipline, overlapping rather than coinciding — are joined on their grid bubbles and the overlap is de-duplicated. A sheet the program cannot confidently place is named beside the total with the reason, because a silently dropped sheet is a silently missing count.