VeXtraBOM is a self-hosted engine, in production use, that turns engineering drawing packages into structured bills of materials. It is built by EdgeXene LLC, Indiana, United States, and runs entirely on the computer it is installed on -- your drawings and workbooks never leave that machine, and EdgeXene receives no copy of either.
Every deployment is tailored to the customer's own drawings. Drawing families differ in ways that matter to extraction: how a parts list is framed and split across columns, where the quantity charts sit and how they are keyed, which codes stand for a variable and which for a part. We work from your actual drawings and tune the engine to read them as they are drawn, rather than asking your team to change how it draws.
Uploaded drawings pass through a text-recognition engine, then a geometric reconstruction step: table regions are located by their header anchors, cropped and skew-corrected, re-read, and clustered back into rows and columns. CAD-exported vector PDFs skip recognition entirely -- their exact embedded text and coordinates are used directly, with an automatic fallback to recognition for scanned pages.
Before uploading, the console shows a map of how each sheet layout is read: a parts-list sheet (one table, usually top-right), a config/variant sheet (BOM chart top-right, calculation tables under the drawing), or a CAD PDF. Auto-detect -- the default -- makes no layout assumptions and finds tables wherever they sit. On variant sheets, rows marked with highlighter are flagged as the engineer's selected configuration -- read from the PDF's own highlight annotations on CAD PDFs; on photos, marked rows are flagged from the ink color in the console preview.
Every result is a draft. The console flags low-confidence reads, every cell is editable, rows can be removed, and nothing is exported until a person has had the chance to review it. A wrong extraction costs a correction during review, not a bad BOM downstream.
Working today: parts-list extraction from photos and scans, vector-PDF extraction with scanned-page fallback, BOM data QC, multi-drawing super-BOM roll-up (indented assembly tree plus a consolidated purchase list; variable quantities carried unresolved), and workbook generation -- a drawing package becomes a draft workbook (parts sheets, quantity charts, drawing list, legend) with amber needs-a-name and blue needs-a-calculation flags; nothing is guessed. Highlighter-marked chart selections are read (PDF highlight annotations on CAD PDFs, stroke-color math on flattened or scanned PDF pages) and resolved into a SELECTED column beside the printed quantity, with a CONFIG SELECTIONS sheet recording every choice; ambiguous or unresolvable selections are flagged, never guessed. Roll-up and generation accept an optional catalog prefix (for example W-) applied to part numbers that start with a digit -- a configured rule the generated workbook's legend records, not a guess. The overview page labels every capability by its real state.
Early access, questions, or drawing sets you would like the pipeline to handle:
contact@edgexene.io| TitleVeXtraBOM -- about | Drawn byEdgeXene LLC | Rev1.0 | StatusRELEASED | PrivacyPOLICY | ©2026 |