quiddity

Deterministic feature recognition for B-Rep geometry in STEP files: reads a STEP model and finds its manufacturing features (holes, pockets, slots, bosses, steps, fillets, chamfers, patterns, springs and turned profiles) with their measured sizes and positions, reads and writes AP242 PMI, and matches features between two revisions of a part; pure Rust with Python bindings and a command line, working from the B-rep alone with no CAD system or OpenCascade.

Open projectLibrary / CLIOpen sourceActiveMechanical

What it does and how it connects

quiddity reads a STEP model and finds its manufacturing features with their measured sizes and positions: holes (through, blind, counterbored, countersunk, spotfaced, plus modelled threads and hole patterns), pockets, slots, channels and grooves, steps, bosses, pads and ribs, fillets, blends and chamfers, turned profiles and helical springs. It works from the B-rep geometry alone, with no CAD system, feature tree or OpenCascade: pure Rust, deterministic (the same input gives the same output), on Windows, macOS, Linux and in the browser via WebAssembly. Every output record names its feature family, the STEP faces it is made of, its measured parameters and whether its evidence is verified or unproved, and a face accounting says what every face of the part is: part of a feature, part of the body, or refused or unexplained, so nothing is silently dropped.

It also reads and writes semantic AP242 PMI (dimensions, tolerances, datums, notes and their callouts): a write keeps everything else in the file and checks the result by reading it back before replacing the destination. Its correspond command matches features between two revisions of a part and reports each as carried, adapted (with what changed), orphaned or ambiguous, matching on revision-independent fingerprints rather than face numbers, so it still works when the part is moved or remodelled.

Measured results published in the repo (version 0.5.1, on the public MFInstSeg test split): all 9,375 models recognised with no failures, 77% recall across all labelled features (96% for holes) with a 14% false-positive share; volume, area and centre of mass agree with the values stored in NIST STEP files on 103 of 106 checks within 0.1% and with OpenCascade on 9,110 of 9,120 checks; a median 13 ms to read a part and 73 ms to recognise it. The packages are the quiddity Rust crate (recognition, fingerprints, correspondence, the PMI JSON and the command line), the haecceity crate (the geometry underneath: STEP reading, the B-rep, measurement and semantic PMI, re-exported as quiddity::kernel) and the quiddity Python wheel over both. Draftwright, the STEP-to-drawing generator in this directory, is built on it.

  • Draftwright

    The drawing generator built on quiddity: it turns a STEP model into a dimensioned engineering drawing (SVG, PDF or editable DXF) using quiddity's feature recognition and PMI reading, rather than exposing the recognition itself as a library, command line and Python package.

  • OCCT

    The Open CASCADE geometry kernel: a general B-rep modelling, data-exchange and visualisation toolkit. quiddity is not a modelling kernel and deliberately works without one: it reads STEP and recognises manufacturing features from the B-rep geometry alone.

  • build123d

    A Python parametric B-rep modelling framework built on the Open Cascade kernel, for authoring new parts as code, rather than a recognition library that reads existing STEP parts and reports the features already in them.

Sources

  1. 1pzfreo/quiddity READMEgithub.com
  2. 2GitHub repo statsapi.github.com

Not verified

  • Real user voices: no real linked post read for this page; the repo was created on 2026-10-06 and has 112 stars.
  • Related Error Index entries: no matching entry verified.

Last verified November 19, 2026. Listings are informational. The evidence score measures public, verifiable evidence and is not an endorsement or a compatibility claim.