specify-core

Add manufacturing intent to a STEP part, and write it as AP242 semantic PMI.

specify-core reads a STEP part, recognises its features (holes with their counterbores and threads, hole patterns, slots, bosses, fillets and turned bodies) and works out what a manufacturer still needs to know: the material, datums, fits, thread classes, hole positions, surface finish and the general tolerance. It asks a question for each gap, with a suggested default and the reason for it. Your answers are written back into the STEP as AP242 semantic PMI (GD&T that CAD and inspection tools can read), and the file is read back to verify it.

It is deterministic: rules and standards tables, no AI. The answers are stored in the file too, so a later revision of the part can be opened and carried on with.

• No dependencies: no OpenCascade, build123d or quiddity. The engine is Rust, in one wheel.
• Wheels for Linux (x86_64, aarch64), macOS (Apple silicon, Intel) and Windows, for CPython 3.12 and later.
• Apache-2.0.

It is the engine behind Draftwright Specify, which adds a 3D review of the same questions in the browser.

Install

pip install specify-core

Python

from specify_core import api

analysis = api.analyse("plate.step")         # features, faces and any existing PMI

for q in api.questions(analysis, {}):         # what is still unknown
    print(q.id, "|", q.prompt, "| default:", q.default)

answers = {"part.material": "6082-T6", "hole.function:holes:9": "fit:H7"}
intent = api.apply(analysis, answers, accept_defaults=True)   # suggested defaults for the rest
report = api.write("plate.step", intent, "plate-pmi.step", answers=answers)

print(report.written)       # ['datum A', 'Ø8 H7 on holes:9', 'position 0.05 on holes:9', ...]
print(report.not_written)   # anything that could not be written, with the reason

Opening the written file again picks up where you left off: api.analyse("plate-pmi.step")["resume"]["answers"] returns the answers stored in it.

The main functions, all in specify_core.api:

Answers are a dict of question id to value, e.g. {"part.material": "6082-T6"}. Requirements no question asked for (a flatness, a fit, a finish on faces you choose) are added as extra.* answers.

Command line

specify-core analyse plate.step -o plate.analysis.json
specify-core questions plate.analysis.json --open                # unanswered questions, as JSON
echo '{"part.material": "6082-T6"}' > answers.json
specify-core write plate.step plate.analysis.json answers.json --accept-defaults -o plate-pmi.step

specify-core parts lists an assembly's parts and --part picks one; write takes an analysis and answers pair for each part to write together. specify-core mesh writes the face-indexed triangles.

For a host program

specify-core serve answers JSON requests, one per line, on stdin and stdout. Each request carries everything it needs, so a restarted process loses nothing. This is how an application drives it as a child process:

{"id": 1, "op": "analyse", "step": "/path/part.step"}
{"id": 2, "op": "questions", "analysis": {...}, "answers": {...}}
{"id": 3, "op": "write", "step": "/path/part.step", "output": "/path/out.step", "analysis": {...}, "answers": {...}, "accept_defaults": true}

Each reply is {"id", "ok": true, "result"} or {"id", "ok": false, "error": {"code", "message"}}, where the code is incomplete (with what is missing), invalid or failed. analyse replies with the analysis and the mesh; write with the report and the intent. specify_core.serve.reply(request) answers the same requests in-process.

What it writes

AP242 semantic PMI: datums and datum systems, size dimensions with ± and fit tolerances, location dimensions, position, flatness, perpendicularity, parallelism, profile and runout tolerances, threads, knurls, surface texture, material, notes and the general tolerance. Every file is read back and checked before it is kept.

Links

• Source and issues: github.com/pzfreo/specify-core-rust
• Changelog: CHANGELOG.md
• Building and testing this package: docs/python.md
• Try it in the browser: editor.draftwright.io/specify

Status: early (0.x). The JSON formats may change between minor versions; changes are listed in the changelog. Draftwright® is a registered trade mark of Betwixt Ltd.