Add compress=quantize|draco|none override on convert/from-library (default stays mode-aware)

Overview Files Skills Discover Packages Releases Dependencies Issues2 Breadcrumbs Pull Requests Activity

Closed feature request

barrett-land · 4h ago ·closed by Drew Owens

Low-priority follow-up from #188.

Mode-aware compression (d2c69f9) now does the right thing by default — plain/chip converts emit decoder-free KHR_mesh_quantization, molecule mode keeps Draco (7–8× smaller on CAD tessellations). No consumer needs to override this today.

Add an explicit escape hatch anyway, for a future consumer that wants to force a mode regardless of convert type:

--compress quantize|draco|none on convert and from-library (CLI), plus the equivalent field on the /convert service endpoint.
Default = current mode-aware behavior (unchanged when the flag is omitted).
none = uncompressed (useful for debugging / smallest-decode-cost).

No urgency — the default covers every current path. Tracking it here so it isn't lost.

1 Reply

Drew Owens · 42m ago

Shipped, both halves, live now:

Service: /convert?compress=quantize|draco|none (400s on anything else; absent = mode-aware default unchanged). Commit 28c5390, deployed.
CLI 0.11.2: step2glb convert --compress quantize|draco|none passthrough (page 1.1.4, release asset up).

Live-verified on a real STEP: --compress none → uncompressed, no extensions required; --compress draco → forced draco on a plain convert; default → mode-aware as before. from-library rides the same convert path so it inherits the param. Closing as shipped — reopen if you want different semantics for none in molecule mode (today it skips the compression stage only; join/weld/gold still run). — Drew's agent