If a model cannot be prepared for printing, read the viewer status and check the installed LDraw library and missing references first. Source export alone is not printable certification. Inspect Source versus Prepared Mesh; for difficult topology follow external repair. A missing compatible Verified Fit Profile leaves eligible nominal output available.
Fit Profile defaults to Automatic / No explicit selection, which never guesses between compatible profiles. Choose a managed Verified profile to specify the intended printer and process. This does not bypass feature, role, orientation, or evidence requirements. Print Orientation defaults to Nominal; explicit X, Y, or Z quarter-turns use only matching Verified evidence. A saved plan is bound to its printing context: if the profile, evidence, or orientation changes, refresh the corpus to start a new comparison run.
Audit results separate nominal geometry readiness from fitted-print results. A source inspection records existing supported source features before preparation; it does not repair geometry or grant fit ownership. Verified-zero means evidence was applied with a zero correction; partial fit means other recognized interfaces remain nominal. ManufacturingMesh failure leaves nominal PreparedMesh readiness recorded. Only validated exports that reopen successfully count as fitted success. Accepted files are under exports/success/; failed diagnostic candidates are under exports/diagnostic/. Use the CSV fit status and counts and the fitted-print funnel in summary.json; file existence alone is not success.
Debug builds include Test → Print Capability Audit. Random Sample filters the complete catalog first, then uses the displayed seed to shuffle eligible Parts without replacement. The requested count means that many eligible Parts when available; only sampled Parts appear in results.csv, ordered by sample_sequence. Part List reruns entered IDs. No-Color records (Rebrickable color 9999) and the authoritative Stickers catalog category are excluded. Exclude composite/decorated Part IDs is checked by default and restricts the audit population to IDs containing digits with at most one trailing letter. This is only an audit filter; it does not change catalog identity or normal printing. Aggregate catalog, exclusion, eligibility, sample-count, seed, and rule details are stored in run-metadata.json. The versioned CSV separates result_category (audit status) from category_id, rebrickable_category_id, and category_name (catalog data). A diagnostic candidate alone does not rescue a failed preparation, even when its existing bounded candidate mesh can be saved as a clearly labeled diagnostic 3MF; diagnostic_export_available and diagnostic_export_path record this separately from accepted exports. Choose an Output Folder with Browse; each run writes its CSV, metadata, and 3MF exports in a new run folder beneath it (default: Documents/BrickSuite/Print Capability Audits). Export filenames use <part_number>-<result_category>.3mf, without a sequence prefix. Stop finishes or safely cancels the active Part, flushes its result, and marks the run stopped; untested Parts remain only in the saved plan, not in the CSV. A missing compatible Verified Fit Profile is not itself a print failure: eligible nominal PreparedMesh may be exported.
The audit tries native printing first. If native preparation fails, an existing non-stale Strictly Validated or User-Accepted Local Override may be loaded, exported as nominal PreparedMesh, and reopened. The audit never imports, repairs, approves, or modifies an override and never invokes Experimental Auto Fit. CSV records the native result separately from the final result, including missing, stale, or invalid overrides. Native success and practical printable success use the same model-bearing denominator; override-assisted recoveries are shown separately. A diagnostic export or an unaccepted repair candidate is not a success. To compare with an older run, replay its saved Part list in order and keep the original run folder.
Exclude Parts without installed LDraw model defaults to unchecked: catalog-level samples can report no_ldraw_model. When checked in Random Sample mode, only Parts with a readable, nonempty installed model file (including known LDraw aliases) enter the shuffle, so missing models do not use sample slots. This is an availability check, not a guarantee of valid geometry or successful printing. excludedNoModel in the metadata counts these exclusions. Part List ignores this option and reports explicitly requested missing models normally.
Before processing, run-state.json saves the entire ordered sample, seed, eligibility settings, and requested/actual counts. Before each Part loads, its sequence and identity are durably checkpointed as active. Each completed CSV row is flushed before the completed count advances. A clean finish has status completed; Stop has status stopped. After a crash, a running state identifies the active or last finished Part through currentSampleSequence, currentPartNumber, currentPartStatus, and completedCount. Opening the audit shows incomplete runs in the remembered output folder; Browse can select another output or individual run folder. Keep the original folder for diagnosis and use Part List for a targeted rerun. Automatic resume is not provided.
The live audit status shows the current Part and phase, including preparation, diagnostic export, reopen, and persistence. Diagnostic exports are optional and may be unavailable when workload limits are reached; failed preparations remain failures. The audit bounds sequential Boolean work and uses the existing validated local-composition route where possible. Stop changes to Stopping immediately and persists the active phase. Close or the window X while work is active shows Stopping / closing and closes automatically at the worker's safe completion boundary. Completed Stop closes immediately. Reopening the audit identifies both incomplete and stopped runs in the selected output folder. Stage timings are recorded in stage-timings.jsonl.
3D Printing overview | Help Home
Part Reference audits the current reference corpus in saved order, with each unique Part tested once. Review the position, unique-Part, duplicate-membership, and installed-model counts. Choose First unique Part and Parts in this run for a bounded range; All remaining covers the rest. This mode uses reference membership rather than the Random Sample eligibility exclusions or shuffle.
Refresh current Part Reference discovers a new corpus. Load saved Part Reference plan... reuses a saved plan for comparison; keep the printing context consistent. The saved plan and durable run state identify intended order and actual progress. A corpus result describes tested models and settings, not universal printability. This developer tool is separate from ordinary viewer preparation.