Coverage and proof

What works today, and what is still experimental

This one page combines available checks, their source support, software testing, and the limits that still prevent authoritative use.

79drawing-profile checks
Works in the review workspace

Ontario Part 9 drawing review

79 checks compare PDF evidence with mapped requirements. Every result still requires qualified human confirmation.

Open review workspace
Validated with public decisions

Official decision replays

2 decisions, 17 source facts, and 6 checks verify an exact match and a safe escalation.

See decision replays
Technical demonstrations

DXF, IFC, and fixed scenarios

These examples prove measurement, rule routing, and result states. They are synthetic and are not permit decisions.

Open technical demo

Product proof

What the testing actually establishes

17source-linked facts
39/39regression outcomes
6/6safe escalations
1000/1000simulation assertions

Demonstrated

  • The same inputs produce the same results.
  • Public-case facts are linked to their source.
  • The Ontario replay matches the published outcome.
  • Missing or conflicting evidence does not become a false pass.

Not yet demonstrated

  • Complete provincial or municipal code coverage.
  • Measured accuracy on complete submissions paired with decisions.
  • Approval or endorsement by an authority.
  • Authority to make a permit decision.
0
matched field cases acquired

Acquire a rights-cleared matched permit submission, correction history, and clause-level outcome, then complete independent double review before measuring field accuracy.

Accuracy claim blocked
Need the exact rule detail?

Open the explorer below to inspect sources, required inputs, executed expressions, and the status of each pack.

Explore rules, sources, and scenariosTechnical details

Available packs

Choose coverage to inspect

Official decision and citations linked2012 Ontario Building Code, as applicable to the disputeOpen source
Case-scoped rulesThese checks reproduce only the provisions and facts in the published case. They are not yet a complete province-wide rule pack.

Executable checks

2024 ONBCC 19

2 / 2 shownReproducible for the case
ON12-9.8.4.5-1A-WINDER-TURN

Maximum winder turn

2012 OBC, Division B, Clause 9.8.4.5.(1)(a)

Required inputs
  • Observed turn
  • Applicable maximum
Logicobserved turn <= 90°
Deterministic
ON12-9.8.4.5-2-WINDER-SEPARATION

Separation between winder sets

2012 OBC, Division B, Sentence 9.8.4.5.(2)

Required inputs
  • Number of sets
  • Measured separation
Logicone set OR separation >= 1,200 mm
Deterministic
Sources and validation material28
28researched sources
18usable now
6decision sources
0matched cohorts
Sample & Test FilesbuildingSMART International · International
cleared open
ACC function-generation reproduction packageTechnical University of Munich research team · International
IFC-Bench v2 question and answer datasetIFC-Bench authors · International
cleared open
Two-bedroom laneway housing reference package L2H-0City of Toronto · Toronto, Ontario
Building Permit Hub source, integration schema, and HOT2000 fixturesProvince of British Columbia · British Columbia
cleared open
Ishigaki-IDS-BenchONESTRUCTION Inc. · International, English and Japanese
cleared open
TERMIUM Plus construction and civil-engineering terminologyPublic Services and Procurement Canada, Translation Bureau · Canada
cleared open
EnerGuide Rating System Open DataNatural Resources Canada · Canada
cleared open
CiviX Server raw XML legislation and bylaw APIKing's Printer for British Columbia · British Columbia
cleared open
Machine-readable BC Building Code JSON and planned APIProvince of British Columbia · British Columbia
authority agreement required
CODE-ACCORD regulatory sentence corpus v1.0.0ACCORD Horizon Europe project · England and Finland
cleared open
CCMC Registry of Product Assessments JSON APINational Research Council Canada · Canada
Building Code Appeal Board decisionsProvince of British Columbia · British Columbia
Safety Codes Council appeal decisionsSafety Codes Council · Alberta
Saskatchewan Construction Standards Appeal Board decisionsGovernment of Saskatchewan · Saskatchewan
Building Code Commission rulingsOntario Building Code Commission · Ontario
Building and fire STANDATA interpretationsGovernment of Alberta · Alberta
cleared open
Housing Design Catalogue technical packagesCanada Mortgage and Housing Corporation · Seven Canadian regions
permission required
BC Standardized Housing DesignsProvince of British Columbia · British Columbia
permission required
NRC digitalized-code sample dataNational Research Council Canada · Canada
program access required
Issued building permitsCity of Vancouver · Vancouver, British Columbia
cleared open
Building PermitsCity of Calgary · Calgary, Alberta
cleared open
Owner-authorized building plans and issued e-permit packageCity of Vancouver · Vancouver, British Columbia
contributor agreement required
Applicant-held electronic permit package and examiner noticesCity of Toronto · Toronto, Ontario
contributor agreement required
Search of Records and plans requestCity of Edmonton · Edmonton, Alberta
contributor agreement required
Property Research recordsCity of Calgary · Calgary, Alberta
permission required
AEC-owned de-identified permit contributionsParticipating architects, engineers, designers, and applicants · Selected Canadian jurisdictions
contributor agreement required
Authority-held matched permit cohortParticipating authority having jurisdiction · Pilot jurisdiction to be selected
authority agreement required
Software test detailsValid
Regression100%39/39 expected outcomes

Checks engine states and pathways with internal test cases.

Safety replay6/6conservative dispositions

Checks that disputes and missing evidence are escalated rather than invented.

Simulation1000/1000250 generated cases

Checks determinism, state accounting, and missing-data safety.

These tests prove software behaviour. They do not measure field accuracy.
PACK d8708a39bd1b
Complete prototype requirement matrix11 demonstrated · 3 partial
M1mandatory

Multiformat permit input

Process a bounded mixed-file permit-package ZIP; reconcile facts and conflicts across embedded-text PDF, ASCII DXF, HOT2000 H2K, JSON/JSON-LD, and IFC members; normalize one shared stair vocabulary across PDF, mapped DXF dimensions, structured facts, and physically plausible IfcStairFlight attributes; extract a bounded, non-identifying set of HOT2000 energy facts; and execute an IFC/IFCZIP geometry profile that resolves project units, world placements, a review-confirmed role pair, and common tessellated, swept, BRep, Boolean, and mapped representations.

The package profile permits 50 files and 100 MB expanded data, rejects unsafe paths, and does not process nested archives. Mapped DXF and IFC dimensions require human confirmation. Contradictory or implausible stair values are withheld. OCR, proprietary DWG, long-tail vendor semantics, and ambiguous geometry remain human-review or later-phase work.
lib/input-formats.tslib/extraction.tslib/stair-validation-corpus.tsscripts/sync-stair-validation-corpus.tslib/part3-dxf-demo.tslib/ifc-geometry.tslib/ifc-broad-geometry.tstests/extraction.test.tstests/stair-corpus-live.test.tstests/input-formats.test.tstests/live-evidence.test.tstests/part3-dxf-demo.test.tstests/ifc-geometry.test.tstests/ifc-broad-geometry.test.tsPOST /api/applications/{id}/artifacts
Demonstrated
M2mandatory

Part 3 and Part 9 classification

Compute Part 3 or Part 9 scope from occupancy, storeys, area, and high-building status, then distinguish houses, suites, small buildings, and Groups A through F. Demonstrate that changing actual DXF footprint vertices across the size boundary changes the selected pathway.

The public technical demo proves dynamic classification with a clearly labelled synthetic DXF fixture. Its 600 m² scope mapping is illustrative, not an authoritative jurisdictional rule.
lib/compliance/classify.tslib/compliance/rules.tslib/part3-dxf-demo.tstests/classification.test.tstests/part3-dxf-demo.test.tsGET /challenge-lab
Demonstrated
M3mandatory

Machine-readable deterministic rules

Evaluate constrained, versioned logic without a generative model in the decision path; produce repeatable fingerprints; run 10-check NBC 2020 and NBC 2025 Part 9 profiles, a 30-check Ontario Part 9 profile, and a focused 23-check Ontario Part 3 exit-stair profile; preserve all four result states; link public case checks to cited provisions; and execute one exact IFC surface-distance rule over normalized product geometry.

The Ontario and national stair mappings remain preliminary, review-gated candidate validations, not complete code coverage or automated permit determinations. The separate IFC pipe-clearance rule remains illustrative. Complete authoritative coverage remains a later milestone.
lib/compliance/evaluate.tslib/compliance/nbc-2020-stair-candidates.tslib/compliance/obc-2024-stair-candidates.tslib/compliance/nbc-2025-candidates.tslib/stair-validation.tsscripts/run-stair-validation.tstests/nbc-stair-candidates.test.tstests/obc-2024-stair-candidates.test.tstests/obc-2024-part3-stair-candidates.test.tstests/stair-validation.test.tslib/ifc-geometry.tslib/ifc-broad-geometry.tslib/bc-pilot.tslib/ontario-pilot.ts
Demonstrated
M4mandatory

Human review and traceability

Preserve immutable machine findings, require reasons for adjudication, link evidence, expose evaluation traces, and record append-only audit events.

Prototype roles are simulated; production authorization requires an identity provider and RLS policy review.
components/application-workspace.tsxlib/store/state.tstests/store.test.ts
Demonstrated
M5mandatory

Digital code format compatibility

Validate canonical JSON/JSON-LD rule packs; round-trip the recognized CodeTrace XML Rule Pack Exchange 1.0 profile into canonical executable rules; quarantine unmatched XML/DITA for mapping; reconcile English and French source variants; carry IFC dictionary mappings; and pin official sources by hash.

Only the documented XML 1.0 profile is executable. Every other NRC or AHJ vocabulary, derived mapping, and bilingual source pair needs an approved adapter and qualified review.
lib/rule-packs.tslib/rule-pack-xml.tslib/bilingual-rule-reconciliation.tsscripts/sync-code-sources.mjstests/rule-packs.test.tstests/candidate-sandbox.test.tstests/official-sources.test.ts
Demonstrated
M6mandatory

Four result categories

Return Meets, Does Not Meet, Information Not Available, or Uncertain, while tracking non-applicability separately as an excluded disposition.

No confidence score can convert missing or uncertain information into a pass.
lib/compliance/evaluate.tstests/engine.test.tslib/assurance/benchmark.ts
Demonstrated
M7mandatory

User interface and open API

Provide a guided Canadian intake-to-decision workspace, separate public-decision replays, a clearly bounded synthetic technical demo, source-file and extraction review, finding, revision, audit, and combined coverage-and-proof views, validation-only APIs for fixed candidate scenarios and generalized NBC 2020, Ontario Code 2024 Parts 3 and 9, and NBC 2025 stair facts, and persistent Ontario Code 2024 or NBC 2025 stair candidate runs in the normal workspace results flow.

Each bounded stair profile persists findings, exact pack metadata, audit events, and human decisions, but remains clearly labelled review-gated candidate validation rather than an automated permit determination. General CA-DEMO checks remain illustrative. Production authentication and role-based authorization remain implementation gates.
app/workspace/page.tsxcomponents/review-workspace.tsxcomponents/application-workspace.tsxapp/api/revisions/[id]/run/route.tscomponents/rules-catalogue.tsxapp/api/rule-packs/sandbox/route.tsapp/api/rule-packs/stairs/route.tslib/openapi.tstests/stair-validation.test.tstests/store.test.ts
Demonstrated
M8mandatory

Canadian residency and on-premises path

Run fully local without external AI, build a non-root read-only on-prem container, and operate the hosted prototype with Supabase Canada Central plus Vercel Montréal functions.

The hosted prototype verifies regional configuration. Contractual residency, logs, backups, CDN behaviour, privacy controls, and subprocessors still require production review.
Dockerfile.runtimecompose.yamldocs/ON_PREM.mdvercel.jsonsupabase/config.toml
Demonstrated
M9mandatory

Exports and integration

Export PDF, SpreadsheetML, CSV, canonical JSON, and BCF 3.0 issue archives. Geometry issues include selected IFC GUIDs, a viewpoint, closest-point line, measurement, and deterministic trace.

The geometry topic and viewpoint validate against the official BCF 3.0 XSDs. Receiving-system certification, production APIs, and municipal connectors remain Phase 2.
lib/exports.tstests/exports.test.tslib/openapi.ts
Demonstrated
M10mandatory

English and French

Provide en-CA and fr-CA interface and rule content without changing deterministic IDs, values, states, or fingerprints, and report logic drift or missing source-language documents before candidate activation.

Shared canonical logic and source hashes detect drift; they do not replace qualified review of authoritative English and French legal text.
components/language-provider.tsxlib/compliance/rules.tslib/bilingual-rule-reconciliation.tstests/bilingual.test.tstests/candidate-sandbox.test.ts
Demonstrated
A1additional

openBIM exchange

Preflight IFC2X3, IFC4, IFC4X3, and IFCZIP; resolve a bounded ISO 16739 geometry profile with an exact fast path and web-ifc normalization fallback; export a BCF 3.0 selected-object viewpoint; preflight IDS structure; and document ISO 19650 as a process boundary.

This is a verified bounded profile, not universal IFC support, buildingSMART software certification, or ISO 19650 certification. Ambiguous selection and geometry outside limits fail safely for human review.
lib/ifc-geometry.tslib/ifc-broad-geometry.tslib/standards.tslib/exports.tstests/ifc-geometry.test.tstests/ifc-broad-geometry.test.tstests/exports.test.tsPOST /api/standards/validate
Demonstrated
A2additional

Accuracy measurement

Run versioned simple and complex cohorts, publish exact synthetic results, exercise full Part 3 and Part 9 golden workflows, pin an independent 12-model IFC and Solibri benchmark, maintain a two-person-reviewed Ontario drawing control, register official case-discovery sources, and block field claims until a representative matched case passes its acceptance gates.

Synthetic workflows, external model benchmarks, and public drawings prove software behaviour and conservative evidence handling. They are not matched Ontario permit decisions. Field accuracy and representative-environment readiness remain blocked until the MTO package is received, independently adjudicated, and replayed through every required gate.
lib/assurance/benchmark.tslib/external-stair-benchmarks.tslib/validation-manifests.tsvalidation/toronto-l2h-0-stair-answer-key.jsonvalidation/mto-representative-case.jsonscripts/sync-external-stair-benchmarks.tstests/external-stair-benchmarks-live.test.tstests/validation-manifests.test.tstests/golden-workflows.test.tslib/permit-case-sources.tstests/permit-case-sources.test.ts
Partial
A3additional

Revision identity and trust

Use append-only revisions, artifact SHA-256 hashes, run fingerprints, and Ed25519 revision seals pinned to a configured public-key trust anchor.

Development-session keys prove integrity during a run, not a professional engineer’s identity. Production must use managed keys and verified credentials.
lib/seal.tsPOST /api/seals/verifytests/seal.test.ts
Partial
A4additional

Configurable code updates

Provide bounded JSON, JSON-LD, and recognized XML rule-pack ingestion plus hash-verified official-source acquisition, duplicate detection, unmatched-format quarantine, bilingual logic reconciliation, and explicit activation state.

No imported or mapped pack becomes authoritative automatically. Permission, official-text reconciliation, independent bilingual review, and administrator approval are mandatory.
lib/rule-packs.tslib/rule-pack-xml.tslib/bilingual-rule-reconciliation.tsscripts/sync-code-sources.mjstests/rule-packs.test.tstests/vbbl-sources.test.ts
Partial
PACK d8708a39bd1b