Skip to content

Curriculum Quality Dashboard

This document defines the persisted quality-status layer used by the local Workbench dashboard at /quality-dashboard.

The dashboard is intentionally read-only. It reads generated status snapshots from docs/qa-ci/status/ and does not reimplement validator logic in React.

Detailed semantics for maturity levels, QA scopes, route coverage, and every CQR-* rule are documented in:

  • docs/qa-ci/curriculum-quality-maturity-and-routes.md

Files

  • Generator: app/scripts/generateCurriculumQualityStatus.ts
  • JSON snapshot: docs/qa-ci/status/curriculum-quality-status.json
  • Markdown snapshot: docs/qa-ci/status/curriculum-quality-status.md
  • Memory-card review checker: app/scripts/memoryCardReview.ts
  • Memory-card review config runner: app/scripts/runMemoryCardReviewConfigs.ts
  • Memory-card rollout triage: app/scripts/reportMemoryCardReviewRollout.ts
  • Memory-card review ledgers: curricula/DE/Gymnasium/quality/memory-card-review/*.review.jsonl and *.cards.review.jsonl
  • Memory-card review audit reports: docs/qa-ci/status/memory-card-review-canonical-math-full.md, docs/qa-ci/status/memory-card-review-canonical-physics-full.md, docs/qa-ci/status/memory-card-review-canonical-chemistry-full.md, docs/qa-ci/status/memory-card-review-canonical-biology-full.md, docs/qa-ci/status/memory-card-review-canonical-german-full.md, docs/qa-ci/status/memory-card-review-canonical-history-full.md, docs/qa-ci/status/memory-card-review-canonical-informatics-full.md, docs/qa-ci/status/memory-card-review-canonical-latin-full.md, docs/qa-ci/status/memory-card-review-canonical-politics-economics-full.md, docs/qa-ci/status/memory-card-review-canonical-economics-full.md
  • Memory-card rollout report: docs/qa-ci/status/memory-card-review-rollout.md
  • M0 remediation report generator: app/scripts/reportM0RemediationPlan.ts
  • M0 remediation report: docs/qa-ci/status/m0-remediation-plan.md
  • Englisch M0 remediation pilot generator: app/scripts/reportEnglishRemediationPilot.ts
  • Englisch M0 remediation pilot report: docs/qa-ci/status/english-remediation-pilot.md
  • Source-coverage audit generator: app/scripts/generateCurriculumSourceCoverageAudit.ts
  • Source-coverage audit JSON: docs/qa-ci/status/curriculum-source-coverage-audit.json
  • Source-coverage audit Markdown: docs/qa-ci/status/curriculum-source-coverage-audit.md
  • MEM SPARQL consistency audit generator: app/scripts/generateMemSparqlConsistencyAudit.ts
  • MEM SPARQL consistency audit config: curricula/DE/Gymnasium/quality/mem-sparql-consistency/canonical-math-poc.config.json
  • MEM SPARQL consistency audit JSON: docs/qa-ci/status/mem-sparql-consistency-audit.json
  • MEM SPARQL consistency audit Markdown: docs/qa-ci/status/mem-sparql-consistency-audit.md
  • Workbench route: /quality-dashboard
  • Local dev endpoint: /__quality-dashboard/status
  • Local status-file endpoint: /__quality-dashboard/file?path=docs/qa-ci/status/<file>.md

Regenerate the snapshots with:

cd app
npm run quality:curriculum-status
npm run quality:memory-card-review:report:all
npm run quality:memory-card-review:check:all
npm run quality:memory-card-review:rollout
npm run quality:memory-card-review:pilot-dossier
npm run quality:m0-remediation
npm run quality:english-remediation-pilot
npm run quality:source-coverage-audit
npm run quality:mem-sparql-consistency

quality:memory-card-review:*:all discovers all *.config.json files in curricula/DE/Gymnasium/quality/memory-card-review/. Add a config only when the CQR-302 ledger is meant to be enforced; Mathematik, Physik, Chemie, Biologie, Deutsch, Geschichte, Informatik, Latein, Politik und Wirtschaft, and Wirtschaftswissenschaften are currently completed and enforced in CI.

quality:mem-sparql-consistency is deliberately non-blocking. It probes the live MEM/FWU SPARQL endpoint, writes a Mathematik/Gymnasium review report, and keeps endpoint gaps or source-text discrepancies as triage issues instead of CI failures.

quality:memory-card-review:pilot-dossier is a non-enforced preparation step for subject-level pilots. It writes a semantic triage dossier without creating or changing the enforced CQR-302 config or ledger.

To initialize a new visible-but-open queue for a curriculum, create the config file first and then run:

cd app
npm run quality:memory-card-review -- --mode=bootstrap --config=curricula/DE/Gymnasium/quality/memory-card-review/<review-id>.config.json

The bootstrap mode writes only needs_developer_review records for missing ordinary atomic goals. It is a visibility step, not a semantic pass decision.

Maturity Levels

The dashboard reports one conservative maturity level per curriculum and, where available, per configured route scope. The top-level M0-M6 cards are aggregate maturity counts. They change only when at least one curriculum changes maturity. To make that explicit, the dashboard also shows an M5/M6 review gates strip for CQR-301, CQR-401, CQR-501, and CQR-302; these gate cards show how many rule instances pass and how many are still open. The first three rules carry the core M5 level. CQR-302 is the additional M6 memory layer, so missing memory-card configuration blocks M6, not M5.

Level Meaning
M0 Basic graph quality is visible, but source-ingestion is not yet complete or original source URLs are missing.
M1 Original source inventories are readable, linked to official HTTP(S) source URLs, and their extracted goals are registered in the source membership/closure ledger.
M2 Bundesland composition-view atoms are fully source-backed, and registered source original goals are fully covered by the Bundesland view.
M3 Every configured learner-facing QA scope has a clean route from motivation through atomic learning goals to terminal autonomy.
M4 Terminal autonomy goals in every configured QA scope are exam-mode-capable via examData or an explicit reviewed exception convention.
M5 Core review/readiness layer is clean: semantic atomicity is current, composition views exist, and no active or obsolete applicability warning debt remains.
M6 M5 plus reviewed memory-card layer: memory-card decision tracing is current or explicitly reviewed, every kept primary card is traced, and configured learner-facing views expose required memory nodes.

M5 is intentionally strict for the core curriculum layer. M6 is stricter again and is reserved for curricula whose optional memory-card layer has also been semantically reviewed.

Source References And Local Working Copies

The dashboard treats original curriculum sources in two separate layers:

  • The committed, durable layer is the structured source reference in Git: sourceDocument/sourceDocuments with official HTTP(S) URLs, titles, roles, and optional local path hints.
  • Local PDFs or HTML files are working copies for extraction, page checks, and local review. They are normally ignored by Git through curricula/**/*.pdf and similar rules.

This distinction is intentional. A green source badge means the official source reference is documented for the currently declared source scope and the extracted or retained source inventory is readable and registered. It does not mean that the original PDF is committed to Git.

Local source-file availability remains useful diagnostic information:

  • If a generator must re-run PDF extraction, the local file must exist or be downloaded from the documented official URL first.
  • If the persisted extraction is already present, the dashboard can still be green with 0/n local working copies, as long as n/n official URLs are documented and the extracted inventory is readable.
  • The UI therefore labels this as source links, not as bundled source files.

For German Gymnasium quality work, "source scope is green" means the source situation is explicitly decided for every relevant cell that the curriculum claims to cover: Bundesland, subject, Sekundarstufe I/II, and G8/G9 or other duration models where applicable. A cell can be green either through its own retained source inventory or through an explicit reviewed decision that no separate source/duration split is required.

Rule Families

The current rule catalog is versioned as curriculum-quality-v2.

Rule Target Meaning
CQR-000 M1 Source inventory ingestion: original source inventories are readable, linked to official HTTP(S) source URLs, and their extracted goals are registered.
CQR-001 M0 Basic graph integrity: IDs, local/global references, self-reference guards, and direct DAG checks.
CQR-002 M0 Explicit type metadata matches structural atomic/cluster classification.
CQR-003 M2 Bundesland atomic coverage: every declared jurisdiction view has full source-backed coverage for its source-coverage atoms, no unsupported assigned source-coverage atom, and no registered source original goal missing from the view mapping.
CQR-004 M2 Course-level mapping consistency: source goals marked GK_LK, LK, or unspecified map only to canonical SkillPilot goals with compatible GK/LK tags; unspecified defaults to GK_LK unless an LK-only decision is explicitly reviewed.
CQR-101 M3 Effective full route coverage through R_eff: motivation anchor -> selected atomic goals -> terminal autonomy.
CQR-102 M3 Direct atomic route coverage through authored atomic requires.
CQR-103 M3 No route-scope cluster-level requires remain for ordinary sequencing.
CQR-201 M4 Terminal autonomy goals have examData.
CQR-301 M5 Semantic atomicity review ledgers for content leaf goals are complete, current, and resolved.
CQR-302 M6 Memory-card review ledgers explicitly decide whether ordinary atomic goals justify memorization support; every kept primary card must trace to such a goal-level decision, every memory deck must remain traceable, and configured learner-facing views must expose the referenced memory nodes. Missing review configuration blocks M6, but not M5.
CQR-401 M5 At least one learner-facing composition view exists for the curriculum.
CQR-501 M5 Applicability warnings are split into active, accepted-current, and obsolete-accepted counts.

CQR-* rules are dashboard/readiness rules. They complement GVR-*, APV-*, and CPV-* validators documented in docs/qa-ci/graph-validation-rules.md; they do not replace those validator families.

Mapping Pipeline Visibility

The dashboard also renders the curriculum-mapping processing pipeline when persisted source-extraction artifacts provide pipelineStatus. This is separate from the M0-M6 rule calculation: it shows how far the source-to-SkillPilot implementation has progressed without pretending that a later step is complete.

The pipeline is reported per canonical curriculum and per source landscape:

  • MAPPING-1: official original Lehrplan passages extracted.
  • MAPPING-2: source goals created from those passages.
  • MAPPING-3: source goals mapped to canonical SkillPilot goals.

Each step is complete, incomplete, or blocked. The dashboard table shows completed source pipelines over all expected source pipelines, for example 1/31. This is intentionally source-based: one completed Sek-II extraction must not make a combined Sek-I/Sek-II curriculum look complete. If a mapping source has no persisted source-extraction artifact yet, the dashboard creates an explicit 0/3 placeholder with MAPPING-1 open and dependent steps blocked. The detail panel shows the current open step, the source landscape, jurisdiction, number of passages, number of source goals, and each step badge.

For Hessen Mathematik Sek II the current expected state is:

  • MAPPING-1 = complete
  • MAPPING-2 = complete
  • MAPPING-3 = complete, with the complete slices E.1 to E.7, Q1.1 to Q1.5, Q2.1 to Q2.5, Q3.1 to Q3.5, and Q4.1 to Q4.3: 316/316 source goals reviewed, 316 mapped, 0 canonical-goal gaps, and 0 placement/view questions. The source-goal denominator is now 316 because two former Q3.2 entries were corrected as Normalverteilung formula extraction artifacts and one Q4.2 entry was corrected as a domain-heading artifact (Analytische Geometrie:), not an official learning goal.

That means the official text extraction and source-goal creation are currently considered completed by the persisted checks, while canonical SkillPilot mapping remains explicitly open and quantifies the remaining work.

For Hessen Mathematik Sek I the current expected state is intentionally different:

  • MAPPING-1 = complete: the primary Kerncurriculum, the G9 Lehrplan passages, and the implementation Leitfaden are visible as official/source-supporting passages.
  • MAPPING-2 = complete: the persisted extraction now contains 99 KC 7.3 content-field source goals, 86 KC 7.1/7.2/6 competency-expectation source goals, and 241 granular G9 topic-line source goals.
  • MAPPING-3 = complete: the review lane now maps all 426/426 reviewed source goals to canonical SkillPilot goals, covering all KC 7.3 content-field source goals (Zahlen, Operationen, Ebene Figuren, Körper, Beziehungen geometrischer Objekte, Umgang mit Größen, Messvorgänge, Zuordnungen, Funktionen und Gleichungen, statistische Erhebungen, Umgang mit dem Zufall), all KC 7.1/7.2/6 competency expectations, and all granular G9 Jahrgänge 5-10 topic-line source goals; no source-goal mapping gaps remain in this lane.

Bundesland Coverage

The dashboard persists and renders Bundesland coverage for learner-facing composition views, not for every raw node in the JSON file.

For each canonical curriculum and each declared jurisdiction, the snapshot stores:

  • raw atomic goals: all leaf nodes in the canonical JSON file,
  • DE source-view atoms: the union of source-coverage-relevant atomic nodes rendered by the Germany-wide de-de-* composition views,
  • Bundesland source-view atoms: the union of source-coverage-relevant atomic nodes rendered by the matching Bundesland composition views,
  • source-backed Bundesland view atoms,
  • Bundesland view atoms without sufficient Lehrplan evidence,
  • registered source Lehrplan atoms,
  • source Lehrplan atoms that map into the Bundesland view,
  • source Lehrplan atoms not mapped into the Bundesland view,
  • extracted source atoms from registered source inventories,
  • extracted source atoms missing from the source membership/closure ledger,
  • registered source original goals,
  • registered source original goals fully covered by Bundesland view atoms,
  • projection errors and warnings.

Atomic nodes count for CQR-003 only when they are relevant for curriculum source coverage. Memory/SRS, practice, assessment, motivation, orientation, and examData goals are excluded from this numerator and denominator; they are checked by their own QA lanes. Cluster nodes do not count for the CQR-003 numerator or denominator.

DE source-view atoms is the national reference maximum for the current canonical view set after this source-coverage filter. For example, mathematics currently has more raw atomic leaves in the JSON than DE source-view atoms because not every raw leaf is part of source-coverage QA or the learner-facing national view.

The table value Bundeslaender is cleanJurisdictions / totalJurisdictions. A jurisdiction is complete only when all checks are clean:

  • every Bundesland source-view atom is backed by accepted Lehrplan evidence,
  • no Bundesland source-view atom is assigned without such evidence,
  • every registered source original goal is fully covered by canonical atoms rendered in that Bundesland view,
  • every extracted source atom from available source inventories is registered in the source membership/closure ledger.

The extraction check has one hard boundary: it can only verify source inventories that exist in the repository and can be parsed. If the official Lehrplan has not yet been turned into a complete source extraction or retained source snapshot, the dashboard cannot infer the missing official goals automatically; that remains explicit source-ingestion work.

Only these evidence kinds count directly for CQR-003 content coverage:

  • provenance: direct canonical-goal provenance from a registered source curriculum.
  • mapping: reviewed mapping from a source curriculum goal to the canonical goal, including reviewed partial mappings when the source goal is still content-covered.
  • reviewed surrogate evidence: an explicitly accepted requires-closure surrogate entry for a real logical prerequisite gap.

These evidence kinds do not count as Lehrplan coverage for CQR-003:

  • override: manual visibility/applicability overlay.
  • child-union: cluster inference from visible children.
  • automatic requires-closure: prerequisite visibility inference without an explicit surrogate review.

Reviewed partial mappings still remain visible as passgenauigkeit warnings. They are content coverage, but not a passgenaue one-to-one source/canonical relation. An accepted APV-201 override warning may keep a projection operationally visible, but it does not satisfy Bundesland coverage. Lower values such as 28/77 belegt remain visible in the detail panel as partial evidence. Values such as 419 nicht belegt are explicitly treated as unsupported assignments, not as coverage.

Current Route Profiles

The currently required route profiles are:

  • canonical-math-sek1
  • landscape: Mathematik (Gymnasium, DE)
  • scope: Sekundarstufe I
  • motivation anchor: Warum Mathematik? – Entdecken, Muster & Alltag
  • terminal autonomy target: Sek-I-Abschlussaufgaben Mathematik

  • terminal autonomy cluster: Übungen Sekundarstufe I

  • canonical-math-sek2

  • landscape: Mathematik (Gymnasium, DE)
  • scope: Sekundarstufe II
  • motivation anchor: Warum Mathematik? – Denken, Muster & Zukunft

  • terminal autonomy targets: the atomic exam-mode goals under Übungen E-Phase, Übungen Q1, Übungen Q2, Übungen Q3, Übungen Q4, and Übungen Prozesskompetenzen

  • canonical-physics-sek1

  • landscape: Physik (Gymnasium, DE)
  • scope: Sekundarstufe I
  • motivation anchor: Warum Physik?

  • terminal autonomy target: atomic exam-mode goals under Übungen Sekundarstufe I Physik

  • canonical-physics-sek2

  • landscape: Physik (Gymnasium, DE)
  • scope: Sekundarstufe II
  • motivation anchor: Warum Physik?
  • terminal autonomy targets: atomic exam-mode goals under Übungen E-Phase, Übungen Q1, Übungen Q2, Übungen Q3, and Übungen Q4

The Sek I scope mirrors the existing GVR-011 / GVR-012 rollout profile. The Sek II scope carries the same dashboard contract for the upper-secondary part of the shared canonical mathematics curriculum.

  • CQR-101 should remain green.
  • CQR-102 is the migration target from effective route coverage to direct atomic route coverage.
  • CQR-103 tracks remaining scoped cluster-level requires.
  • CQR-201 tracks which year-level practice nodes still lack concrete examData.

Mathe Sek I Perfection Path

To bring canonical mathematics Sek I from M3 toward M4 after source ingestion and Bundesland coverage are clean:

  1. Keep GVR-012 / CQR-101 green while editing.
  2. Move broad Sek-I cluster-level requires down into precise atomic prerequisites.
  3. Ensure each selected atomic goal has a direct atomic path back to the motivation anchor.
  4. Preserve the existing direct terminal path into the Sek-I capstone.
  5. Add concrete examData to the J5-J10 terminal autonomy goals or document a machine-readable exception convention.
  6. Regenerate curriculum-quality-status.json and rerun graph validation.

Commands:

cd app
npm run quality:curriculum-status
npm run validate:graph

For a full release-quality pass, also run the relevant view and review validators:

cd app
npm run validate:view-filters
npm run validate:composition-views
npm run quality:semantic-atomicity:check -- --config=curricula/DE/Gymnasium/quality/semantic-atomicity/canonical-math-full.config.json