Question and use cases
Seven logical categories map to six base-table groups with distinct schedules.
Use this data map when adding an audit category, querying historical rows, or changing retention.
Use the diagram during design review, incident diagnosis, code walkthroughs, and onboarding. It intentionally omits classes and projects unrelated to the focused question, so read it together with the source entry point and related topic pages.
This page covers only the named responsibilities and relationships; consult the relevant module documentation for capabilities not shown here.
Reading path
- Step 1: Map the logical category to its canonical base-table group.
- Step 2: Check whether the group is unsharded, monthly, or yearly.
- Step 3: Read retention independently from shard cadence; they answer different lifecycle questions.
Do not skip arrow direction, numbering, or group titles: they express dependency or time direction, execution order, and responsibility boundaries. If the diagram differs from current source or accepted architecture decisions, correct the generated catalog immediately.
Legend and notation
| Visual element | Meaning | What it does not imply |
|---|---|---|
| Coral focus | The decision point or primary path emphasized by this view | That the element is always more important or privileged |
| Blue boundary | An external system, protocol edge, or explicit boundary | That it must be an independently deployed service |
| Muted connector | A dependency, call, transition, or data flow as named by its label | Synchronous, same-transaction, or exactly-once behavior |
| Group frame | A responsibility, layer, or lifecycle phase | A team or physical-machine boundary |
This is a DATA MAP diagram. Use that form to understand the layout, then test your interpretation against the source facts below.
Key relationships and design meaning
Relationship 1
Source fact: Seven logical AuditCategory values map to six physical base-table groups.
Architectural meaning: Logical categories can evolve independently from physical table count, but every query must preserve the mapping.
Review action: Will query discovery include every physical shard for the requested time range?
Relationship 2
Source fact: High-volume request, external-request, and communication logs shard monthly.
Architectural meaning: Monthly sharding bounds high-volume table size and makes time-range shard discovery a first-class query responsibility.
Review action: Does a shared table preserve category discrimination?
Relationship 3
Source fact: Security and exception rows share SysSpecialLog and are distinguished by Level.
Architectural meaning: A shared special-log table saves schema duplication only if level/category discrimination remains reliable in writes, queries, and retention jobs.
Review action: Is retention value zero interpreted as indefinite retention?
Design review questions
Answer each question when reviewing or implementing a related change. If code, tests, or an ADR cannot support the answer, do not decide from the diagram alone.
- Will query discovery include every physical shard for the requested time range?
- Does a shared table preserve category discrimination?
- Is retention value zero interpreted as indefinite retention?
How to use the answers
- Identify the single owner of the capability or rule.
- Confirm that dependency, call, or event direction does not reverse ownership.
- Capture the conclusion with an automated test or repeatable command.
Boundaries and common mistakes
Boundary 1
Retention value 0 means keep indefinitely; it is not an immediate deletion policy.
- Do not infer: A connector in the diagram does not mean every implementation uses synchronous calls, a shared transaction, or shared data ownership.
- Review requirement: Does a shared table preserve category discrimination?
Boundary 2
Query ports discover physical shards by canonical base-table prefixes.
- Do not infer: A connector in the diagram does not mean every implementation uses synchronous calls, a shared transaction, or shared data ownership.
- Review requirement: Is retention value zero interpreted as indefinite retention?
Source verification and regeneration
The primary verification entry point is AuditCategory.cs + AuditTableDefinitions.cs + Audit*Schedule.cs. Inspect it and its direct references before regenerating diagram assets and pages.
# Regenerate bilingual SVG, standalone HTML, and documentation pagesnpm run diagrams
# Check locale pairs, references, safety attributes, accessibility, and canvas boundsnpm run verify:diagrams# Confirm that generation changed only the intended diagrams and pagesgit diff -- scripts/diagrams diagram-sources public/diagrams src/content/docs
# Verify internal links, MDX structure, and professional depth gatesnpm run verify:docsnpm run audit:docs-depthChange-completion checklist
- Responsibilities, order, states, and relationships match current source.
- Every new element helps answer this page’s question instead of turning the diagram into an inventory.
- Arrows have explicit direction and semantics without implying nonexistent synchronous or transactional guarantees.
- Chinese and English titles, labels, facts, and boundaries remain semantically equivalent.
- The diagram remains readable on narrow screens, in fullscreen, and while zoomed, with no overlap or overflow.
- Relevant architecture tests, integration tests, or verification commands have run.
- If a long-lived constraint changed, its ADR or architecture documentation is updated.