worldbox-observer-mod/docs/event-e2e.md

Event E2E suite (maintainable)

Goal: every catalog event is covered by automation that stays cheap when new ids are added.

Live accuracy roadmap: event-live-verification-plan.md.

Observe → confirm → emit: event-observation.md.

Tracks

Track	Scenario	What it proves	How new events join
Inventory audits	nested in event_coverage	Live library vs authored catalog	Add JSON row / live seed; audit fails until matched
A inject coverage	event_inject_coverage / ec30	Every camera-worthy id can register an interest key + label	Auto via EventCatalog.*.AuthoredIds
Live apply loop	event_live_apply_loop	Happiness Signal + status gains via real apply	Auto loop; ledger id / id_drop / fail
Live relationship	event_live_relationship	Vanilla setters/Beh + join negative; leave/family_removed/baby live fires	Shared runner + ledger claims
Live pipelines	event_live_pipelines	Traits/decisions/WorldLog/wars/eras/books/plots + building destroy; boat/combat best-effort (libraries unsupported)	Auto loops + honest gaps
Live outcome smoke	event_live_outcome_smoke	Short family + sample apply	Curated
Live coverage ledger	event_live_coverage	Honest `none	feed

Suite split (weight-aware)

Gate	Includes	When
event_suite	audits + inject + apply loop + relationship + outcome smoke	Day-to-day event gate
regression	event_suite + event_live_pipelines (+ other product smokes)	Full gate
critical_smoke ×3	Spectator/settings/tip/exit	PR gate

event_live_pipelines stays regression-only.

Flake budget is proven (--repeat 3 PASS); it remains heavier than day-to-day.

Coverage levels

Recorded in .harness/event-live-coverage.tsv (gitignored; regenerated each live run).

Level	Meaning	Verified?
none	Inject only; no live fire this run	No
feed	Production feed entry with synthetic payload	Partial
pipeline	Vanilla fire proved the shared hook family; this id may not have fired	Partial
id_drop	Real fire; authored drop matched	Yes (accurate drop)
id	Real fire + expected note/interest	Yes
unsupported	No deterministic fire API; reason in notes	Honest gap
fail	Unexpected miss/drop	Broken

Inject / domain_feed never writes a live level other than leaving none.

Busy policy

Most feeds early-out on AgentHarness.Busy.

Live runners set ForceLiveEventFeeds (relationship also sets ForceRelationshipFeeds).

LiveFeedsAllowed = !Busy || ForceLiveEventFeeds || ForceRelationshipFeeds.

Ledger notes should record busy_bypass=ForceLiveEventFeeds (or the specific flag) when used.

Commands

./scripts/harness-run.sh event_inject_coverage
./scripts/harness-run.sh event_live_apply_loop
./scripts/harness-run.sh event_live_relationship
./scripts/harness-run.sh event_live_pipelines
./scripts/harness-run.sh event_suite
./scripts/harness-run.sh regression

Artifacts (under IdleSpectator/.harness/, gitignored):

• event-inject-coverage.tsv - inject wiring
• event-live-coverage.tsv - source of truth for live proof
• event-trigger-audit.tsv - design / call-site audit (not live proof)

Design rules

1. Loop AuthoredIds - no per-id harness steps for inject/apply/traits.
2. Inject ≠ live. domain_feed is never live.
3. ForceLiveEventFeeds / LiveFeedsAllowed bypass Busy during live runners only.
4. Coverage levels must not over-claim (see plan).
5. Registry max-96: remove/clear during mass live loops so trim does not false-fail.
6. Assert fails on fail > 0; do not fail on camera none until a budget is explicitly tightened.

Code

• EventInjectCoverageHarness, EventLiveApplyLoopHarness, EventLiveRelationshipHarness, EventLivePipelinesHarness
• EventLiveCoverageLedger
• Asserts: event_inject_coverage, event_live_apply_loop, event_live_relationship, event_live_pipelines, event_live_coverage