LiveArtifactRepository uses a hot subscription that does not replay history,
so its cache was always empty when verifyProduces ran immediately after emit.
Fix by reading directly from EventStore (synchronous) instead of the repository.
Also emit ArtifactCreatedEvent for non-llmEmitted produces slots after stage
validation passes, and fix tool parameter decoding to use JsonPrimitive.content
so escape sequences like \n are resolved rather than passed as literal backslash-n.
Deletes 21 Module.kt scaffolding objects that were never wired into any DI
registry. Removes unused imports across 8 production and 3 test files.
Restores StageExecutor, CyclePolicyResolver, PolicyValidation, and
ReplayContractTest — these are deferred features, not dead code.
Removes 6 files audited as fully unused: StageExecutor interface (shadowed
by kernel), CyclePolicyResolver and PolicyValidation (deferred cycle policy),
ApprovalRoutes (never registered), TuiNavigation (superseded by reducer),
and ReplayContractTest abstract class (no concrete subclasses).
ArtifactRelationshipType and CyclePolicy were audited but retained —
both are actively referenced by production code.
- SessionOrchestrator: wire T2/T3/T4 approval gate before tool execution;
emit OrchestrationPausedEvent/ApprovalRequestedEvent, await decision,
return rejection as ERROR context entry so LLM sees the denial;
propagate tool ERROR entries as StageExecutionResult.Failure
- SandboxedToolExecutor: remove dead code and simplify
- InfrastructureModule: minor wiring cleanup
- LlamaCppInferenceProvider / build.gradle: related build fixes
New core:artifacts-store interface + infrastructure/artifacts-cas
implementation: segment files + SQLite index, Blake3 hashing,
group-commit fsync via flushBefore, recovery tail-scan, manual
compactor, and oldest-first disk-cap evictor.
Inference events now carry promptArtifactId / responseArtifactId;
orchestrators put bytes before emitting. SqliteEventStore wraps
its txn in artifactStore.flushBefore so segment data is fsynced
before the event commit, making the crash window non-corrupting
(TailScanner re-indexes orphan tail records on reopen).
Compactor and evictor are mutually exclusive via maintenanceMutex.
Step 9 (cloud sync) deferred to a later epic.
See docs/reviews/2026-05-18-cas-steps-1-8-review.md for the final
review.
kami