Summary
Ensures every unit of work at every level emits a CaseLedgerEntry node so causedByEntryId chains are traversable end-to-end. Introduces SPAWN_STARTED / SPAWN_COMPLETED as unified ledger event types replacing the per-model family of SUBCASE_STARTED, WORKFLOW_STEP_DISPATCHED, PLAN_STEP_STARTED, etc.
Consolidation — unified SPAWN event types
All spawning operations share identical ledger structure: parentLedgerEntryId, childId, spawnType, returnTo, policy. Instead of N event types with per-type metadata:
CaseHubEventType.SPAWN_STARTED // replaces SUBCASE_STARTED, WORKFLOW_STEP_DISPATCHED,
// PLAN_STEP_STARTED, AGENT_INVOCATION_STARTED,
// WORKITEM_CHILD_CREATED
CaseHubEventType.SPAWN_COMPLETED // replaces SUBCASE_COMPLETED, etc.
Standard metadata fields (all SPAWN_STARTED entries carry these):
spawnType: SUBCASE | WORKFLOW_STEP | PLAN_STEP | AGENT_INVOCATION | WORKITEM_CHILD | WORKITEM_GROUPchildId — what was spawnedreturnTo — where output routes backpolicyType — ALL | M_OF_N | FIRST | ANYpolicyThreshold — for M_OF_NgroupId — groups parallel spawns (same value for all children in a multi-instance group)
Lineage traversal becomes uniform: every node is SPAWN_STARTED / SPAWN_COMPLETED, regardless of what was spawned. findCausalChain() queries one event type, not N.
Current partial state
SpawnRecord
CompletionTracker (epic #204) creates a SpawnRecord for each SPAWN_STARTED entry:
record SpawnRecord(
UUID spawnId,
UUID parentLedgerEntryId,
SpawnType type,
List<UUID> childIds,
Object returnTo,
CompletionPolicy policy,
int threshold
) {}The SpawnRecord is the runtime counterpart to the ledger entry — the tracker holds it in memory; the ledger entry is the durable record.
Lineage query API
interface LineageRepository {
CausalGraph findCausalChain(UUID rootEntryId, int maxDepth);
List<CaseLedgerEntry> findDirectChildren(UUID spawnStartedEntryId);
List<CaseLedgerEntry> findRootCauses(UUID caseId);
List<CaseLedgerEntry> findByGroup(UUID groupId); // all children in a multi-instance group
}quarkus-work multi-instance
Part of
#201
Summary
Ensures every unit of work at every level emits a
CaseLedgerEntrynode socausedByEntryIdchains are traversable end-to-end. IntroducesSPAWN_STARTED/SPAWN_COMPLETEDas unified ledger event types replacing the per-model family of SUBCASE_STARTED, WORKFLOW_STEP_DISPATCHED, PLAN_STEP_STARTED, etc.Consolidation — unified SPAWN event types
All spawning operations share identical ledger structure:
parentLedgerEntryId,childId,spawnType,returnTo,policy. Instead of N event types with per-type metadata:Standard metadata fields (all SPAWN_STARTED entries carry these):
spawnType: SUBCASE | WORKFLOW_STEP | PLAN_STEP | AGENT_INVOCATION | WORKITEM_CHILD | WORKITEM_GROUPchildId— what was spawnedreturnTo— where output routes backpolicyType— ALL | M_OF_N | FIRST | ANYpolicyThreshold— for M_OF_NgroupId— groups parallel spawns (same value for all children in a multi-instance group)Lineage traversal becomes uniform: every node is
SPAWN_STARTED/SPAWN_COMPLETED, regardless of what was spawned.findCausalChain()queries one event type, not N.Current partial state
SpawnRecord
CompletionTracker(epic #204) creates aSpawnRecordfor eachSPAWN_STARTEDentry:The
SpawnRecordis the runtime counterpart to the ledger entry — the tracker holds it in memory; the ledger entry is the durable record.Lineage query API
quarkus-work multi-instance
completedCountandthresholdin metadataPart of
#201