agents-relay 2.0.156 → 2.0.158

package/workflow/task-lifecycle.mmd

@@ -11,7 +11,13 @@ blocked["state: BLOCKED"]
 succeeded["state: SUCCEEDED"]
 failed["state: FAILED"]
 cancelled["state: CANCELLED"]
-platformFallbackSafe["guard: platformFallbackSafe"]
+draftBlockAllowed["guard: troubleshooterTask"]
+queuedReviewBlockAllowed["guard: troubleshooterTask"]
+queuedBlockAllowed["guard: troubleshooterTask"]
+runningBlockAllowed["guard: troubleshooterTask"]
+unblockAllowed["guard: troubleshooterTask"]
+interactiveWaitAllowed["guard: awaitingInteractiveDecision"]
+recoverBlockAllowed["guard: troubleshooterTask"]
 
 depSat["guard: dependenciesSatisfied"]
 depUnsat["guard: dependenciesUnsatisfied"]
@@ -28,14 +34,9 @@ recoveryReopenClearReason["action: mutation.clearFailureReason"]
 recoveryReopenClearLease["action: mutation.clearLease"]
 recoveryReopenClearExecution["action: mutation.clearExecution"]
 
-retryAvailableRunning["guard: retryAvailable"]
 plannerRetryAvailable["guard: retryAvailable"]
 plannerUnavailableMark["action: mutation.markPlannerUnavailable"]
-orphanRetryAvailable["guard: retryAvailable"]
-orphanRetryClearError["action: mutation.clearError"]
-orphanRetryClearLease["action: mutation.clearLease"]
-orphanRetryClearExecution["action: mutation.clearExecution"]
-orphanFailMark["action: mutation.markStartupOrphanedExhausted"]
+orphanFailMark["action: mutation.markUnknownOutcome"]
 orphanFailClearLease["action: mutation.clearLease"]
 orphanFailClearExecution["action: mutation.clearExecution"]
 plannerRetryClearError["action: mutation.clearError"]
@@ -43,17 +44,10 @@ plannerRetryClearLease["action: mutation.clearLease"]
 plannerRetryClearExecution["action: mutation.clearExecution"]
 plannerFailClearLease["action: mutation.clearLease"]
 plannerFailClearExecution["action: mutation.clearExecution"]
-retryAvailableQueued["guard: retryAvailable"]
-leaseRetryRunningClearError["action: mutation.clearError"]
-leaseRetryRunningClearLease["action: mutation.clearLease"]
-leaseRetryRunningClearExecution["action: mutation.clearExecution"]
-leaseFailRunningMark["action: mutation.markLeaseExpired"]
+leaseFailRunningMark["action: mutation.markUnknownOutcome"]
 leaseFailRunningClearLease["action: mutation.clearLease"]
 leaseFailRunningClearExecution["action: mutation.clearExecution"]
-leaseRetryQueuedClearError["action: mutation.clearError"]
-leaseRetryQueuedClearLease["action: mutation.clearLease"]
-leaseRetryQueuedClearExecution["action: mutation.clearExecution"]
-leaseFailQueuedMark["action: mutation.markLeaseExpired"]
+leaseFailQueuedMark["action: mutation.markUnknownOutcome"]
 leaseFailQueuedClearLease["action: mutation.clearLease"]
 leaseFailQueuedClearExecution["action: mutation.clearExecution"]
 
@@ -80,8 +74,6 @@ completeClearError["action: mutation.clearError"]
 completeClearReason["action: mutation.clearFailureReason"]
 unblockRetryClearError["action: mutation.clearError"]
 unblockRetryClearReason["action: mutation.clearFailureReason"]
-unblockDraftClearError["action: mutation.clearError"]
-unblockDraftClearReason["action: mutation.clearFailureReason"]
 failedRetryClearError["action: mutation.clearError"]
 failedRetryClearReason["action: mutation.clearFailureReason"]
 recoverStartClearError["action: mutation.clearError"]
@@ -97,7 +89,9 @@ draft -->|on: reconcile| depSat
 depSat -->|true| reconcileDraftClearError
 reconcileDraftClearError --> reconcileDraftClearReason
 reconcileDraftClearReason --> queued
-draft -->|on: block| blocked
+draft -->|on: block| draftBlockAllowed
+draftBlockAllowed -->|true| blocked
+draftBlockAllowed -->|false| failed
 draft -->|on: cancel| cancelled
 
 queued -->|on: hold| draft
@@ -122,35 +116,41 @@ reviewApproveQueuedClearReason --> succeeded
 queued -->|on: review.changes_required| reviewChangesQueuedClearError
 reviewChangesQueuedClearError --> reviewChangesQueuedClearReason
 reviewChangesQueuedClearReason --> succeeded
-queued -->|on: review.blocked| blocked
-queued -->|on: block| blocked
+queued -->|on: review.blocked| queuedReviewBlockAllowed
+queuedReviewBlockAllowed -->|true| blocked
+queuedReviewBlockAllowed -->|false| failed
+queued -->|on: block| queuedBlockAllowed
+queuedBlockAllowed -->|true| blocked
+queuedBlockAllowed -->|false| failed
 queued -->|on: fail| failed
 queued -->|on: effect.launchTask.failed| failed
 queued -->|on: watchdog.failed| failed
 queued -->|on: worker.delivery_failed| failed
 queued -->|on: worker.start_ack_failed| failed
-queued -->|on: worker.platform_failure| platformFallbackSafe
-platformFallbackSafe -->|true| platformFallbackClearFailure
-platformFallbackClearFailure["action: mutation.preparePlatformFallback"]
-platformFallbackClearFailure --> queued
-platformFallbackSafe -->|false| platformFallbackQuarantine
+queued -->|on: worker.platform_failure| platformFallbackQuarantine
 platformFallbackQuarantine["action: mutation.quarantinePlatformFailure"]
-platformFallbackQuarantine --> blocked
+platformFallbackQuarantine --> failed
 queued -->|on: planner.unavailable| plannerUnavailableMark
 plannerUnavailableMark --> failed
-queued -->|on: execution.orphaned| orphanRetryAvailable
-queued -->|on: lease.expired| retryAvailableQueued
-retryAvailableQueued -->|true| leaseRetryQueuedClearError
-leaseRetryQueuedClearError --> leaseRetryQueuedClearLease
-leaseRetryQueuedClearLease --> leaseRetryQueuedClearExecution
-leaseRetryQueuedClearExecution --> queued
-retryAvailableQueued -->|false| leaseFailQueuedMark
+queued -->|on: execution.orphaned| orphanFailMark
+orphanFailMark --> orphanFailClearLease
+orphanFailClearLease --> orphanFailClearExecution
+orphanFailClearExecution --> failed
+queued -->|on: lease.expired| leaseFailQueuedMark
 leaseFailQueuedMark --> leaseFailQueuedClearLease
 leaseFailQueuedClearLease --> leaseFailQueuedClearExecution
 leaseFailQueuedClearExecution --> failed
 queued -->|on: cancel| cancelled
 
-running -->|on: task.blocked| blocked
+running -->|on: interactive.waiting| interactiveWaitAllowed
+interactiveWaitAllowed -->|true| runningWaitClearLease
+runningWaitClearLease["action: mutation.clearLease"]
+runningWaitClearLease --> runningWaitClearExecution
+runningWaitClearExecution["action: mutation.clearExecution"]
+runningWaitClearExecution --> draft
+running -->|on: task.blocked| runningBlockAllowed
+runningBlockAllowed -->|true| blocked
+runningBlockAllowed -->|false| failed
 running -->|on: task.progress| running
 running -->|on: review.approved| reviewApproveRunningClearError
 reviewApproveRunningClearError --> reviewApproveRunningClearReason
@@ -158,25 +158,17 @@ reviewApproveRunningClearReason --> succeeded
 running -->|on: review.changes_required| reviewChangesRunningClearError
 reviewChangesRunningClearError --> reviewChangesRunningClearReason
 reviewChangesRunningClearReason --> succeeded
-running -->|on: review.blocked| blocked
+running -->|on: review.blocked| runningBlockAllowed
 running -->|on: task.completed| completeClearError
 completeClearError --> completeClearReason
 completeClearReason --> succeeded
 running -->|on: task.failed| failed
 running -->|on: worker.platform_failure| runningPlatformFailureQuarantine
 runningPlatformFailureQuarantine["action: mutation.quarantinePlatformFailure"]
-runningPlatformFailureQuarantine --> blocked
+runningPlatformFailureQuarantine --> failed
 running -->|on: watchdog.failed| failed
 running -->|on: planner.failed| plannerRetryAvailable
-running -->|on: execution.orphaned| orphanRetryAvailable
-orphanRetryAvailable -->|true| orphanRetryClearError
-orphanRetryClearError --> orphanRetryClearLease
-orphanRetryClearLease --> orphanRetryClearExecution
-orphanRetryClearExecution --> queued
-orphanRetryAvailable -->|false| orphanFailMark
-orphanFailMark --> orphanFailClearLease
-orphanFailClearLease --> orphanFailClearExecution
-orphanFailClearExecution --> failed
+running -->|on: execution.orphaned| orphanFailMark
 plannerRetryAvailable -->|true| plannerRetryClearError
 plannerRetryClearError --> plannerRetryClearLease
 plannerRetryClearLease --> plannerRetryClearExecution
@@ -184,12 +176,7 @@ plannerRetryClearExecution --> queued
 plannerRetryAvailable -->|false| plannerFailClearLease
 plannerFailClearLease --> plannerFailClearExecution
 plannerFailClearExecution --> failed
-running -->|on: lease.expired| retryAvailableRunning
-retryAvailableRunning -->|true| leaseRetryRunningClearError
-leaseRetryRunningClearError --> leaseRetryRunningClearLease
-leaseRetryRunningClearLease --> leaseRetryRunningClearExecution
-leaseRetryRunningClearExecution --> queued
-retryAvailableRunning -->|false| leaseFailRunningMark
+running -->|on: lease.expired| leaseFailRunningMark
 leaseFailRunningMark --> leaseFailRunningClearLease
 leaseFailRunningClearLease --> leaseFailRunningClearExecution
 leaseFailRunningClearExecution --> failed
@@ -197,12 +184,9 @@ running -->|on: task.cancelled| cancelled
 running -->|on: pause| queued
 running -->|on: cancel| cancelled
 
-autonomousJob["guard: autonomousJob"]
-blocked -->|on: task.unblocked| autonomousJob
-autonomousJob -->|true| unblockDraftClearError
-unblockDraftClearError --> unblockDraftClearReason
-unblockDraftClearReason --> draft
-autonomousJob -->|false| unblockRetryClearError
+blocked -->|on: task.unblocked| unblockAllowed
+unblockAllowed -->|true| unblockRetryClearError
+unblockAllowed -->|false| blocked
 unblockRetryClearError --> unblockRetryClearReason
 unblockRetryClearReason --> queued
 blocked -->|on: cancel| cancelled
@@ -220,7 +204,9 @@ recoverComplete -->|true| recoverCompleteClearError
 recoverCompleteClearError --> recoverCompleteClearReason
 recoverCompleteClearReason --> succeeded
 failed -->|on: task.blocked| recoverBlocked
-recoverBlocked -->|true| blocked
+recoverBlocked -->|true| recoverBlockAllowed
+recoverBlockAllowed -->|true| blocked
+recoverBlockAllowed -->|false| failed
 failed -->|on: task.failed| recoverFailed
 recoverFailed -->|true| failed
 failed -->|on: task.cancelled| recoverCancelled

package/workflow/task-resolution.mmd

@@ -7,27 +7,39 @@ flowchart LR
 unresolved["state: unresolved"]
 checkOwner["state: check_owner"]
 checkPath["state: check_path"]
+checkFreshExecution["state: check_fresh_execution"]
 activeRecovery["action: resolution.activateRecovery"]
 sameJob["action: resolution.attachSameJobDependency"]
 blockerJob["action: resolution.registerBlockerDependency"]
 escalate["action: resolution.createEscalationGate"]
 obsolete["action: resolution.cancelSupersede"]
 resolved["state: resolved"]
+rejected["state: rejected"]
 recoveryOwner["guard: recoveryOwnerAvailable"]
 plannerOwner["guard: plannerOwnerAvailable"]
 exact["guard: exactOriginalRetry"]
 same["guard: sameJobDependency"]
+fresh["guard: freshRecoveryExecution"]
 blocker["guard: blockerJobDependency"]
 gate["guard: escalationGate"]
 obsoleteGuard["guard: genuinelyObsolete"]
+classifiable["guard: classifiableFailure"]
+classificationKind["guard: classificationSuperseded"]
+classifySuperseded["action: resolution.classifySuperseded"]
+classifyInvalid["action: resolution.classifyInvalid"]
+supersededEvidence["guard: supersededEvidenceComplete"]
+invalidEvidence["guard: invalidEvidenceComplete"]
 
 unresolved -->|on: resolution.propose| checkOwner
 checkOwner -->|on: resolution.propose| recoveryOwner
 recoveryOwner -->|true| checkPath
 recoveryOwner -->|false| plannerOwner
 plannerOwner -->|true| checkPath
-plannerOwner -->|false| resolved
-checkPath -->|on: resolution.propose| exact
+plannerOwner -->|false| rejected
+checkPath -->|on: resolution.propose| fresh
+fresh -->|true| checkFreshExecution
+fresh -->|false| rejected
+checkFreshExecution -->|on: resolution.propose| exact
 exact -->|true| activeRecovery
 exact -->|false| same
 same -->|true| sameJob
@@ -37,9 +49,18 @@ blocker -->|false| gate
 gate -->|true| escalate
 gate -->|false| obsoleteGuard
 obsoleteGuard -->|true| obsolete
-obsoleteGuard -->|false| resolved
+obsoleteGuard -->|false| rejected
 activeRecovery -->|to: resolved| resolved
 sameJob -->|to: resolved| resolved
 blockerJob -->|to: resolved| resolved
 escalate -->|to: resolved| resolved
 obsolete -->|to: resolved| resolved
+unresolved -->|on: resolution.classify| classifiable
+classifiable -->|true| classificationKind
+classifiable -->|false| rejected
+classificationKind -->|true| supersededEvidence
+classificationKind -->|false| invalidEvidence
+supersededEvidence -->|true| classifySuperseded
+invalidEvidence -->|true| classifyInvalid
+classifySuperseded --> resolved
+classifyInvalid --> resolved

package/workflow/troubleshooter.mmd

@@ -1,6 +1,6 @@
 ---
 title: Troubleshooter
-description: Coordinates diagnosis, recovery, escalation, and resolution of failed Tasks.
+description: Re-runs diagnosis after authoritative unblock, requires fresh resolution evidence, then authorizes recovery of FAILED work.
 ---
 flowchart LR
 %% relay: initial idle
@@ -12,6 +12,9 @@ checkId["state: check_id"]
 checkBarrier["state: check_barrier"]
 skip["state: skip"]
 create["state: create"]
+resume["state: resumed_after_authoritative_unblock"]
+freshProposal["state: fresh_resolution_proposed"]
+failClosed["state: remain_failed_without_original_mutation"]
 
 failed["guard: failedWork"]
 ordinary["guard: ordinaryWork"]
@@ -20,8 +23,13 @@ needsRecovery["guard: needsRecovery"]
 idUnused["guard: idUnused"]
 noBarrier["guard: noRecoveryBarrier"]
 createAction["action: troubleshooter.create"]
+verifyFresh["guard: freshConditionVerification"]
+proposeResolution["action: troubleshooter.proposeFreshResolution"]
+resolutionFresh["guard: resolutionForCurrentExecution"]
+actOnOriginal["action: troubleshooter.applyAuthorizedResolution"]
 
 idle -->|on: troubleshooter.reconcile| failed
+idle -->|on: troubleshooter.resumed| resume
 failed -->|false| skip
 failed -->|true| checkOrdinary
 
@@ -45,3 +53,14 @@ checkBarrier -->|on: troubleshooter.reconcile| noBarrier
 noBarrier -->|false| skip
 noBarrier -->|true| createAction
 createAction --> create
+
+%% A human response only queues this same Troubleshooter. It must start a new
+%% execution, reverify the condition, and create a new resolution before acting.
+resume -->|on: task.started| verifyFresh
+verifyFresh -->|true| proposeResolution
+verifyFresh -->|false| failClosed
+proposeResolution --> freshProposal
+freshProposal -->|on: resolution.propose| resolutionFresh
+resolutionFresh -->|true| actOnOriginal
+resolutionFresh -->|false| failClosed
+actOnOriginal --> resume