Interface: LoopContinuationFrame
Defined in: ports/loop-continuation-frame.ts:26
Serialized continuation frame for an iterative P.loop execution (kind: 'loop-frame').
Preserves cumulative telemetry, .act()-only tripwire counters, loop-scoped capability leases (grantedLoopLeases), active skills, and deduplicated lesson outcome attribution across iterations, cross-model escalations, and fork suspensions.
Properties
| Property | Modifier | Type | Description | Defined in |
|---|---|---|---|---|
accumulatedLessonOutcomes | readonly | Readonly<Record<string, LessonOutcome>> | Per-loop deduplicated lesson outcome map (lessonId -> worst 4-state outcome observed). | ports/loop-continuation-frame.ts:82 |
actCostUsd | readonly | number | Cumulative .act()-only USD cost evaluated against the .act() actor's P.policy.maxCostUsd. | ports/loop-continuation-frame.ts:58 |
actElapsedDurationMs | readonly | number | Cumulative .act()-only elapsed duration in milliseconds evaluated against P.policy.maxDurationSeconds. | ports/loop-continuation-frame.ts:60 |
activeSkillActivations | readonly | Readonly<Record<string, { callId: string; messageId: string; }>> | Map of .act() active skillId to its { messageId, callId } activation coordinates. | ports/loop-continuation-frame.ts:66 |
activeSkillIds | readonly | readonly string[] | Active skill IDs mounted by the .act() actor across iterations. | ports/loop-continuation-frame.ts:64 |
actTurnsTaken | readonly | number | Cumulative .act()-only LLM turns evaluated against the .act() actor's P.policy.maxTurns. | ports/loop-continuation-frame.ts:56 |
bufferedGateRecoveries? | readonly | readonly unknown[] | Buffered gate recovery pairs awaiting .evaluate() acceptance before distillation. | ports/loop-continuation-frame.ts:80 |
conflictCheckArmed? | readonly | boolean | Whether scopeNode.conflictCleanCheck is armed on the loop's scope handle. | ports/loop-continuation-frame.ts:74 |
costUsd | readonly | number | Cumulative total USD cost across both .act() and .evaluate() iterations. | ports/loop-continuation-frame.ts:52 |
elapsedDurationMs | readonly | number | Cumulative total elapsed duration in milliseconds across the loop. | ports/loop-continuation-frame.ts:54 |
grantedLoopLeases | readonly | readonly LeaseDelta[] | Loop-scoped capability leases inherited across all .act() iterations. | ports/loop-continuation-frame.ts:62 |
injectedScopedRuleIds | readonly | readonly string[] | Identifiers of ScopedRuleNode directives injected into the .act() actor across iterations. | ports/loop-continuation-frame.ts:72 |
iteration | readonly | number | Zero-based current iteration index (0..maxIterations - 1). | ports/loop-continuation-frame.ts:40 |
kind | readonly | "loop-frame" | Discriminant identifying a loop continuation frame. | ports/loop-continuation-frame.ts:28 |
lastActModelId? | readonly | string | Provider model identifier used on the most recent .act() iteration. | ports/loop-continuation-frame.ts:34 |
lastActOutputJson? | readonly | string | JSON-serialized .act() output persisted when transitioning to phase: 'evaluate'. | ports/loop-continuation-frame.ts:46 |
lastActScopeReceiptJson? | readonly | string | JSON-serialized scope receipt persisted when transitioning to phase: 'evaluate'. | ports/loop-continuation-frame.ts:48 |
lastRejectedIterationJson? | readonly | string | Optional JSON-serialized snapshot of the most recent rejected iteration for .onLoopRecovery(). | ports/loop-continuation-frame.ts:78 |
loopId | readonly | string | Logical identifier of the LoopNode. | ports/loop-continuation-frame.ts:30 |
loopInputJson | readonly | string | JSON-serialized validated input payload for the loop. | ports/loop-continuation-frame.ts:38 |
observedEffects | readonly | readonly ToolEffect[] | Cumulative ToolEffect descriptors recorded across the loop. | ports/loop-continuation-frame.ts:70 |
phase | readonly | "act" | "evaluate" | Active phase within the current iteration ('act' vs 'evaluate'). | ports/loop-continuation-frame.ts:42 |
preservedActLeafMessageId | readonly | string | null | Leaf message ID of the .act() actor's structured-output completion turn. | ports/loop-continuation-frame.ts:44 |
recalledLessons | readonly | readonly RecalledLoopLessonRecord[] | Active lessons recalled at loop start and forwarded to .evaluate(). | ports/loop-continuation-frame.ts:76 |
roleKey | readonly | string | Canonical roleKey of the loop's .act() actor. | ports/loop-continuation-frame.ts:32 |
stepExecId | readonly | string | Unique step execution identifier in LedgerStore. | ports/loop-continuation-frame.ts:36 |
turnsTaken | readonly | number | Cumulative total LLM turns across both .act() and .evaluate() iterations. | ports/loop-continuation-frame.ts:50 |