@tangle-network/agent-runtime 0.142.0 → 0.142.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/agent.js +1 -1
- package/dist/index.js +2 -2
- package/dist/kernel.js +1 -1
- package/dist/{loop-runner-bin-B18hKrB1.js → loop-runner-bin-DKEzvd7f.js} +2 -2
- package/dist/{loop-runner-bin-B18hKrB1.js.map → loop-runner-bin-DKEzvd7f.js.map} +1 -1
- package/dist/loop-runner-bin.js +1 -1
- package/dist/mcp/index.js +1 -1
- package/dist/{runtime-DCU6DOqI.js → runtime-WOQ24LYd.js} +10 -19
- package/dist/runtime-WOQ24LYd.js.map +1 -0
- package/dist/testing.js +8 -8
- package/package.json +1 -1
- package/dist/runtime-DCU6DOqI.js.map +0 -1
package/dist/agent.js
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
import { o as mapSandboxEvent } from "./sandbox-events-UKNxOsgK.js";
|
|
2
2
|
import { $t as AGENT_PROFILE_MATERIALIZATION_AXES, an as promptControlProfileMaterialization, cn as promptResourceProfileMaterialization, dn as validateProfileMaterialization, en as assertProfileMaterialization, fn as worktreeCliProfileMaterialization, in as profileMaterializationAxes, ln as renderProfileMaterializationIssues, lt as createSandboxForSpec, nn as defineProfileMaterializationContract, on as promptModelProfileMaterialization, rn as fullProfileMaterialization, sn as promptOnlyProfileMaterialization, tn as controlProfileMaterialization, un as sandboxActProfileMaterialization } from "./supervisor-BO7B2eim.js";
|
|
3
|
-
import "./runtime-
|
|
3
|
+
import "./runtime-WOQ24LYd.js";
|
|
4
4
|
import { createHash } from "node:crypto";
|
|
5
5
|
import { isAbsolute, join, relative, resolve, sep } from "node:path";
|
|
6
6
|
import { existsSync, readFileSync, statSync } from "node:fs";
|
package/dist/index.js
CHANGED
|
@@ -9,10 +9,10 @@ import { At as exportEvalRuns, Bt as sanitizeKnowledgeReadinessReport, Dt as bui
|
|
|
9
9
|
import { i as notifyRuntimeHookEvent, n as defineRuntimeHooks, r as notifyRuntimeDecisionPoint, t as composeRuntimeHooks } from "./runtime-hooks-tXpAarhW.js";
|
|
10
10
|
import { D as optimizerMethod, O as strategyAuthorMethod } from "./structural-rollout-Bncu0lF7.js";
|
|
11
11
|
import { B as ROLLOUT_POLICY_EXTENSION, G as structuralRolloutPolicyFromProfile, H as normalizeRolloutPolicy, J as defaultBuildPrompt, K as agenticGenerator, L as improve, R as withMethodRuntimeControls, U as parseRolloutPolicy, V as applyRolloutPolicyToProfile, W as serializeRolloutPolicy, q as commandVerifier, z as rawTraceDistiller } from "./improvement-cycle-Bk4Lw_Rq.js";
|
|
12
|
-
import { Bt as connectStdioMcp, zt as McpSpawnFault } from "./runtime-
|
|
12
|
+
import { Bt as connectStdioMcp, zt as McpSpawnFault } from "./runtime-WOQ24LYd.js";
|
|
13
13
|
import { n as defaultRedactorIdentityMaterial, r as resolveRedactor, t as defaultRedactor } from "./redact-D-u-rrcn.js";
|
|
14
14
|
import { a as createSupervisedKnowledgeUpdater, c as runSupervisedKnowledgeUpdate, i as RESEARCH_SUPERVISOR_SYSTEM_PROMPT, l as createKnowledgeImprovementActivationExecutor, n as createAgentKnowledgeReadinessCheck, o as formatSupervisedKnowledgeTask, r as runKnowledgeImprovementJob, s as knowledgeReadinessDeliverable, t as buildKnowledgeImprovementExperimentBundles } from "./knowledge-Cf4HsD2i.js";
|
|
15
|
-
import { a as isDelegatedLoopMode, c as worktreeLoopRunner, i as auditLoopRunner, n as runLoopRunnerCli, o as researchLoopRunner, r as DELEGATED_LOOP_MODES, s as runDelegatedLoop, t as parseLoopRunnerArgv } from "./loop-runner-bin-
|
|
15
|
+
import { a as isDelegatedLoopMode, c as worktreeLoopRunner, i as auditLoopRunner, n as runLoopRunnerCli, o as researchLoopRunner, r as DELEGATED_LOOP_MODES, s as runDelegatedLoop, t as parseLoopRunnerArgv } from "./loop-runner-bin-DKEzvd7f.js";
|
|
16
16
|
import { n as mcpToolsForRuntimeMcpSubset, t as mcpToolsForRuntimeMcp } from "./openai-tools-DtMJcw3r.js";
|
|
17
17
|
import { agentProfileSchema, canonicalAgentProfileDigest } from "@tangle-network/agent-interface";
|
|
18
18
|
import { FAILURE_CLASSES, acquisitionPlansForKnowledgeGaps, blockingKnowledgeEval, canonicalJson, runAgentControlLoop, scoreKnowledgeReadiness, userQuestionsForKnowledgeGaps } from "@tangle-network/agent-eval";
|
package/dist/kernel.js
CHANGED
|
@@ -5,7 +5,7 @@ import { n as assertModelAllowed, r as assertProfileModelsAllowed } from "./mode
|
|
|
5
5
|
import { $ as createInbox, A as createBudgetPool, C as effectiveConcurrency, E as rollingDispatch, F as cliWorktreeExecutor, G as DEFAULT_SANDBOX_STEERING_MAX_TURNS, H as workerTraceEnv, I as createExecutor, J as decodeToolPart, K as createSteerableSandboxSession, L as createExecutorRegistry, Ot as createOpenInferenceFileExporter, Q as readWorkerProgress, S as waitUntil, T as queueOf, U as workerTraceHeaders, V as readWorkerTraceContext, W as workerTraceSeamKey, X as DEFAULT_STALL_AFTER_MS, Y as sandboxSessionTraceSource, Z as createActivityLog, _ as createWaitProbes, a as pickBestDelivered, at as isPeerMailEnvelope, b as timerAt, ct as TERMINAL_DECISIONS, dt as isTerminalDecision, et as AUTHORITY_MARKERS, ft as runAgentRounds, g as settledToIteration, ht as acquireSandbox, i as collectDelivered, it as createPeerMailbox, j as spendFromUsageEvents, kt as createOtelExporter, mt as probeSandboxCapabilities, n as createSupervisor, nt as PEER_MAIL_WIRE_KEY, o as runFinalizer, ot as peerMailTools, pt as createSandboxLineage, q as createPushTraceSource, r as bestDelivered, rt as claimsAuthority, st as peerMailVerbNames, t as createRootHandle, tt as DEFAULT_PEER_MAIL_LIMITS, u as createScope, ut as defaultSelectWinner, v as isWaitOutcome, w as freeSlots, x as validateWaitSpec, y as pollFor, z as createWorktreeCliExecutor } from "./supervisor-BO7B2eim.js";
|
|
6
6
|
import { a as sandboxClientAsProvider, i as resolveAgentEnvironmentProvider, n as providerAsExecutor, o as createTangleSandboxExactProcessProvider, r as providerAsSandboxClient, t as createAgentEnvironmentProviderRegistry } from "./environment-provider-D1qUgO6B.js";
|
|
7
7
|
import { C as observe, E as profileOptimizerModelCall, S as defaultAnalystInstruction, T as profileChatClient, _ as depthStrategy, a as defaultStructuralRolloutPolicy, b as sample, c as officialChecksFromMeta, d as selectBestIndex, f as structuralRollout, g as defineStrategy, h as breadthStrategy, i as defaultExtractCandidate, l as resolveEntrySymbol, m as adaptiveRefine, n as compareCheckOutcomes, o as filterAuthoredAsserts, p as visibleCheckScore, r as composeCheckSources, s as modelAuthoredChecks, t as canDisplace, u as sandboxCheckRunner, v as refine, w as renderReport, x as sampleThenRefine, y as runAgentic } from "./structural-rollout-Bncu0lF7.js";
|
|
8
|
-
import { $ as runPersonified, A as pickChampion, At as stopSentinel, B as runBenchmark, Bt as connectStdioMcp, C as analyzeTrace, Ct as inlineSandboxClient, D as chatWorkerSeam, Dt as completionAuthorizes, E as chatTransportExecutor, Et as superviseDispatch, F as strategyAuthorContract, Ft as renderLeaderboardSvg, G as reconnectRetainedRun, Gt as mcpSecretEnvMetadataKey, H as reconnectRetainedInteractiveRun, Ht as createMcpEnvironment, I as strategyAuthorSystemPrompt, It as renderPairwiseMarkdown, J as startRetainedRunInEnvironment, Jt as secretEnvOfMcpServer, K as recoverRetainedRun, Kt as resolveMcpServerLaunch, L as SandboxRunAbortError, Lt as auditIntent, M as selectChampion, Mt as pairwiseSignificance, N as assertStrategyContract, Nt as renderLeaderboardHtml, O as createChatSessionStore, Ot as deterministicCompletion, P as authorStrategy, Pt as renderLeaderboardMarkdown, Q as definePersona, R as openSandboxRun, Rt as defaultAuditorInstruction, S as withUntrackedArtifacts, St as localSandboxClient, Tt as loopDispatch, U as recoverRetainedInteractiveRun, Ut as sanitizeMcpToolSchema, V as claimRetainedInteractiveControl, Vt as materializeLocalMcp, W as startRetainedInteractiveRun, Wt as envKeyProvider, X as equalKOnCost, Y as promotionGate, Z as trajectoryReport, _ as VERIFY_TAIL_CHARS, _t as registryScopeAnalyst, a as localShell, at as renderCorpusToInstructions, b as settledWorkerOut, bt as defineLeaderboard, c as createVerifierEnvironment, ct as loopUntil, d as harvestSurfaceDiffs, dt as selectValidWinner, et as builtinShapes, f as failuresAnalyst, ft as verify, g as NOTE_MAX_CHARS, gt as createScopeAnalyst, h as EVIDENCE_MAX_CHARS, ht as buildSteerContext, i as jjWorkspace, it as InMemoryCorpus, j as runStrategyEvolution, jt as leaderboard, k as discriminatingMeans, kt as sentinelCompletion, l as boxSurfaceReader, lt as panel, m as worktreeFanout, mt as assertTraceDerivedFindings, n as makeFinding, nt as registerShape, o as runInWorkspace, ot as fanout, p as superviseSurface, pt as widen, q as startRetainedRun, qt as resolveSecretEnv, r as gitWorkspace, rt as FileCorpus, s as createWaterfallCollector, st as flatWidenGate, t as computeFindingId, tt as createShapeRegistry, u as fsSurfaceReader, ut as pipeline, v as closingWorkerNote, vt as inProcessSandboxClient, w as patchDelivered, wt as loopCampaignDispatch, x as copyUntrackedIntoClone, xt as resolveSandboxClient, y as composeWorkerEvidence, yt as harvestCorpus, z as printBenchmarkReport, zt as McpSpawnFault } from "./runtime-
|
|
8
|
+
import { $ as runPersonified, A as pickChampion, At as stopSentinel, B as runBenchmark, Bt as connectStdioMcp, C as analyzeTrace, Ct as inlineSandboxClient, D as chatWorkerSeam, Dt as completionAuthorizes, E as chatTransportExecutor, Et as superviseDispatch, F as strategyAuthorContract, Ft as renderLeaderboardSvg, G as reconnectRetainedRun, Gt as mcpSecretEnvMetadataKey, H as reconnectRetainedInteractiveRun, Ht as createMcpEnvironment, I as strategyAuthorSystemPrompt, It as renderPairwiseMarkdown, J as startRetainedRunInEnvironment, Jt as secretEnvOfMcpServer, K as recoverRetainedRun, Kt as resolveMcpServerLaunch, L as SandboxRunAbortError, Lt as auditIntent, M as selectChampion, Mt as pairwiseSignificance, N as assertStrategyContract, Nt as renderLeaderboardHtml, O as createChatSessionStore, Ot as deterministicCompletion, P as authorStrategy, Pt as renderLeaderboardMarkdown, Q as definePersona, R as openSandboxRun, Rt as defaultAuditorInstruction, S as withUntrackedArtifacts, St as localSandboxClient, Tt as loopDispatch, U as recoverRetainedInteractiveRun, Ut as sanitizeMcpToolSchema, V as claimRetainedInteractiveControl, Vt as materializeLocalMcp, W as startRetainedInteractiveRun, Wt as envKeyProvider, X as equalKOnCost, Y as promotionGate, Z as trajectoryReport, _ as VERIFY_TAIL_CHARS, _t as registryScopeAnalyst, a as localShell, at as renderCorpusToInstructions, b as settledWorkerOut, bt as defineLeaderboard, c as createVerifierEnvironment, ct as loopUntil, d as harvestSurfaceDiffs, dt as selectValidWinner, et as builtinShapes, f as failuresAnalyst, ft as verify, g as NOTE_MAX_CHARS, gt as createScopeAnalyst, h as EVIDENCE_MAX_CHARS, ht as buildSteerContext, i as jjWorkspace, it as InMemoryCorpus, j as runStrategyEvolution, jt as leaderboard, k as discriminatingMeans, kt as sentinelCompletion, l as boxSurfaceReader, lt as panel, m as worktreeFanout, mt as assertTraceDerivedFindings, n as makeFinding, nt as registerShape, o as runInWorkspace, ot as fanout, p as superviseSurface, pt as widen, q as startRetainedRun, qt as resolveSecretEnv, r as gitWorkspace, rt as FileCorpus, s as createWaterfallCollector, st as flatWidenGate, t as computeFindingId, tt as createShapeRegistry, u as fsSurfaceReader, ut as pipeline, v as closingWorkerNote, vt as inProcessSandboxClient, w as patchDelivered, wt as loopCampaignDispatch, x as copyUntrackedIntoClone, xt as resolveSandboxClient, y as composeWorkerEvidence, yt as harvestCorpus, z as printBenchmarkReport, zt as McpSpawnFault } from "./runtime-WOQ24LYd.js";
|
|
9
9
|
import { At as defaultToolDetectors, Ct as plateauLength, Et as canonicalFindingEvent, Ft as createSupervisorSpanRecorder, It as gateOnDeliverable, Lt as mapExecutorResult, Mt as createFileRunContext, Nt as createInMemoryRunContext, Ot as normalizeAnalyzeOnSettle, Pt as FileCoordinationLog, St as bestSoFar, Tt as DEFAULT_AWAIT_EVENT_TIMEOUT_MS, X as delegate, Y as defaultDelegateBudget, _ as promptHandle, _t as noProgressFor, a as assertCoordinationBinding, b as classifyDriverFailure, bt as anytimeReport, d as delegatesWorkerBriefPrompt, f as dumbContinuationFailPrompt, ft as finalizeBestDelivered, g as naiveContinuationPrompt, gt as createProgressTracker, h as kernelPromptRegistry, ht as anyOf, i as workerFromBackend, jt as watchTrace, kt as createEventBus, l as analyzesFindingsReportPrompt, m as formatPromptHandle, mt as allWorkersStalled, n as supervise, o as resolveSupervisorProfile, p as dumbContinuationPassPrompt, pt as allOf, s as supervisorAgent, t as DEFAULT_AUTHORED_PROFILE_SECURITY_POLICY, u as createPromptRegistry, v as supervisorPolicyPrompt, vt as plateau, wt as renderAnytimeTable, x as serveCoordinationMcp, xt as areaUnderCurve, y as DriverAttemptsExhaustedError, yt as sampleFromSettled } from "./supervise-BA2aQ8vU.js";
|
|
10
10
|
import { a as readWorkerCancellation, c as supervisorRunDir, d as workerCancelRequestsFile, f as workerCancellationFile, g as workerInboxFileFromEventDir, h as workerInboxFile, i as readWorkerCancelRequests, l as supervisorRunsRoot, m as workerControlLogFile, n as legacySupervisorRunDir, o as readWorkerSteerRequests, p as workerCancellationsDir, r as legacySupervisorRunsRoot, s as safeWorkerFile, t as cancelWorker, u as supervisorWorkersDir, v as writeWorkerSteer } from "./run-layout-Be_Z0dF8.js";
|
|
11
11
|
import { a as supervisorInstructions, i as profileRichnessFinding, n as assessAuthoredProfile, r as defaultProfileRichnessThresholds, t as asAuthoredProfile } from "./authoring-Y-OUn4rW.js";
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
import { i as ConfigError } from "./errors-CDZ8XsVj.js";
|
|
2
2
|
import { L as createExecutorRegistry } from "./supervisor-BO7B2eim.js";
|
|
3
|
-
import { $ as runPersonified, Q as definePersona, m as worktreeFanout } from "./runtime-
|
|
3
|
+
import { $ as runPersonified, Q as definePersona, m as worktreeFanout } from "./runtime-WOQ24LYd.js";
|
|
4
4
|
import { t as runAnalystLoop } from "./analyst-loop-xZWlVqS8.js";
|
|
5
5
|
import { t as createKbGate } from "./kb-gate-DpaSwXVx.js";
|
|
6
6
|
//#region src/loop-runner.ts
|
|
@@ -243,4 +243,4 @@ if (invokedScript && /loop-runner-bin\.(js|ts|mjs)$/.test(invokedScript)) main()
|
|
|
243
243
|
//#endregion
|
|
244
244
|
export { isDelegatedLoopMode as a, worktreeLoopRunner as c, auditLoopRunner as i, runLoopRunnerCli as n, researchLoopRunner as o, DELEGATED_LOOP_MODES as r, runDelegatedLoop as s, parseLoopRunnerArgv as t };
|
|
245
245
|
|
|
246
|
-
//# sourceMappingURL=loop-runner-bin-
|
|
246
|
+
//# sourceMappingURL=loop-runner-bin-DKEzvd7f.js.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"loop-runner-bin-B18hKrB1.js","names":[],"sources":["../src/loop-runner.ts","../src/loop-runner-bin.ts"],"sourcesContent":["/**\n *\n * `runDelegatedLoop` — the configured delegated loop-runner.\n *\n * One typed entrypoint a worker agent (or a scheduled routine) calls to run a\n * disciplined loop in a chosen MODE, over agent-runtime's hardened engines:\n *\n * code → build-in-a-loop on the GENERIC recursive path (worktreeLoopRunner: author one\n * `AgentProfile` per harness → worktree-CLI leaves → `patchDelivered` gate)\n * review → caller-registered runner — a `code` runner with an approval gate over candidates\n * research → research-in-a-loop with valid-only KB growth (createKbGate)\n * audit → analyze trace/run data → findings (runAnalystLoop, caller-wired)\n * self-improve → caller-registered `improve(profile, options)` run\n *\n * It is intentionally a thin façade: the value is that EVERY product reuses the\n * one hardened engine instead of forking delegation logic. The dispatcher owns\n * mode routing, timing, fail-loud on an unregistered mode, and a uniform result\n * shape; each mode's engine is a pre-configured runner in the registry (build it\n * with the factories below, or inject your own / a stub).\n *\n * @experimental\n */\n\nimport type { AgentProfile } from '@tangle-network/agent-interface'\nimport { runAnalystLoop } from './analyst-loop'\nimport type { RunAnalystLoopOpts, RunAnalystLoopResult } from './analyst-loop/types'\nimport { ConfigError } from './errors'\nimport { type CreateKbGateOptions, createKbGate, type FactCandidate } from './mcp/kb-gate'\nimport {\n type AuthoredHarness,\n type Budget,\n createExecutorRegistry,\n definePersona,\n runPersonified,\n type WinnerStrategy,\n type WorktreeFanoutOptions,\n type WorktreePatchArtifact,\n worktreeFanout,\n} from './runtime'\n\n/** All valid delegated-loop mode names — used for validation and CLI surfaces. @experimental */\nexport const DELEGATED_LOOP_MODES = ['code', 'review', 'research', 'audit', 'self-improve'] as const\n\n/** @experimental */\nexport type DelegatedLoopMode = (typeof DELEGATED_LOOP_MODES)[number]\n\n/** Type guard — returns true when `value` is a valid `DelegatedLoopMode` string. @experimental */\nexport function isDelegatedLoopMode(value: unknown): value is DelegatedLoopMode {\n return typeof value === 'string' && (DELEGATED_LOOP_MODES as readonly string[]).includes(value)\n}\n\n/** @experimental A pre-configured loop for one mode. Returns the mode's raw\n * output; the dispatcher wraps it in a {@link DelegatedLoopResult}. */\nexport type DelegatedLoopRunner<T = unknown> = (signal: AbortSignal) => Promise<T>\n\n/** @experimental Mode → configured runner. Partial: only register the modes a\n * given product/routine actually uses. */\nexport type DelegatedLoopRegistry = Partial<Record<DelegatedLoopMode, DelegatedLoopRunner>>\n\n/** @experimental Uniform result — never throws from a registered runner; a\n * thrown engine becomes `{ ok: false, error }` so a routine can record + move on. */\nexport interface DelegatedLoopResult<T = unknown> {\n mode: DelegatedLoopMode\n ok: boolean\n output?: T\n error?: string\n durationMs: number\n}\n\n/** @experimental */\nexport interface RunDelegatedLoopOptions {\n signal?: AbortSignal\n /** Clock override for deterministic tests. */\n now?: () => number\n}\n\n/**\n *\n * Dispatch a configured loop by mode. Fails loud (throws `ConfigError`) when no\n * runner is registered for the mode — a routine pointed at an unwired mode is a\n * config bug, not a silent no-op. A runner that throws is captured as\n * `{ ok: false }` so unattended runs record the failure rather than crash.\n *\n * @experimental\n */\nexport async function runDelegatedLoop<T = unknown>(\n mode: DelegatedLoopMode,\n registry: DelegatedLoopRegistry,\n options: RunDelegatedLoopOptions = {},\n): Promise<DelegatedLoopResult<T>> {\n const runner = registry[mode] as DelegatedLoopRunner<T> | undefined\n if (!runner) {\n throw new ConfigError(\n `runDelegatedLoop: no runner registered for mode '${mode}' (registered: ${\n Object.keys(registry).join(', ') || 'none'\n })`,\n )\n }\n const now = options.now ?? Date.now\n const signal = options.signal ?? new AbortController().signal\n const start = now()\n try {\n const output = await runner(signal)\n return { mode, ok: true, output, durationMs: now() - start }\n } catch (err) {\n return {\n mode,\n ok: false,\n error: err instanceof Error ? err.message : String(err),\n durationMs: now() - start,\n }\n }\n}\n\n/** @experimental Options for the local-repo `code` runner over the GENERIC recursive path. */\nexport interface WorktreeLoopRunnerOptions {\n /** Exact profile carried by the personified root that owns this fanout. */\n rootProfile: AgentProfile\n /** Absolute path to the local git checkout each worktree is cut from. */\n repoRoot: string\n /** The instruction handed to every authored harness (composed under each profile's systemPrompt). */\n taskPrompt: string\n /** The supervisor-authored harness profiles — one fanout item (one worktree-CLI leaf) each. */\n harnesses: ReadonlyArray<AuthoredHarness>\n /** Conserved budget pool bounding the fanout (equal-k holds by construction). */\n budget: Budget\n /** Shell command run in each worktree to derive the tests-PASS signal. */\n testCmd?: string\n /** Shell command run in each worktree to derive the typecheck-PASS signal. */\n typecheckCmd?: string\n /** Which verification signals the deliverable REQUIRES present-and-passing (default none). */\n require?: ReadonlyArray<'tests' | 'typecheck'>\n /** Diff-size cap (lines). */\n maxDiffLines?: number\n /** Literal path prefixes the patch must not touch (the secret-floor is always on regardless). */\n forbiddenPaths?: string[]\n /** Winner-selection strategy among gated candidates. Default `highest-score`. */\n winnerStrategy?: WinnerStrategy\n /** Test seams forwarded to the worktree-CLI leaves so the runner drives offline. */\n runGit?: WorktreeFanoutOptions['runGit']\n runHarness?: WorktreeFanoutOptions['runHarness']\n runCommand?: WorktreeFanoutOptions['runCommand']\n}\n\n/**\n *\n * `code` mode on the GENERIC recursive path: author one `AgentProfile` per harness, run them as a\n * `worktreeFanout` (N `createWorktreeCliExecutor` leaves, each `gateOnDeliverable`) through\n * `runPersonified` on the keystone Supervisor. The sandbox-session counterpart that drives the in-box\n * harness over a `SandboxClient` is `detachedSessionDelegate` (`./mcp/delegates`); here there is no\n * `runAgentRounds` driver, no role-coupled delegate — the harness list is the fanout, the gate is\n * `patchDelivered`,\n * the winner is the shared valid-only selector (NOT `defaultSelectWinner`, whose non-valid fallback\n * would surface an ungated patch). Equal-k holds by the conserved budget pool. Returns the winning\n * patch artifact, or throws when no candidate is delivered (fail loud, never a vacuous done).\n *\n * @experimental\n */\nexport function worktreeLoopRunner(\n options: WorktreeLoopRunnerOptions,\n): DelegatedLoopRunner<WorktreePatchArtifact> {\n const shape = worktreeFanout<string>({\n repoRoot: options.repoRoot,\n taskPrompt: options.taskPrompt,\n harnesses: options.harnesses,\n ...(options.testCmd !== undefined ? { testCmd: options.testCmd } : {}),\n ...(options.typecheckCmd !== undefined ? { typecheckCmd: options.typecheckCmd } : {}),\n ...(options.require !== undefined ? { require: options.require } : {}),\n ...(options.maxDiffLines !== undefined ? { maxDiffLines: options.maxDiffLines } : {}),\n ...(options.forbiddenPaths !== undefined ? { forbiddenPaths: options.forbiddenPaths } : {}),\n ...(options.winnerStrategy !== undefined ? { winnerStrategy: options.winnerStrategy } : {}),\n ...(options.runGit ? { runGit: options.runGit } : {}),\n ...(options.runHarness ? { runHarness: options.runHarness } : {}),\n ...(options.runCommand ? { runCommand: options.runCommand } : {}),\n })\n // The persona's only role here is to carry the fanout shape onto the Supervisor; each item's\n // executor is BYO (the gated worktree-CLI leaf), so the registry only needs to pass BYO through.\n const persona = definePersona<WorktreePatchArtifact>({\n name: 'worktree-coder',\n root: { profile: options.rootProfile, harness: null },\n directive: 'deliver a minimal validated patch on a fresh worktree',\n context: { role: 'coder' },\n executors: { registry: createExecutorRegistry() },\n })\n return async (signal) => {\n const result = await runPersonified<string, WorktreePatchArtifact>({\n persona,\n shape,\n task: options.taskPrompt,\n budget: options.budget,\n signal,\n })\n if (result.kind !== 'winner' || result.out.kind !== 'done') {\n const blockers =\n result.kind === 'winner' && result.out.kind === 'blocked'\n ? result.out.blockers.join('; ')\n : `supervisor settled ${result.kind}`\n throw new Error(`worktreeLoopRunner: no delivered patch (${blockers})`)\n }\n return result.out.deliverable\n }\n}\n\n/** @experimental A fact rejected at the KB gate — surfaced, never dropped. */\nexport interface VetoedFact {\n candidate: FactCandidate\n vetoedBy?: string\n reason?: string\n}\n\n/** @experimental */\nexport interface ResearchLoopResult {\n /** Facts that passed the fail-closed gate — safe to write to the KB. */\n accepted: FactCandidate[]\n /** Facts the gate vetoed in the final round — escalate, do not silently drop. */\n vetoed: VetoedFact[]\n /** Research rounds actually run. */\n rounds: number\n}\n\n/** @experimental Options for the default `research` runner. */\nexport interface ResearchLoopRunnerOptions {\n /**\n * The research engine (the consumer's web/doc searcher + extractor). Called\n * each round with the prior round's vetoes so it can re-research the gaps.\n * Returns fact candidates carrying their grounding (`verbatimPassage` +\n * `sourceText`).\n */\n research: (round: number, vetoed: VetoedFact[]) => Promise<FactCandidate[]>\n /** Gate config (extra judges, self-artifact kinds, …). The floor is always on. */\n gate?: CreateKbGateOptions\n /** Max research rounds (correct-on-veto remediation). Default 1. */\n maxRounds?: number\n}\n\n/**\n * `research` mode — research-in-a-loop with valid-only KB growth.\n *\n * Each round: research → gate every candidate (fail-closed; passage MUST be in\n * the source) → accept the clean ones → re-research the vetoed ones next round,\n * up to `maxRounds`. Vetoed facts in the final round are RETURNED (escalate,\n * never silently dropped) so the caller audits vs retries.\n *\n * @experimental\n */\nexport function researchLoopRunner(\n o: ResearchLoopRunnerOptions,\n): DelegatedLoopRunner<ResearchLoopResult> {\n const gate = createKbGate(o.gate)\n const maxRounds = Math.max(1, Math.trunc(o.maxRounds ?? 1))\n return async (signal) => {\n const accepted: FactCandidate[] = []\n let vetoed: VetoedFact[] = []\n let rounds = 0\n for (let round = 0; round < maxRounds; round += 1) {\n if (signal.aborted) break\n rounds += 1\n const candidates = await o.research(round, vetoed)\n if (candidates.length === 0) break\n vetoed = []\n for (const c of candidates) {\n const v = await gate(c)\n if (v.accepted) accepted.push(c)\n else vetoed.push({ candidate: c, vetoedBy: v.vetoedBy, reason: v.reason })\n }\n if (vetoed.length === 0) break\n }\n return { accepted, vetoed, rounds }\n }\n}\n\n/**\n * `audit` mode — analyst loop over captured trace/run data.\n *\n * @experimental\n */\nexport function auditLoopRunner<TProposal = unknown, TEdit = unknown>(\n options: RunAnalystLoopOpts,\n): DelegatedLoopRunner<RunAnalystLoopResult<TProposal, TEdit>> {\n return async () => runAnalystLoop<TProposal, TEdit>(options)\n}\n","#!/usr/bin/env node\n/**\n *\n * `agent-runtime-loop` — the schedulable entrypoint for the configured\n * delegated loop-runner. A cron job / routine / Makefile target invokes:\n *\n * agent-runtime-loop --mode research --config ./loops.config.js\n *\n * The config module wires the registry (with full access to env / creds —\n * which is why the deps live there, not in this generic bin). It must default-\n * export a `DelegatedLoopRegistry`, or a `() => DelegatedLoopRegistry | Promise<…>`.\n * The bin runs the selected mode, prints the `DelegatedLoopResult` as JSON, and\n * exits 0 on `ok`, 1 on a recorded failure, 2 on a usage/config error.\n *\n * @experimental\n */\n\nimport {\n DELEGATED_LOOP_MODES,\n type DelegatedLoopMode,\n type DelegatedLoopRegistry,\n type DelegatedLoopResult,\n isDelegatedLoopMode,\n runDelegatedLoop,\n} from './loop-runner'\n\n/** @experimental Parsed CLI invocation. */\nexport interface LoopRunnerCliArgs {\n mode: string\n /** Loads the registry — the bin wires this from `--config`; tests inject a stub. */\n loadRegistry: () => Promise<DelegatedLoopRegistry> | DelegatedLoopRegistry\n now?: () => number\n}\n\n/** @experimental */\nexport interface LoopRunnerCliResult {\n exitCode: number\n result?: DelegatedLoopResult\n error?: string\n}\n\n/**\n *\n * Pure CLI core (no process / argv / IO) so it's unit-testable: validate the\n * mode, load the registry, dispatch, map to an exit code (0 ok / 1 failed /\n * 2 usage). Exported for embedding in custom runners + tests.\n *\n * @experimental\n */\nexport async function runLoopRunnerCli(args: LoopRunnerCliArgs): Promise<LoopRunnerCliResult> {\n if (!isDelegatedLoopMode(args.mode)) {\n return {\n exitCode: 2,\n error: `unknown mode '${args.mode}' (expected one of: ${DELEGATED_LOOP_MODES.join(', ')})`,\n }\n }\n let registry: DelegatedLoopRegistry\n try {\n registry = await args.loadRegistry()\n } catch (err) {\n return { exitCode: 2, error: `failed to load registry: ${errMsg(err)}` }\n }\n if (!registry[args.mode]) {\n return {\n exitCode: 2,\n error: `config registers no runner for mode '${args.mode}' (registered: ${\n Object.keys(registry).join(', ') || 'none'\n })`,\n }\n }\n // runDelegatedLoop throws only on a missing runner (guarded above); a failing\n // engine is captured as { ok: false } → exit 1, not a crash.\n const result = await runDelegatedLoop(args.mode as DelegatedLoopMode, registry, {\n ...(args.now ? { now: args.now } : {}),\n })\n return { exitCode: result.ok ? 0 : 1, result }\n}\n\n/** Parse `--mode X --config Y` from an argv tail (`process.argv.slice(2)`). */\nexport function parseLoopRunnerArgv(argv: string[]): { mode?: string; config?: string } {\n const out: { mode?: string; config?: string } = {}\n for (let i = 0; i < argv.length; i += 1) {\n const a = argv[i]\n if (a === '--mode') out.mode = argv[++i]\n else if (a === '--config') out.config = argv[++i]\n else if (a?.startsWith('--mode=')) out.mode = a.slice('--mode='.length)\n else if (a?.startsWith('--config=')) out.config = a.slice('--config='.length)\n }\n return out\n}\n\n/** Normalize a config module's default export → a registry. */\nfunction resolveRegistry(mod: unknown): DelegatedLoopRegistry {\n const def = (mod as { default?: unknown })?.default ?? mod\n const value = typeof def === 'function' ? (def as () => unknown)() : def\n return value as DelegatedLoopRegistry\n}\n\nfunction errMsg(err: unknown): string {\n return err instanceof Error ? err.message : String(err)\n}\n\n/** The argv → IO → exit shell. Kept thin; logic lives in `runLoopRunnerCli`. */\nasync function main(): Promise<void> {\n const { mode, config } = parseLoopRunnerArgv(process.argv.slice(2))\n if (!mode || !config) {\n process.stderr.write(\n 'usage: agent-runtime-loop --mode <mode> --config <module>\\n' +\n ` modes: ${DELEGATED_LOOP_MODES.join(' | ')}\\n` +\n ' config: a JS/TS module default-exporting a DelegatedLoopRegistry (or a factory)\\n',\n )\n process.exit(2)\n }\n const { pathToFileURL } = await import('node:url')\n const { resolve } = await import('node:path')\n const cli = await runLoopRunnerCli({\n mode,\n loadRegistry: async () => resolveRegistry(await import(pathToFileURL(resolve(config)).href)),\n })\n process.stdout.write(`${JSON.stringify(cli.result ?? { error: cli.error }, null, 2)}\\n`)\n if (cli.error) process.stderr.write(`${cli.error}\\n`)\n process.exit(cli.exitCode)\n}\n\n// Run only when executed as the bin — never when imported for the testable\n// core, and never when bundled into a runtime that has no `process.argv`\n// (e.g. Cloudflare Workers, where `process` is a shim without `argv`). Reading\n// `process.argv[1]` directly would throw at module load there; `process.argv?.`\n// keeps the guard a no-op instead of crashing the Worker on startup.\nconst invokedScript = typeof process !== 'undefined' ? process.argv?.[1] : undefined\nif (invokedScript && /loop-runner-bin\\.(js|ts|mjs)$/.test(invokedScript)) {\n void main()\n}\n"],"mappings":";;;;;;;AAyCA,MAAa,uBAAuB;CAAC;CAAQ;CAAU;CAAY;CAAS;AAAc;;AAM1F,SAAgB,oBAAoB,OAA4C;CAC9E,OAAO,OAAO,UAAU,YAAa,qBAA2C,SAAS,KAAK;AAChG;;;;;;;;;;AAoCA,eAAsB,iBACpB,MACA,UACA,UAAmC,CAAC,GACH;CACjC,MAAM,SAAS,SAAS;CACxB,IAAI,CAAC,QACH,MAAM,IAAI,YACR,oDAAoD,KAAK,iBACvD,OAAO,KAAK,QAAQ,CAAC,CAAC,KAAK,IAAI,KAAK,OACrC,EACH;CAEF,MAAM,MAAM,QAAQ,OAAO,KAAK;CAChC,MAAM,SAAS,QAAQ,UAAU,IAAI,gBAAgB,CAAC,CAAC;CACvD,MAAM,QAAQ,IAAI;CAClB,IAAI;EAEF,OAAO;GAAE;GAAM,IAAI;GAAM,QAAA,MADJ,OAAO,MAAM;GACD,YAAY,IAAI,IAAI;EAAM;CAC7D,SAAS,KAAK;EACZ,OAAO;GACL;GACA,IAAI;GACJ,OAAO,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;GACtD,YAAY,IAAI,IAAI;EACtB;CACF;AACF;;;;;;;;;;;;;;;AA8CA,SAAgB,mBACd,SAC4C;CAC5C,MAAM,QAAQ,eAAuB;EACnC,UAAU,QAAQ;EAClB,YAAY,QAAQ;EACpB,WAAW,QAAQ;EACnB,GAAI,QAAQ,YAAY,KAAA,IAAY,EAAE,SAAS,QAAQ,QAAQ,IAAI,CAAC;EACpE,GAAI,QAAQ,iBAAiB,KAAA,IAAY,EAAE,cAAc,QAAQ,aAAa,IAAI,CAAC;EACnF,GAAI,QAAQ,YAAY,KAAA,IAAY,EAAE,SAAS,QAAQ,QAAQ,IAAI,CAAC;EACpE,GAAI,QAAQ,iBAAiB,KAAA,IAAY,EAAE,cAAc,QAAQ,aAAa,IAAI,CAAC;EACnF,GAAI,QAAQ,mBAAmB,KAAA,IAAY,EAAE,gBAAgB,QAAQ,eAAe,IAAI,CAAC;EACzF,GAAI,QAAQ,mBAAmB,KAAA,IAAY,EAAE,gBAAgB,QAAQ,eAAe,IAAI,CAAC;EACzF,GAAI,QAAQ,SAAS,EAAE,QAAQ,QAAQ,OAAO,IAAI,CAAC;EACnD,GAAI,QAAQ,aAAa,EAAE,YAAY,QAAQ,WAAW,IAAI,CAAC;EAC/D,GAAI,QAAQ,aAAa,EAAE,YAAY,QAAQ,WAAW,IAAI,CAAC;CACjE,CAAC;CAGD,MAAM,UAAU,cAAqC;EACnD,MAAM;EACN,MAAM;GAAE,SAAS,QAAQ;GAAa,SAAS;EAAK;EACpD,WAAW;EACX,SAAS,EAAE,MAAM,QAAQ;EACzB,WAAW,EAAE,UAAU,uBAAuB,EAAE;CAClD,CAAC;CACD,OAAO,OAAO,WAAW;EACvB,MAAM,SAAS,MAAM,eAA8C;GACjE;GACA;GACA,MAAM,QAAQ;GACd,QAAQ,QAAQ;GAChB;EACF,CAAC;EACD,IAAI,OAAO,SAAS,YAAY,OAAO,IAAI,SAAS,QAAQ;GAC1D,MAAM,WACJ,OAAO,SAAS,YAAY,OAAO,IAAI,SAAS,YAC5C,OAAO,IAAI,SAAS,KAAK,IAAI,IAC7B,sBAAsB,OAAO;GACnC,MAAM,IAAI,MAAM,2CAA2C,SAAS,EAAE;EACxE;EACA,OAAO,OAAO,IAAI;CACpB;AACF;;;;;;;;;;;AA4CA,SAAgB,mBACd,GACyC;CACzC,MAAM,OAAO,aAAa,EAAE,IAAI;CAChC,MAAM,YAAY,KAAK,IAAI,GAAG,KAAK,MAAM,EAAE,aAAa,CAAC,CAAC;CAC1D,OAAO,OAAO,WAAW;EACvB,MAAM,WAA4B,CAAC;EACnC,IAAI,SAAuB,CAAC;EAC5B,IAAI,SAAS;EACb,KAAK,IAAI,QAAQ,GAAG,QAAQ,WAAW,SAAS,GAAG;GACjD,IAAI,OAAO,SAAS;GACpB,UAAU;GACV,MAAM,aAAa,MAAM,EAAE,SAAS,OAAO,MAAM;GACjD,IAAI,WAAW,WAAW,GAAG;GAC7B,SAAS,CAAC;GACV,KAAK,MAAM,KAAK,YAAY;IAC1B,MAAM,IAAI,MAAM,KAAK,CAAC;IACtB,IAAI,EAAE,UAAU,SAAS,KAAK,CAAC;SAC1B,OAAO,KAAK;KAAE,WAAW;KAAG,UAAU,EAAE;KAAU,QAAQ,EAAE;IAAO,CAAC;GAC3E;GACA,IAAI,OAAO,WAAW,GAAG;EAC3B;EACA,OAAO;GAAE;GAAU;GAAQ;EAAO;CACpC;AACF;;;;;;AAOA,SAAgB,gBACd,SAC6D;CAC7D,OAAO,YAAY,eAAiC,OAAO;AAC7D;;;;;;;;;;;;;;;;;;;;;;;;;;ACvOA,eAAsB,iBAAiB,MAAuD;CAC5F,IAAI,CAAC,oBAAoB,KAAK,IAAI,GAChC,OAAO;EACL,UAAU;EACV,OAAO,iBAAiB,KAAK,KAAK,sBAAsB,qBAAqB,KAAK,IAAI,EAAE;CAC1F;CAEF,IAAI;CACJ,IAAI;EACF,WAAW,MAAM,KAAK,aAAa;CACrC,SAAS,KAAK;EACZ,OAAO;GAAE,UAAU;GAAG,OAAO,4BAA4B,OAAO,GAAG;EAAI;CACzE;CACA,IAAI,CAAC,SAAS,KAAK,OACjB,OAAO;EACL,UAAU;EACV,OAAO,wCAAwC,KAAK,KAAK,iBACvD,OAAO,KAAK,QAAQ,CAAC,CAAC,KAAK,IAAI,KAAK,OACrC;CACH;CAIF,MAAM,SAAS,MAAM,iBAAiB,KAAK,MAA2B,UAAU,EAC9E,GAAI,KAAK,MAAM,EAAE,KAAK,KAAK,IAAI,IAAI,CAAC,EACtC,CAAC;CACD,OAAO;EAAE,UAAU,OAAO,KAAK,IAAI;EAAG;CAAO;AAC/C;;AAGA,SAAgB,oBAAoB,MAAoD;CACtF,MAAM,MAA0C,CAAC;CACjD,KAAK,IAAI,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK,GAAG;EACvC,MAAM,IAAI,KAAK;EACf,IAAI,MAAM,UAAU,IAAI,OAAO,KAAK,EAAE;OACjC,IAAI,MAAM,YAAY,IAAI,SAAS,KAAK,EAAE;OAC1C,IAAI,GAAG,WAAW,SAAS,GAAG,IAAI,OAAO,EAAE,MAAM,CAAgB;OACjE,IAAI,GAAG,WAAW,WAAW,GAAG,IAAI,SAAS,EAAE,MAAM,CAAkB;CAC9E;CACA,OAAO;AACT;;AAGA,SAAS,gBAAgB,KAAqC;CAC5D,MAAM,MAAO,KAA+B,WAAW;CAEvD,OADc,OAAO,QAAQ,aAAc,IAAsB,IAAI;AAEvE;AAEA,SAAS,OAAO,KAAsB;CACpC,OAAO,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;AACxD;;AAGA,eAAe,OAAsB;CACnC,MAAM,EAAE,MAAM,WAAW,oBAAoB,QAAQ,KAAK,MAAM,CAAC,CAAC;CAClE,IAAI,CAAC,QAAQ,CAAC,QAAQ;EACpB,QAAQ,OAAO,MACb;WACc,qBAAqB,KAAK,KAAK,EAAE;CAEjD;EACA,QAAQ,KAAK,CAAC;CAChB;CACA,MAAM,EAAE,kBAAkB,MAAM,OAAO;CACvC,MAAM,EAAE,YAAY,MAAM,OAAO;CACjC,MAAM,MAAM,MAAM,iBAAiB;EACjC;EACA,cAAc,YAAY,gBAAgB,MAAM,OAAO,cAAc,QAAQ,MAAM,CAAC,CAAC,CAAC,KAAK;CAC7F,CAAC;CACD,QAAQ,OAAO,MAAM,GAAG,KAAK,UAAU,IAAI,UAAU,EAAE,OAAO,IAAI,MAAM,GAAG,MAAM,CAAC,EAAE,GAAG;CACvF,IAAI,IAAI,OAAO,QAAQ,OAAO,MAAM,GAAG,IAAI,MAAM,GAAG;CACpD,QAAQ,KAAK,IAAI,QAAQ;AAC3B;AAOA,MAAM,gBAAgB,OAAO,YAAY,cAAc,QAAQ,OAAO,KAAK,KAAA;AAC3E,IAAI,iBAAiB,gCAAgC,KAAK,aAAa,GACrE,KAAU"}
|
|
1
|
+
{"version":3,"file":"loop-runner-bin-DKEzvd7f.js","names":[],"sources":["../src/loop-runner.ts","../src/loop-runner-bin.ts"],"sourcesContent":["/**\n *\n * `runDelegatedLoop` — the configured delegated loop-runner.\n *\n * One typed entrypoint a worker agent (or a scheduled routine) calls to run a\n * disciplined loop in a chosen MODE, over agent-runtime's hardened engines:\n *\n * code → build-in-a-loop on the GENERIC recursive path (worktreeLoopRunner: author one\n * `AgentProfile` per harness → worktree-CLI leaves → `patchDelivered` gate)\n * review → caller-registered runner — a `code` runner with an approval gate over candidates\n * research → research-in-a-loop with valid-only KB growth (createKbGate)\n * audit → analyze trace/run data → findings (runAnalystLoop, caller-wired)\n * self-improve → caller-registered `improve(profile, options)` run\n *\n * It is intentionally a thin façade: the value is that EVERY product reuses the\n * one hardened engine instead of forking delegation logic. The dispatcher owns\n * mode routing, timing, fail-loud on an unregistered mode, and a uniform result\n * shape; each mode's engine is a pre-configured runner in the registry (build it\n * with the factories below, or inject your own / a stub).\n *\n * @experimental\n */\n\nimport type { AgentProfile } from '@tangle-network/agent-interface'\nimport { runAnalystLoop } from './analyst-loop'\nimport type { RunAnalystLoopOpts, RunAnalystLoopResult } from './analyst-loop/types'\nimport { ConfigError } from './errors'\nimport { type CreateKbGateOptions, createKbGate, type FactCandidate } from './mcp/kb-gate'\nimport {\n type AuthoredHarness,\n type Budget,\n createExecutorRegistry,\n definePersona,\n runPersonified,\n type WinnerStrategy,\n type WorktreeFanoutOptions,\n type WorktreePatchArtifact,\n worktreeFanout,\n} from './runtime'\n\n/** All valid delegated-loop mode names — used for validation and CLI surfaces. @experimental */\nexport const DELEGATED_LOOP_MODES = ['code', 'review', 'research', 'audit', 'self-improve'] as const\n\n/** @experimental */\nexport type DelegatedLoopMode = (typeof DELEGATED_LOOP_MODES)[number]\n\n/** Type guard — returns true when `value` is a valid `DelegatedLoopMode` string. @experimental */\nexport function isDelegatedLoopMode(value: unknown): value is DelegatedLoopMode {\n return typeof value === 'string' && (DELEGATED_LOOP_MODES as readonly string[]).includes(value)\n}\n\n/** @experimental A pre-configured loop for one mode. Returns the mode's raw\n * output; the dispatcher wraps it in a {@link DelegatedLoopResult}. */\nexport type DelegatedLoopRunner<T = unknown> = (signal: AbortSignal) => Promise<T>\n\n/** @experimental Mode → configured runner. Partial: only register the modes a\n * given product/routine actually uses. */\nexport type DelegatedLoopRegistry = Partial<Record<DelegatedLoopMode, DelegatedLoopRunner>>\n\n/** @experimental Uniform result — never throws from a registered runner; a\n * thrown engine becomes `{ ok: false, error }` so a routine can record + move on. */\nexport interface DelegatedLoopResult<T = unknown> {\n mode: DelegatedLoopMode\n ok: boolean\n output?: T\n error?: string\n durationMs: number\n}\n\n/** @experimental */\nexport interface RunDelegatedLoopOptions {\n signal?: AbortSignal\n /** Clock override for deterministic tests. */\n now?: () => number\n}\n\n/**\n *\n * Dispatch a configured loop by mode. Fails loud (throws `ConfigError`) when no\n * runner is registered for the mode — a routine pointed at an unwired mode is a\n * config bug, not a silent no-op. A runner that throws is captured as\n * `{ ok: false }` so unattended runs record the failure rather than crash.\n *\n * @experimental\n */\nexport async function runDelegatedLoop<T = unknown>(\n mode: DelegatedLoopMode,\n registry: DelegatedLoopRegistry,\n options: RunDelegatedLoopOptions = {},\n): Promise<DelegatedLoopResult<T>> {\n const runner = registry[mode] as DelegatedLoopRunner<T> | undefined\n if (!runner) {\n throw new ConfigError(\n `runDelegatedLoop: no runner registered for mode '${mode}' (registered: ${\n Object.keys(registry).join(', ') || 'none'\n })`,\n )\n }\n const now = options.now ?? Date.now\n const signal = options.signal ?? new AbortController().signal\n const start = now()\n try {\n const output = await runner(signal)\n return { mode, ok: true, output, durationMs: now() - start }\n } catch (err) {\n return {\n mode,\n ok: false,\n error: err instanceof Error ? err.message : String(err),\n durationMs: now() - start,\n }\n }\n}\n\n/** @experimental Options for the local-repo `code` runner over the GENERIC recursive path. */\nexport interface WorktreeLoopRunnerOptions {\n /** Exact profile carried by the personified root that owns this fanout. */\n rootProfile: AgentProfile\n /** Absolute path to the local git checkout each worktree is cut from. */\n repoRoot: string\n /** The instruction handed to every authored harness (composed under each profile's systemPrompt). */\n taskPrompt: string\n /** The supervisor-authored harness profiles — one fanout item (one worktree-CLI leaf) each. */\n harnesses: ReadonlyArray<AuthoredHarness>\n /** Conserved budget pool bounding the fanout (equal-k holds by construction). */\n budget: Budget\n /** Shell command run in each worktree to derive the tests-PASS signal. */\n testCmd?: string\n /** Shell command run in each worktree to derive the typecheck-PASS signal. */\n typecheckCmd?: string\n /** Which verification signals the deliverable REQUIRES present-and-passing (default none). */\n require?: ReadonlyArray<'tests' | 'typecheck'>\n /** Diff-size cap (lines). */\n maxDiffLines?: number\n /** Literal path prefixes the patch must not touch (the secret-floor is always on regardless). */\n forbiddenPaths?: string[]\n /** Winner-selection strategy among gated candidates. Default `highest-score`. */\n winnerStrategy?: WinnerStrategy\n /** Test seams forwarded to the worktree-CLI leaves so the runner drives offline. */\n runGit?: WorktreeFanoutOptions['runGit']\n runHarness?: WorktreeFanoutOptions['runHarness']\n runCommand?: WorktreeFanoutOptions['runCommand']\n}\n\n/**\n *\n * `code` mode on the GENERIC recursive path: author one `AgentProfile` per harness, run them as a\n * `worktreeFanout` (N `createWorktreeCliExecutor` leaves, each `gateOnDeliverable`) through\n * `runPersonified` on the keystone Supervisor. The sandbox-session counterpart that drives the in-box\n * harness over a `SandboxClient` is `detachedSessionDelegate` (`./mcp/delegates`); here there is no\n * `runAgentRounds` driver, no role-coupled delegate — the harness list is the fanout, the gate is\n * `patchDelivered`,\n * the winner is the shared valid-only selector (NOT `defaultSelectWinner`, whose non-valid fallback\n * would surface an ungated patch). Equal-k holds by the conserved budget pool. Returns the winning\n * patch artifact, or throws when no candidate is delivered (fail loud, never a vacuous done).\n *\n * @experimental\n */\nexport function worktreeLoopRunner(\n options: WorktreeLoopRunnerOptions,\n): DelegatedLoopRunner<WorktreePatchArtifact> {\n const shape = worktreeFanout<string>({\n repoRoot: options.repoRoot,\n taskPrompt: options.taskPrompt,\n harnesses: options.harnesses,\n ...(options.testCmd !== undefined ? { testCmd: options.testCmd } : {}),\n ...(options.typecheckCmd !== undefined ? { typecheckCmd: options.typecheckCmd } : {}),\n ...(options.require !== undefined ? { require: options.require } : {}),\n ...(options.maxDiffLines !== undefined ? { maxDiffLines: options.maxDiffLines } : {}),\n ...(options.forbiddenPaths !== undefined ? { forbiddenPaths: options.forbiddenPaths } : {}),\n ...(options.winnerStrategy !== undefined ? { winnerStrategy: options.winnerStrategy } : {}),\n ...(options.runGit ? { runGit: options.runGit } : {}),\n ...(options.runHarness ? { runHarness: options.runHarness } : {}),\n ...(options.runCommand ? { runCommand: options.runCommand } : {}),\n })\n // The persona's only role here is to carry the fanout shape onto the Supervisor; each item's\n // executor is BYO (the gated worktree-CLI leaf), so the registry only needs to pass BYO through.\n const persona = definePersona<WorktreePatchArtifact>({\n name: 'worktree-coder',\n root: { profile: options.rootProfile, harness: null },\n directive: 'deliver a minimal validated patch on a fresh worktree',\n context: { role: 'coder' },\n executors: { registry: createExecutorRegistry() },\n })\n return async (signal) => {\n const result = await runPersonified<string, WorktreePatchArtifact>({\n persona,\n shape,\n task: options.taskPrompt,\n budget: options.budget,\n signal,\n })\n if (result.kind !== 'winner' || result.out.kind !== 'done') {\n const blockers =\n result.kind === 'winner' && result.out.kind === 'blocked'\n ? result.out.blockers.join('; ')\n : `supervisor settled ${result.kind}`\n throw new Error(`worktreeLoopRunner: no delivered patch (${blockers})`)\n }\n return result.out.deliverable\n }\n}\n\n/** @experimental A fact rejected at the KB gate — surfaced, never dropped. */\nexport interface VetoedFact {\n candidate: FactCandidate\n vetoedBy?: string\n reason?: string\n}\n\n/** @experimental */\nexport interface ResearchLoopResult {\n /** Facts that passed the fail-closed gate — safe to write to the KB. */\n accepted: FactCandidate[]\n /** Facts the gate vetoed in the final round — escalate, do not silently drop. */\n vetoed: VetoedFact[]\n /** Research rounds actually run. */\n rounds: number\n}\n\n/** @experimental Options for the default `research` runner. */\nexport interface ResearchLoopRunnerOptions {\n /**\n * The research engine (the consumer's web/doc searcher + extractor). Called\n * each round with the prior round's vetoes so it can re-research the gaps.\n * Returns fact candidates carrying their grounding (`verbatimPassage` +\n * `sourceText`).\n */\n research: (round: number, vetoed: VetoedFact[]) => Promise<FactCandidate[]>\n /** Gate config (extra judges, self-artifact kinds, …). The floor is always on. */\n gate?: CreateKbGateOptions\n /** Max research rounds (correct-on-veto remediation). Default 1. */\n maxRounds?: number\n}\n\n/**\n * `research` mode — research-in-a-loop with valid-only KB growth.\n *\n * Each round: research → gate every candidate (fail-closed; passage MUST be in\n * the source) → accept the clean ones → re-research the vetoed ones next round,\n * up to `maxRounds`. Vetoed facts in the final round are RETURNED (escalate,\n * never silently dropped) so the caller audits vs retries.\n *\n * @experimental\n */\nexport function researchLoopRunner(\n o: ResearchLoopRunnerOptions,\n): DelegatedLoopRunner<ResearchLoopResult> {\n const gate = createKbGate(o.gate)\n const maxRounds = Math.max(1, Math.trunc(o.maxRounds ?? 1))\n return async (signal) => {\n const accepted: FactCandidate[] = []\n let vetoed: VetoedFact[] = []\n let rounds = 0\n for (let round = 0; round < maxRounds; round += 1) {\n if (signal.aborted) break\n rounds += 1\n const candidates = await o.research(round, vetoed)\n if (candidates.length === 0) break\n vetoed = []\n for (const c of candidates) {\n const v = await gate(c)\n if (v.accepted) accepted.push(c)\n else vetoed.push({ candidate: c, vetoedBy: v.vetoedBy, reason: v.reason })\n }\n if (vetoed.length === 0) break\n }\n return { accepted, vetoed, rounds }\n }\n}\n\n/**\n * `audit` mode — analyst loop over captured trace/run data.\n *\n * @experimental\n */\nexport function auditLoopRunner<TProposal = unknown, TEdit = unknown>(\n options: RunAnalystLoopOpts,\n): DelegatedLoopRunner<RunAnalystLoopResult<TProposal, TEdit>> {\n return async () => runAnalystLoop<TProposal, TEdit>(options)\n}\n","#!/usr/bin/env node\n/**\n *\n * `agent-runtime-loop` — the schedulable entrypoint for the configured\n * delegated loop-runner. A cron job / routine / Makefile target invokes:\n *\n * agent-runtime-loop --mode research --config ./loops.config.js\n *\n * The config module wires the registry (with full access to env / creds —\n * which is why the deps live there, not in this generic bin). It must default-\n * export a `DelegatedLoopRegistry`, or a `() => DelegatedLoopRegistry | Promise<…>`.\n * The bin runs the selected mode, prints the `DelegatedLoopResult` as JSON, and\n * exits 0 on `ok`, 1 on a recorded failure, 2 on a usage/config error.\n *\n * @experimental\n */\n\nimport {\n DELEGATED_LOOP_MODES,\n type DelegatedLoopMode,\n type DelegatedLoopRegistry,\n type DelegatedLoopResult,\n isDelegatedLoopMode,\n runDelegatedLoop,\n} from './loop-runner'\n\n/** @experimental Parsed CLI invocation. */\nexport interface LoopRunnerCliArgs {\n mode: string\n /** Loads the registry — the bin wires this from `--config`; tests inject a stub. */\n loadRegistry: () => Promise<DelegatedLoopRegistry> | DelegatedLoopRegistry\n now?: () => number\n}\n\n/** @experimental */\nexport interface LoopRunnerCliResult {\n exitCode: number\n result?: DelegatedLoopResult\n error?: string\n}\n\n/**\n *\n * Pure CLI core (no process / argv / IO) so it's unit-testable: validate the\n * mode, load the registry, dispatch, map to an exit code (0 ok / 1 failed /\n * 2 usage). Exported for embedding in custom runners + tests.\n *\n * @experimental\n */\nexport async function runLoopRunnerCli(args: LoopRunnerCliArgs): Promise<LoopRunnerCliResult> {\n if (!isDelegatedLoopMode(args.mode)) {\n return {\n exitCode: 2,\n error: `unknown mode '${args.mode}' (expected one of: ${DELEGATED_LOOP_MODES.join(', ')})`,\n }\n }\n let registry: DelegatedLoopRegistry\n try {\n registry = await args.loadRegistry()\n } catch (err) {\n return { exitCode: 2, error: `failed to load registry: ${errMsg(err)}` }\n }\n if (!registry[args.mode]) {\n return {\n exitCode: 2,\n error: `config registers no runner for mode '${args.mode}' (registered: ${\n Object.keys(registry).join(', ') || 'none'\n })`,\n }\n }\n // runDelegatedLoop throws only on a missing runner (guarded above); a failing\n // engine is captured as { ok: false } → exit 1, not a crash.\n const result = await runDelegatedLoop(args.mode as DelegatedLoopMode, registry, {\n ...(args.now ? { now: args.now } : {}),\n })\n return { exitCode: result.ok ? 0 : 1, result }\n}\n\n/** Parse `--mode X --config Y` from an argv tail (`process.argv.slice(2)`). */\nexport function parseLoopRunnerArgv(argv: string[]): { mode?: string; config?: string } {\n const out: { mode?: string; config?: string } = {}\n for (let i = 0; i < argv.length; i += 1) {\n const a = argv[i]\n if (a === '--mode') out.mode = argv[++i]\n else if (a === '--config') out.config = argv[++i]\n else if (a?.startsWith('--mode=')) out.mode = a.slice('--mode='.length)\n else if (a?.startsWith('--config=')) out.config = a.slice('--config='.length)\n }\n return out\n}\n\n/** Normalize a config module's default export → a registry. */\nfunction resolveRegistry(mod: unknown): DelegatedLoopRegistry {\n const def = (mod as { default?: unknown })?.default ?? mod\n const value = typeof def === 'function' ? (def as () => unknown)() : def\n return value as DelegatedLoopRegistry\n}\n\nfunction errMsg(err: unknown): string {\n return err instanceof Error ? err.message : String(err)\n}\n\n/** The argv → IO → exit shell. Kept thin; logic lives in `runLoopRunnerCli`. */\nasync function main(): Promise<void> {\n const { mode, config } = parseLoopRunnerArgv(process.argv.slice(2))\n if (!mode || !config) {\n process.stderr.write(\n 'usage: agent-runtime-loop --mode <mode> --config <module>\\n' +\n ` modes: ${DELEGATED_LOOP_MODES.join(' | ')}\\n` +\n ' config: a JS/TS module default-exporting a DelegatedLoopRegistry (or a factory)\\n',\n )\n process.exit(2)\n }\n const { pathToFileURL } = await import('node:url')\n const { resolve } = await import('node:path')\n const cli = await runLoopRunnerCli({\n mode,\n loadRegistry: async () => resolveRegistry(await import(pathToFileURL(resolve(config)).href)),\n })\n process.stdout.write(`${JSON.stringify(cli.result ?? { error: cli.error }, null, 2)}\\n`)\n if (cli.error) process.stderr.write(`${cli.error}\\n`)\n process.exit(cli.exitCode)\n}\n\n// Run only when executed as the bin — never when imported for the testable\n// core, and never when bundled into a runtime that has no `process.argv`\n// (e.g. Cloudflare Workers, where `process` is a shim without `argv`). Reading\n// `process.argv[1]` directly would throw at module load there; `process.argv?.`\n// keeps the guard a no-op instead of crashing the Worker on startup.\nconst invokedScript = typeof process !== 'undefined' ? process.argv?.[1] : undefined\nif (invokedScript && /loop-runner-bin\\.(js|ts|mjs)$/.test(invokedScript)) {\n void main()\n}\n"],"mappings":";;;;;;;AAyCA,MAAa,uBAAuB;CAAC;CAAQ;CAAU;CAAY;CAAS;AAAc;;AAM1F,SAAgB,oBAAoB,OAA4C;CAC9E,OAAO,OAAO,UAAU,YAAa,qBAA2C,SAAS,KAAK;AAChG;;;;;;;;;;AAoCA,eAAsB,iBACpB,MACA,UACA,UAAmC,CAAC,GACH;CACjC,MAAM,SAAS,SAAS;CACxB,IAAI,CAAC,QACH,MAAM,IAAI,YACR,oDAAoD,KAAK,iBACvD,OAAO,KAAK,QAAQ,CAAC,CAAC,KAAK,IAAI,KAAK,OACrC,EACH;CAEF,MAAM,MAAM,QAAQ,OAAO,KAAK;CAChC,MAAM,SAAS,QAAQ,UAAU,IAAI,gBAAgB,CAAC,CAAC;CACvD,MAAM,QAAQ,IAAI;CAClB,IAAI;EAEF,OAAO;GAAE;GAAM,IAAI;GAAM,QAAA,MADJ,OAAO,MAAM;GACD,YAAY,IAAI,IAAI;EAAM;CAC7D,SAAS,KAAK;EACZ,OAAO;GACL;GACA,IAAI;GACJ,OAAO,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;GACtD,YAAY,IAAI,IAAI;EACtB;CACF;AACF;;;;;;;;;;;;;;;AA8CA,SAAgB,mBACd,SAC4C;CAC5C,MAAM,QAAQ,eAAuB;EACnC,UAAU,QAAQ;EAClB,YAAY,QAAQ;EACpB,WAAW,QAAQ;EACnB,GAAI,QAAQ,YAAY,KAAA,IAAY,EAAE,SAAS,QAAQ,QAAQ,IAAI,CAAC;EACpE,GAAI,QAAQ,iBAAiB,KAAA,IAAY,EAAE,cAAc,QAAQ,aAAa,IAAI,CAAC;EACnF,GAAI,QAAQ,YAAY,KAAA,IAAY,EAAE,SAAS,QAAQ,QAAQ,IAAI,CAAC;EACpE,GAAI,QAAQ,iBAAiB,KAAA,IAAY,EAAE,cAAc,QAAQ,aAAa,IAAI,CAAC;EACnF,GAAI,QAAQ,mBAAmB,KAAA,IAAY,EAAE,gBAAgB,QAAQ,eAAe,IAAI,CAAC;EACzF,GAAI,QAAQ,mBAAmB,KAAA,IAAY,EAAE,gBAAgB,QAAQ,eAAe,IAAI,CAAC;EACzF,GAAI,QAAQ,SAAS,EAAE,QAAQ,QAAQ,OAAO,IAAI,CAAC;EACnD,GAAI,QAAQ,aAAa,EAAE,YAAY,QAAQ,WAAW,IAAI,CAAC;EAC/D,GAAI,QAAQ,aAAa,EAAE,YAAY,QAAQ,WAAW,IAAI,CAAC;CACjE,CAAC;CAGD,MAAM,UAAU,cAAqC;EACnD,MAAM;EACN,MAAM;GAAE,SAAS,QAAQ;GAAa,SAAS;EAAK;EACpD,WAAW;EACX,SAAS,EAAE,MAAM,QAAQ;EACzB,WAAW,EAAE,UAAU,uBAAuB,EAAE;CAClD,CAAC;CACD,OAAO,OAAO,WAAW;EACvB,MAAM,SAAS,MAAM,eAA8C;GACjE;GACA;GACA,MAAM,QAAQ;GACd,QAAQ,QAAQ;GAChB;EACF,CAAC;EACD,IAAI,OAAO,SAAS,YAAY,OAAO,IAAI,SAAS,QAAQ;GAC1D,MAAM,WACJ,OAAO,SAAS,YAAY,OAAO,IAAI,SAAS,YAC5C,OAAO,IAAI,SAAS,KAAK,IAAI,IAC7B,sBAAsB,OAAO;GACnC,MAAM,IAAI,MAAM,2CAA2C,SAAS,EAAE;EACxE;EACA,OAAO,OAAO,IAAI;CACpB;AACF;;;;;;;;;;;AA4CA,SAAgB,mBACd,GACyC;CACzC,MAAM,OAAO,aAAa,EAAE,IAAI;CAChC,MAAM,YAAY,KAAK,IAAI,GAAG,KAAK,MAAM,EAAE,aAAa,CAAC,CAAC;CAC1D,OAAO,OAAO,WAAW;EACvB,MAAM,WAA4B,CAAC;EACnC,IAAI,SAAuB,CAAC;EAC5B,IAAI,SAAS;EACb,KAAK,IAAI,QAAQ,GAAG,QAAQ,WAAW,SAAS,GAAG;GACjD,IAAI,OAAO,SAAS;GACpB,UAAU;GACV,MAAM,aAAa,MAAM,EAAE,SAAS,OAAO,MAAM;GACjD,IAAI,WAAW,WAAW,GAAG;GAC7B,SAAS,CAAC;GACV,KAAK,MAAM,KAAK,YAAY;IAC1B,MAAM,IAAI,MAAM,KAAK,CAAC;IACtB,IAAI,EAAE,UAAU,SAAS,KAAK,CAAC;SAC1B,OAAO,KAAK;KAAE,WAAW;KAAG,UAAU,EAAE;KAAU,QAAQ,EAAE;IAAO,CAAC;GAC3E;GACA,IAAI,OAAO,WAAW,GAAG;EAC3B;EACA,OAAO;GAAE;GAAU;GAAQ;EAAO;CACpC;AACF;;;;;;AAOA,SAAgB,gBACd,SAC6D;CAC7D,OAAO,YAAY,eAAiC,OAAO;AAC7D;;;;;;;;;;;;;;;;;;;;;;;;;;ACvOA,eAAsB,iBAAiB,MAAuD;CAC5F,IAAI,CAAC,oBAAoB,KAAK,IAAI,GAChC,OAAO;EACL,UAAU;EACV,OAAO,iBAAiB,KAAK,KAAK,sBAAsB,qBAAqB,KAAK,IAAI,EAAE;CAC1F;CAEF,IAAI;CACJ,IAAI;EACF,WAAW,MAAM,KAAK,aAAa;CACrC,SAAS,KAAK;EACZ,OAAO;GAAE,UAAU;GAAG,OAAO,4BAA4B,OAAO,GAAG;EAAI;CACzE;CACA,IAAI,CAAC,SAAS,KAAK,OACjB,OAAO;EACL,UAAU;EACV,OAAO,wCAAwC,KAAK,KAAK,iBACvD,OAAO,KAAK,QAAQ,CAAC,CAAC,KAAK,IAAI,KAAK,OACrC;CACH;CAIF,MAAM,SAAS,MAAM,iBAAiB,KAAK,MAA2B,UAAU,EAC9E,GAAI,KAAK,MAAM,EAAE,KAAK,KAAK,IAAI,IAAI,CAAC,EACtC,CAAC;CACD,OAAO;EAAE,UAAU,OAAO,KAAK,IAAI;EAAG;CAAO;AAC/C;;AAGA,SAAgB,oBAAoB,MAAoD;CACtF,MAAM,MAA0C,CAAC;CACjD,KAAK,IAAI,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK,GAAG;EACvC,MAAM,IAAI,KAAK;EACf,IAAI,MAAM,UAAU,IAAI,OAAO,KAAK,EAAE;OACjC,IAAI,MAAM,YAAY,IAAI,SAAS,KAAK,EAAE;OAC1C,IAAI,GAAG,WAAW,SAAS,GAAG,IAAI,OAAO,EAAE,MAAM,CAAgB;OACjE,IAAI,GAAG,WAAW,WAAW,GAAG,IAAI,SAAS,EAAE,MAAM,CAAkB;CAC9E;CACA,OAAO;AACT;;AAGA,SAAS,gBAAgB,KAAqC;CAC5D,MAAM,MAAO,KAA+B,WAAW;CAEvD,OADc,OAAO,QAAQ,aAAc,IAAsB,IAAI;AAEvE;AAEA,SAAS,OAAO,KAAsB;CACpC,OAAO,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;AACxD;;AAGA,eAAe,OAAsB;CACnC,MAAM,EAAE,MAAM,WAAW,oBAAoB,QAAQ,KAAK,MAAM,CAAC,CAAC;CAClE,IAAI,CAAC,QAAQ,CAAC,QAAQ;EACpB,QAAQ,OAAO,MACb;WACc,qBAAqB,KAAK,KAAK,EAAE;CAEjD;EACA,QAAQ,KAAK,CAAC;CAChB;CACA,MAAM,EAAE,kBAAkB,MAAM,OAAO;CACvC,MAAM,EAAE,YAAY,MAAM,OAAO;CACjC,MAAM,MAAM,MAAM,iBAAiB;EACjC;EACA,cAAc,YAAY,gBAAgB,MAAM,OAAO,cAAc,QAAQ,MAAM,CAAC,CAAC,CAAC,KAAK;CAC7F,CAAC;CACD,QAAQ,OAAO,MAAM,GAAG,KAAK,UAAU,IAAI,UAAU,EAAE,OAAO,IAAI,MAAM,GAAG,MAAM,CAAC,EAAE,GAAG;CACvF,IAAI,IAAI,OAAO,QAAQ,OAAO,MAAM,GAAG,IAAI,MAAM,GAAG;CACpD,QAAQ,KAAK,IAAI,QAAQ;AAC3B;AAOA,MAAM,gBAAgB,OAAO,YAAY,cAAc,QAAQ,OAAO,KAAK,KAAA;AAC3E,IAAI,iBAAiB,gCAAgC,KAAK,aAAa,GACrE,KAAU"}
|
package/dist/loop-runner-bin.js
CHANGED
package/dist/mcp/index.js
CHANGED
|
@@ -2,7 +2,7 @@ import { i as ConfigError, m as ValidationError } from "../errors-CDZ8XsVj.js";
|
|
|
2
2
|
import { d as sleep, i as deleteBoxSafe, m as throwIfAborted, p as throwAbort } from "../util-C1bv6yog.js";
|
|
3
3
|
import { t as assertExecutableAgentProfile } from "../model-policy-DgRJIJse.js";
|
|
4
4
|
import { Ct as traceContextToEnv, Gt as createWorktree, Jt as LOCAL_HARNESSES, Kt as removeWorktree, Qt as CodexExecutionDiagnosticError, St as readTraceContextFromEnv, Ut as runWorktreeHarness, Wt as captureWorktreeDiff, Xt as localHarnessExecutable, Yt as harnessSupportsReasoningEffort, Zt as parseCodexTokenUsage, bt as createPropagatingTraceEmitter, ft as runAgentRounds, lt as createSandboxForSpec, qt as DEFAULT_LOCAL_HARNESS, xt as mergeTraceEnv } from "../supervisor-BO7B2eim.js";
|
|
5
|
-
import { T as runCoderChecks, dt as selectValidWinner } from "../runtime-
|
|
5
|
+
import { T as runCoderChecks, dt as selectValidWinner } from "../runtime-WOQ24LYd.js";
|
|
6
6
|
import { $ as DELEGATION_TRACE_MAX_BYTES, A as DELEGATION_HISTORY_INPUT_SCHEMA, B as DELEGATE_FEEDBACK_INPUT_SCHEMA, C as createMcpServer, D as createDelegationStatusHandler, Dt as createCoordinationTools, E as DELEGATION_STATUS_TOOL_NAME, F as DELEGATE_UI_AUDIT_INPUT_SCHEMA, G as DELEGATE_INPUT_SCHEMA, H as createDelegateFeedbackHandler, I as DELEGATE_UI_AUDIT_TOOL_NAME, J as validateDelegateArgs, K as DELEGATE_TOOL_NAME, L as createDelegateUiAuditHandler, M as createDelegationHistoryHandler, N as validateDelegationHistoryArgs, O as validateDelegationStatusArgs, P as DELEGATE_UI_AUDIT_DESCRIPTION, Q as hashIdempotencyInput, R as validateDelegateUiAuditArgs, S as createInProcessTransport, T as DELEGATION_STATUS_INPUT_SCHEMA, Tt as DEFAULT_AWAIT_EVENT_TIMEOUT_MS, U as validateDelegateFeedbackArgs, V as DELEGATE_FEEDBACK_TOOL_NAME, W as DELEGATE_DESCRIPTION, Z as DelegationTaskQueue, at as DelegationPersistenceError, ct as InMemoryDelegationStore, et as DELEGATION_TRACE_MAX_SPANS, it as createDelegationTraceCollector, j as DELEGATION_HISTORY_TOOL_NAME, k as DELEGATION_HISTORY_DESCRIPTION, lt as InMemoryFeedbackStore, nt as capDelegationTrace, ot as DelegationStateCorruptError, q as createDelegateHandler, rt as composeLoopTraceEmitters, st as FileDelegationStore, tt as buildDelegationTraceSpans, ut as eventToSnapshot, w as DELEGATION_STATUS_DESCRIPTION, z as DELEGATE_FEEDBACK_DESCRIPTION } from "../supervise-BA2aQ8vU.js";
|
|
7
7
|
import { t as createStdioToolServer } from "../tool-server-DEmLr9YY.js";
|
|
8
8
|
import { t as createKbGate } from "../kb-gate-DpaSwXVx.js";
|
|
@@ -3667,6 +3667,13 @@ function retainedCreateMaterial(environment) {
|
|
|
3667
3667
|
...environment.providerOptions === void 0 ? {} : { providerOptionNames: retainedObjectNames(environment.providerOptions) }
|
|
3668
3668
|
};
|
|
3669
3669
|
}
|
|
3670
|
+
/** Add the retained ownership marker without attaching turn identity. */
|
|
3671
|
+
function retainedEnvironmentMetadata(metadata, idempotencyKey) {
|
|
3672
|
+
return {
|
|
3673
|
+
...metadata,
|
|
3674
|
+
retainedIdempotencyKey: idempotencyKey
|
|
3675
|
+
};
|
|
3676
|
+
}
|
|
3670
3677
|
function publicWorkspaceMaterial(workspace) {
|
|
3671
3678
|
return {
|
|
3672
3679
|
...workspace.environment === void 0 ? {} : { environment: workspace.environment },
|
|
@@ -3972,14 +3979,7 @@ async function startRetainedRun(options) {
|
|
|
3972
3979
|
assertRequestedInteractionCapabilities(options.provider.name, options.turn.interactions, providerCapabilities);
|
|
3973
3980
|
const environment = await options.provider.create({
|
|
3974
3981
|
...options.environment,
|
|
3975
|
-
metadata:
|
|
3976
|
-
...options.environment.metadata,
|
|
3977
|
-
retainedIdempotencyKey: options.environment.idempotencyKey,
|
|
3978
|
-
retainedIntentDigest: intent.requestDigest,
|
|
3979
|
-
retainedRunId: intent.runId,
|
|
3980
|
-
sessionId: identity.sessionId,
|
|
3981
|
-
executionId: identity.executionId
|
|
3982
|
-
}
|
|
3982
|
+
metadata: retainedEnvironmentMetadata(options.environment.metadata, options.environment.idempotencyKey)
|
|
3983
3983
|
});
|
|
3984
3984
|
let capabilities;
|
|
3985
3985
|
try {
|
|
@@ -4287,16 +4287,7 @@ async function startRetainedInteractiveRun(options) {
|
|
|
4287
4287
|
...options.environment,
|
|
4288
4288
|
profile,
|
|
4289
4289
|
signal: options.signal,
|
|
4290
|
-
metadata:
|
|
4291
|
-
...options.environment.metadata,
|
|
4292
|
-
retainedIdempotencyKey: options.environment.idempotencyKey,
|
|
4293
|
-
interactiveIdempotencyKey: options.interactiveIdempotencyKey,
|
|
4294
|
-
requestedProfileDigest,
|
|
4295
|
-
interactiveIntentDigest: intent.requestDigest,
|
|
4296
|
-
interactiveRunId: intent.runId,
|
|
4297
|
-
sessionId: identity.sessionId,
|
|
4298
|
-
executionId: identity.executionId
|
|
4299
|
-
}
|
|
4290
|
+
metadata: retainedEnvironmentMetadata(options.environment.metadata, options.environment.idempotencyKey)
|
|
4300
4291
|
}), options.signal);
|
|
4301
4292
|
let capabilities;
|
|
4302
4293
|
try {
|
|
@@ -7100,4 +7091,4 @@ function tail(s) {
|
|
|
7100
7091
|
//#endregion
|
|
7101
7092
|
export { runPersonified as $, pickChampion as A, stopSentinel as At, runBenchmark as B, connectStdioMcp as Bt, analyzeTrace as C, inlineSandboxClient as Ct, chatWorkerSeam as D, completionAuthorizes as Dt, chatTransportExecutor as E, superviseDispatch as Et, strategyAuthorContract as F, renderLeaderboardSvg as Ft, reconnectRetainedRun as G, mcpSecretEnvMetadataKey as Gt, reconnectRetainedInteractiveRun as H, createMcpEnvironment as Ht, strategyAuthorSystemPrompt as I, renderPairwiseMarkdown as It, startRetainedRunInEnvironment as J, secretEnvOfMcpServer as Jt, recoverRetainedRun as K, resolveMcpServerLaunch as Kt, SandboxRunAbortError as L, auditIntent as Lt, selectChampion as M, pairwiseSignificance as Mt, assertStrategyContract as N, renderLeaderboardHtml as Nt, createChatSessionStore as O, deterministicCompletion as Ot, authorStrategy as P, renderLeaderboardMarkdown as Pt, definePersona as Q, openSandboxRun as R, defaultAuditorInstruction as Rt, withUntrackedArtifacts as S, localSandboxClient as St, runCoderChecks as T, loopDispatch as Tt, recoverRetainedInteractiveRun as U, sanitizeMcpToolSchema as Ut, claimRetainedInteractiveControl as V, materializeLocalMcp as Vt, startRetainedInteractiveRun as W, envKeyProvider as Wt, equalKOnCost as X, promotionGate as Y, trajectoryReport as Z, VERIFY_TAIL_CHARS as _, registryScopeAnalyst as _t, localShell as a, renderCorpusToInstructions as at, settledWorkerOut as b, defineLeaderboard as bt, createVerifierEnvironment as c, loopUntil as ct, harvestSurfaceDiffs as d, selectValidWinner as dt, builtinShapes as et, failuresAnalyst as f, verify as ft, NOTE_MAX_CHARS as g, createScopeAnalyst as gt, EVIDENCE_MAX_CHARS as h, buildSteerContext as ht, jjWorkspace as i, InMemoryCorpus as it, runStrategyEvolution as j, leaderboard as jt, discriminatingMeans as k, sentinelCompletion as kt, boxSurfaceReader as l, panel as lt, worktreeFanout as m, assertTraceDerivedFindings as mt, makeFinding$1 as n, registerShape as nt, runInWorkspace as o, fanout as ot, superviseSurface as p, widen as pt, startRetainedRun as q, resolveSecretEnv as qt, gitWorkspace as r, FileCorpus as rt, createWaterfallCollector as s, flatWidenGate as st, computeFindingId$1 as t, createShapeRegistry as tt, fsSurfaceReader as u, pipeline as ut, closingWorkerNote as v, inProcessSandboxClient as vt, patchDelivered as w, loopCampaignDispatch as wt, copyUntrackedIntoClone as x, resolveSandboxClient as xt, composeWorkerEvidence as y, harvestCorpus as yt, printBenchmarkReport as z, McpSpawnFault as zt };
|
|
7102
7093
|
|
|
7103
|
-
//# sourceMappingURL=runtime-
|
|
7094
|
+
//# sourceMappingURL=runtime-WOQ24LYd.js.map
|