@exellix/ai-tasks 9.0.6 → 9.1.1
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/CHANGELOG.md +22 -4
- package/README.md +5 -5
- package/RUNTASK_REQUEST.md +17 -15
- package/dist/compile/compileTaskConfiguration.d.ts.map +1 -1
- package/dist/compile/compileTaskConfiguration.js +3 -0
- package/dist/compile/compileTaskConfiguration.js.map +1 -1
- package/dist/core/task-sdk.d.ts.map +1 -1
- package/dist/core/task-sdk.js +112 -166
- package/dist/core/task-sdk.js.map +1 -1
- package/dist/invocation/types.d.ts +1 -1
- package/dist/narrix/narrixUnitExecution.js +2 -2
- package/dist/narrix/narrixUnitExecution.js.map +1 -1
- package/dist/node-execution/buildRequestFromNodePlan.d.ts +4 -0
- package/dist/node-execution/buildRequestFromNodePlan.d.ts.map +1 -1
- package/dist/node-execution/buildRequestFromNodePlan.js +4 -13
- package/dist/node-execution/buildRequestFromNodePlan.js.map +1 -1
- package/dist/node-execution/compileProfessionalAnswerRequest.d.ts +2 -0
- package/dist/node-execution/compileProfessionalAnswerRequest.d.ts.map +1 -0
- package/dist/node-execution/compileProfessionalAnswerRequest.js +4 -0
- package/dist/node-execution/compileProfessionalAnswerRequest.js.map +1 -0
- package/dist/node-execution/rejectForbiddenWireFields.d.ts +2 -0
- package/dist/node-execution/rejectForbiddenWireFields.d.ts.map +1 -1
- package/dist/node-execution/rejectForbiddenWireFields.js +42 -7
- package/dist/node-execution/rejectForbiddenWireFields.js.map +1 -1
- package/dist/node-execution/resolveUnitModelSelection.d.ts.map +1 -1
- package/dist/node-execution/resolveUnitModelSelection.js +10 -1
- package/dist/node-execution/resolveUnitModelSelection.js.map +1 -1
- package/dist/observability/graphExecutionRunLogContract.d.ts +1 -1
- package/dist/observability/graphExecutionRunLogContract.js +1 -1
- package/dist/post-steps/audit/auditChecklistFuncxEnvelope.d.ts +19 -3
- package/dist/post-steps/audit/auditChecklistFuncxEnvelope.d.ts.map +1 -1
- package/dist/post-steps/audit/auditChecklistFuncxEnvelope.js +7 -1
- package/dist/post-steps/audit/auditChecklistFuncxEnvelope.js.map +1 -1
- package/dist/post-steps/audit/loadAuditTemplates.d.ts +2 -55
- package/dist/post-steps/audit/loadAuditTemplates.d.ts.map +1 -1
- package/dist/post-steps/audit/loadAuditTemplates.js +3 -38
- package/dist/post-steps/audit/loadAuditTemplates.js.map +1 -1
- package/dist/post-steps/audit/parseAuditFuncxOutput.d.ts +8 -0
- package/dist/post-steps/audit/parseAuditFuncxOutput.d.ts.map +1 -0
- package/dist/post-steps/audit/parseAuditFuncxOutput.js +62 -0
- package/dist/post-steps/audit/parseAuditFuncxOutput.js.map +1 -0
- package/dist/post-steps/audit/parseAuditOutput.d.ts +2 -0
- package/dist/post-steps/audit/parseAuditOutput.d.ts.map +1 -1
- package/dist/post-steps/audit/parseAuditOutput.js +56 -0
- package/dist/post-steps/audit/parseAuditOutput.js.map +1 -1
- package/dist/post-steps/audit/runAudit.d.ts.map +1 -1
- package/dist/post-steps/audit/runAudit.js +53 -113
- package/dist/post-steps/audit/runAudit.js.map +1 -1
- package/dist/post-steps/audit/runAuditFuncxCall.d.ts +18 -0
- package/dist/post-steps/audit/runAuditFuncxCall.d.ts.map +1 -0
- package/dist/post-steps/audit/runAuditFuncxCall.js +59 -0
- package/dist/post-steps/audit/runAuditFuncxCall.js.map +1 -0
- package/dist/synthesis/resolveSourceMaterial.d.ts.map +1 -1
- package/dist/synthesis/resolveSourceMaterial.js +14 -0
- package/dist/synthesis/resolveSourceMaterial.js.map +1 -1
- package/dist/types/task-types.d.ts +4 -3
- package/dist/types/task-types.d.ts.map +1 -1
- package/dist/utils/bridgeRunSkillGatewayMemory.d.ts.map +1 -1
- package/dist/utils/bridgeRunSkillGatewayMemory.js +1 -0
- package/dist/utils/bridgeRunSkillGatewayMemory.js.map +1 -1
- package/dist/utils/executionMemoryInputRecord.d.ts +12 -0
- package/dist/utils/executionMemoryInputRecord.d.ts.map +1 -0
- package/dist/utils/executionMemoryInputRecord.js +28 -0
- package/dist/utils/executionMemoryInputRecord.js.map +1 -0
- package/dist/utils/resolveAiProfileModel.d.ts +1 -1
- package/dist/utils/resolveAiProfileModel.d.ts.map +1 -1
- package/dist/utils/skillTemplateVariables.d.ts +3 -2
- package/dist/utils/skillTemplateVariables.d.ts.map +1 -1
- package/dist/utils/skillTemplateVariables.js +3 -2
- package/dist/utils/skillTemplateVariables.js.map +1 -1
- package/dist/validation/validateProfessionalAnswerContract.d.ts +8 -0
- package/dist/validation/validateProfessionalAnswerContract.d.ts.map +1 -0
- package/dist/validation/validateProfessionalAnswerContract.js +45 -0
- package/dist/validation/validateProfessionalAnswerContract.js.map +1 -0
- package/dist/validation/validateRunTaskConfig.d.ts.map +1 -1
- package/dist/validation/validateRunTaskConfig.js +2 -0
- package/dist/validation/validateRunTaskConfig.js.map +1 -1
- package/documenations/record-and-template-variables.md +21 -13
- package/documenations/run-task-execution-flow.md +1 -1
- package/documenations/schemas/v1/run-task-request.json +1 -1
- package/documenations/upstream-feature-requests/README.md +9 -5
- package/documenations/upstream-feature-requests/ai-skills-orchestrator-invoke-contract-5.9.md +1 -1
- package/documenations/upstream-feature-requests/funcx-4.9.13-open-items.md +62 -0
- package/documenations/upstream-feature-requests/funcx-gap-analysis-cr-fr.md +401 -0
- package/documenations/upstream-feature-requests/funcx-pre-post-sidekick-actions.md +1 -0
- package/documenations/upstream-feature-requests/graph-engine-runtask-contract-alignment-investigation.md +370 -0
- package/documenations/upstream-feature-requests/xynthesis-ai-profiles-2.1-import-break.md +2 -2
- package/documenations/upstream-feature-requests/xynthesis-openrouter-wire-model-double-prefix-bug.md +1 -1
- package/documenations/upstream-feature-requests/xynthesis-orchestrator-invoke-contract-4.2.md +1 -1
- package/package.json +14 -17
- package/.docs/DOWNSTREAM_ENV.md +0 -42
- package/.docs/FEEDBACK_TO_CLIENT_DOWNSTREAM_FIXES.md +0 -64
- package/.docs/INTERMEDIATE_STEPS.md +0 -82
- package/.docs/activity-structure.md +0 -31
- package/.docs/ai-task-ai-scoping-spec.md +0 -338
- package/.docs/ai-tasks-model-profile-aliases-7x.md +0 -96
- package/.docs/blockers-and-issues.md +0 -346
- package/.docs/building-runTask-sdk.md +0 -659
- package/.docs/building-skill-execution-orchestrator.md +0 -968
- package/.docs/code-used-before/run-task.txt +0 -39
- package/.docs/code-used-before/task-executor.ts.old +0 -57
- package/.docs/code-used-before/test-run-task.ts.old +0 -42
- package/.docs/code-used-before/types.txt +0 -23
- package/.docs/env-ready-policy.md +0 -40
- package/.docs/flow-io/flow-README.md +0 -76
- package/.docs/flow-io/narrix.md +0 -124
- package/.docs/flow-io/web-scoping.md +0 -135
- package/.docs/flow-io/xynthesis-post.md +0 -154
- package/.docs/flow-io/xynthesis-pre.md +0 -181
- package/.docs/gap-analysis.md +0 -201
- package/.docs/integration-facts-ai-tasks.md +0 -109
- package/.docs/investigation/ai-skills.md +0 -170
- package/.docs/investigation/external-packages-assignments.md +0 -66
- package/.docs/investigation/integration-summary.md +0 -20
- package/.docs/investigation/narrix-catalox.md +0 -29
- package/.docs/investigation/workplan-close-graph-engine-gaps.md +0 -101
- package/.docs/logging-stack.md +0 -30
- package/.docs/memory-narrix-adapter-developer-guide.md +0 -402
- package/.docs/memory-narrix-adapter-requirements.md +0 -112
- package/.docs/narrix-context-consumption-gap.md +0 -184
- package/.docs/narrix-context-downstream-report.md +0 -30
- package/.docs/narrix-ingest-and-packs-library-spec.md +0 -240
- package/.docs/narrix-record-input-current-design.md +0 -48
- package/.docs/pacakge.md +0 -48
- package/.docs/possible-components/README.md +0 -11
- package/.docs/possible-components/integration/README.md +0 -10
- package/.docs/possible-components/integration/gaps-when-merging.md +0 -16
- package/.docs/possible-components/integration/platform.md +0 -54
- package/.docs/possible-components/integration/reintegrate-into-ai-tasks.md +0 -26
- package/.docs/possible-components/integration/roadmap-and-checklists.md +0 -54
- package/.docs/possible-components/post-component/README.md +0 -18
- package/.docs/possible-components/post-component/builder-guide.md +0 -175
- package/.docs/possible-components/post-component/gaps-and-artifacts.md +0 -52
- package/.docs/possible-components/post-component/handler-audit.md +0 -47
- package/.docs/possible-components/post-component/handler-polish.md +0 -41
- package/.docs/possible-components/post-component/unified-protocol.md +0 -59
- package/.docs/possible-components/pre-component/README.md +0 -22
- package/.docs/possible-components/pre-component/builder-guide.md +0 -127
- package/.docs/possible-components/pre-component/gaps-and-artifacts.md +0 -35
- package/.docs/possible-components/pre-component/handler-ai-scoping.md +0 -45
- package/.docs/possible-components/pre-component/handler-narrix-preprocessor.md +0 -49
- package/.docs/possible-components/pre-component/handler-narrix-system2.md +0 -35
- package/.docs/possible-components/pre-component/handler-synthesized-context.md +0 -65
- package/.docs/possible-components/pre-component/handler-web-scope.md +0 -29
- package/.docs/possible-components/pre-component/unified-protocol.md +0 -89
- package/.docs/prefer-openrouter-routing-policy.md +0 -114
- package/.docs/questions-for-ai-skills.md +0 -123
- package/.docs/realtime-narrixing-gap-analysis.md +0 -40
- package/.docs/realtime-narrixing.md +0 -433
- package/.docs/run-context-object.md +0 -32
- package/.docs/session-id-usage.md +0 -26
- package/.docs/skill-library-spec.md +0 -249
- package/.docs/synthesized-context-strategy-spec.md +0 -906
- package/.docs/upstream-issue/2026-03-21_woroces-ai-tasks_ISSUE-006_web-scope-question-from-cni-entity.md +0 -46
- package/.docs/web-scopper-embed.md +0 -93
- package/.docs/xynthesis-wiring-and-io.md +0 -12
- package/documenations/activix-feature-request-identity.md +0 -123
- package/documenations/bug-report-xynthesis-and-synthesis-call.md +0 -217
- package/documenations/feature-request-ai-skills-raw-template-access.md +0 -82
- package/documenations/feature-request-athenix-core-directive.md +0 -145
- package/documenations/feature-request-athenix-token-extraction.md +0 -124
- package/documenations/funcx-upstream-github-issues-draft.md +0 -153
- package/documenations/identity-metadata-contract.md +0 -165
- package/documenations/run-task-single-run-checklist.md +0 -109
- package/documenations/sessions/2026-06-08-subnets-model-resolution/CR-1-no-concrete-wire-in-graph-plans.md +0 -93
- package/documenations/sessions/2026-06-08-subnets-model-resolution/CR-2-skillModel-profile-only-at-storage.md +0 -88
- package/documenations/sessions/2026-06-08-subnets-model-resolution/CR-3-reject-concrete-models-in-catalog-rows.md +0 -76
- package/documenations/sessions/2026-06-08-subnets-model-resolution/FR-1-suggested-profile-in-catalogs.md +0 -96
- package/documenations/sessions/2026-06-08-subnets-model-resolution/FR-2-graph-engine-failure-phase-attribution.md +0 -92
- package/documenations/sessions/2026-06-08-subnets-model-resolution/INVESTIGATION-original-bug.md +0 -182
- package/documenations/sessions/2026-06-08-subnets-model-resolution/PROBLEM.md +0 -236
- package/documenations/sessions/2026-06-08-subnets-model-resolution/README.md +0 -11
- package/documenations/sessions/2026-06-08-subnets-model-resolution/funcx-test-resolveModel.cheapDefaultWireSlug.test.ts +0 -117
- package/documenations/upstream-feature-requests/ai-tasks-wrap-up-after-upstream.md +0 -129
- package/documenations/upstream-feedback-request-shape-clarification.md +0 -101
- package/documenations/web-context-precedence.md +0 -33
- package/documenations/xynthesis-activix-telemetry.md +0 -28
|
@@ -58,16 +58,22 @@ export function buildAuditMergeFuncxEnvelope(input) {
|
|
|
58
58
|
tags.taskId = input.taskId;
|
|
59
59
|
if (input.agentId)
|
|
60
60
|
tags.agentId = input.agentId;
|
|
61
|
+
const args = {};
|
|
62
|
+
if (input.customSynthesisGuidelines?.trim()) {
|
|
63
|
+
args.customSynthesisGuidelines = input.customSynthesisGuidelines.trim();
|
|
64
|
+
}
|
|
61
65
|
return {
|
|
62
66
|
goal: "Merge the best parts of two candidate outputs into a single improved answer.",
|
|
63
67
|
input: {
|
|
68
|
+
originalInput: input.originalInput,
|
|
69
|
+
promptContext: input.promptContext,
|
|
64
70
|
candidateA: input.candidateA,
|
|
65
71
|
candidateB: input.candidateB,
|
|
66
72
|
},
|
|
67
73
|
context: {
|
|
68
74
|
templateCores: ["analysis"],
|
|
69
|
-
task: input.taskDescription,
|
|
70
75
|
},
|
|
76
|
+
...(Object.keys(args).length > 0 ? { args } : {}),
|
|
71
77
|
attribution: {
|
|
72
78
|
functionId: FUNCX_POST_AUDIT_MERGE_ID,
|
|
73
79
|
...(input.jobId ? { runId: input.jobId } : {}),
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"auditChecklistFuncxEnvelope.js","sourceRoot":"","sources":["../../../src/post-steps/audit/auditChecklistFuncxEnvelope.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAEH,OAAO,EAAE,UAAU,EAAE,MAAM,aAAa,CAAC;AAiBzC,MAAM,CAAC,MAAM,6BAA6B,GAAG,sBAAsB,CAAC;AACpE,MAAM,CAAC,MAAM,yBAAyB,GAAG,kBAAkB,CAAC;AAmB5D,gGAAgG;AAChG,MAAM,UAAU,gCAAgC,CAC9C,KAAuC;IAEvC,MAAM,IAAI,GAAG,uEAAuE,CAAC;IACrF,MAAM,IAAI,GAA4B,EAAE,CAAC;IACzC,IAAI,KAAK,CAAC,qBAAqB,EAAE,IAAI,EAAE,EAAE,CAAC;QACxC,IAAI,CAAC,qBAAqB,GAAG,KAAK,CAAC,qBAAqB,CAAC,IAAI,EAAE,CAAC;IAClE,CAAC;IACD,IAAI,KAAK,CAAC,UAAU,KAAK,SAAS;QAAE,IAAI,CAAC,UAAU,GAAG,KAAK,CAAC,UAAU,CAAC;IACvE,IAAI,KAAK,CAAC,KAAK,KAAK,SAAS;QAAE,IAAI,CAAC,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC;IACxD,IAAI,KAAK,CAAC,SAAS,KAAK,SAAS;QAAE,IAAI,CAAC,SAAS,GAAG,KAAK,CAAC,SAAS,CAAC;IAEpE,MAAM,IAAI,GAA2B,EAAE,MAAM,EAAE,UAAU,EAAE,KAAK,EAAE,YAAY,EAAE,CAAC;IACjF,IAAI,KAAK,CAAC,KAAK;QAAE,IAAI,CAAC,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC;IAC1C,IAAI,KAAK,CAAC,MAAM;QAAE,IAAI,CAAC,MAAM,GAAG,KAAK,CAAC,MAAM,CAAC;IAC7C,IAAI,KAAK,CAAC,OAAO;QAAE,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC,OAAO,CAAC;IAEhD,OAAO;QACL,IAAI;QACJ,KAAK,EAAE;YACL,aAAa,EAAE,KAAK,CAAC,aAAa;YAClC,aAAa,EAAE,KAAK,CAAC,aAAa;YAClC,eAAe,EAAE,KAAK,CAAC,eAAe;YACtC,UAAU,EAAE,KAAK,CAAC,UAAU;YAC5B,YAAY,EAAE,KAAK,CAAC,YAAY;YAChC,GAAG,CAAC,KAAK,CAAC,gBAAgB,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,gBAAgB,EAAE,KAAK,CAAC,gBAAgB,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YAC7F,GAAG,CAAC,KAAK,CAAC,aAAa,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,aAAa,EAAE,KAAK,CAAC,aAAa,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;SACrF;QACD,OAAO,EAAE;YACP,aAAa,EAAE,CAAC,UAAU,CAAC;SAC5B;QACD,GAAG,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACjD,WAAW,EAAE;YACX,UAAU,EAAE,6BAA6B;YACzC,GAAG,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YAC9C,GAAG,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,UAAU,EAAE,EAAE,CAAC;YAC7E,GAAG,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YACpD,IAAI;SACL;KACF,CAAC;AACJ,CAAC;
|
|
1
|
+
{"version":3,"file":"auditChecklistFuncxEnvelope.js","sourceRoot":"","sources":["../../../src/post-steps/audit/auditChecklistFuncxEnvelope.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAEH,OAAO,EAAE,UAAU,EAAE,MAAM,aAAa,CAAC;AAiBzC,MAAM,CAAC,MAAM,6BAA6B,GAAG,sBAAsB,CAAC;AACpE,MAAM,CAAC,MAAM,yBAAyB,GAAG,kBAAkB,CAAC;AAmB5D,gGAAgG;AAChG,MAAM,UAAU,gCAAgC,CAC9C,KAAuC;IAEvC,MAAM,IAAI,GAAG,uEAAuE,CAAC;IACrF,MAAM,IAAI,GAA4B,EAAE,CAAC;IACzC,IAAI,KAAK,CAAC,qBAAqB,EAAE,IAAI,EAAE,EAAE,CAAC;QACxC,IAAI,CAAC,qBAAqB,GAAG,KAAK,CAAC,qBAAqB,CAAC,IAAI,EAAE,CAAC;IAClE,CAAC;IACD,IAAI,KAAK,CAAC,UAAU,KAAK,SAAS;QAAE,IAAI,CAAC,UAAU,GAAG,KAAK,CAAC,UAAU,CAAC;IACvE,IAAI,KAAK,CAAC,KAAK,KAAK,SAAS;QAAE,IAAI,CAAC,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC;IACxD,IAAI,KAAK,CAAC,SAAS,KAAK,SAAS;QAAE,IAAI,CAAC,SAAS,GAAG,KAAK,CAAC,SAAS,CAAC;IAEpE,MAAM,IAAI,GAA2B,EAAE,MAAM,EAAE,UAAU,EAAE,KAAK,EAAE,YAAY,EAAE,CAAC;IACjF,IAAI,KAAK,CAAC,KAAK;QAAE,IAAI,CAAC,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC;IAC1C,IAAI,KAAK,CAAC,MAAM;QAAE,IAAI,CAAC,MAAM,GAAG,KAAK,CAAC,MAAM,CAAC;IAC7C,IAAI,KAAK,CAAC,OAAO;QAAE,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC,OAAO,CAAC;IAEhD,OAAO;QACL,IAAI;QACJ,KAAK,EAAE;YACL,aAAa,EAAE,KAAK,CAAC,aAAa;YAClC,aAAa,EAAE,KAAK,CAAC,aAAa;YAClC,eAAe,EAAE,KAAK,CAAC,eAAe;YACtC,UAAU,EAAE,KAAK,CAAC,UAAU;YAC5B,YAAY,EAAE,KAAK,CAAC,YAAY;YAChC,GAAG,CAAC,KAAK,CAAC,gBAAgB,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,gBAAgB,EAAE,KAAK,CAAC,gBAAgB,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YAC7F,GAAG,CAAC,KAAK,CAAC,aAAa,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,aAAa,EAAE,KAAK,CAAC,aAAa,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;SACrF;QACD,OAAO,EAAE;YACP,aAAa,EAAE,CAAC,UAAU,CAAC;SAC5B;QACD,GAAG,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACjD,WAAW,EAAE;YACX,UAAU,EAAE,6BAA6B;YACzC,GAAG,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YAC9C,GAAG,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,UAAU,EAAE,EAAE,CAAC;YAC7E,GAAG,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YACpD,IAAI;SACL;KACF,CAAC;AACJ,CAAC;AAqBD,2EAA2E;AAC3E,MAAM,UAAU,4BAA4B,CAC1C,KAAwC;IAExC,MAAM,IAAI,GAA2B,EAAE,MAAM,EAAE,UAAU,EAAE,KAAK,EAAE,kBAAkB,EAAE,CAAC;IACvF,IAAI,KAAK,CAAC,KAAK;QAAE,IAAI,CAAC,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC;IAC1C,IAAI,KAAK,CAAC,MAAM;QAAE,IAAI,CAAC,MAAM,GAAG,KAAK,CAAC,MAAM,CAAC;IAC7C,IAAI,KAAK,CAAC,OAAO;QAAE,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC,OAAO,CAAC;IAEhD,MAAM,IAAI,GAA4B,EAAE,CAAC;IACzC,IAAI,KAAK,CAAC,yBAAyB,EAAE,IAAI,EAAE,EAAE,CAAC;QAC5C,IAAI,CAAC,yBAAyB,GAAG,KAAK,CAAC,yBAAyB,CAAC,IAAI,EAAE,CAAC;IAC1E,CAAC;IAED,OAAO;QACL,IAAI,EAAE,8EAA8E;QACpF,KAAK,EAAE;YACL,aAAa,EAAE,KAAK,CAAC,aAAa;YAClC,aAAa,EAAE,KAAK,CAAC,aAAa;YAClC,UAAU,EAAE,KAAK,CAAC,UAAU;YAC5B,UAAU,EAAE,KAAK,CAAC,UAAU;SAC7B;QACD,OAAO,EAAE;YACP,aAAa,EAAE,CAAC,UAAU,CAAC;SAC5B;QACD,GAAG,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACjD,WAAW,EAAE;YACX,UAAU,EAAE,yBAAyB;YACrC,GAAG,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YAC9C,GAAG,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,SAAS,EAAE,UAAU,EAAE,EAAE,CAAC;YAC7E,GAAG,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YACpD,IAAI;SACL;KACF,CAAC;AACJ,CAAC"}
|
|
@@ -1,31 +1,11 @@
|
|
|
1
1
|
/**
|
|
2
|
-
* Load audit POST step
|
|
2
|
+
* Load audit POST step feedback-injection template from disk.
|
|
3
|
+
* Base path = AUDIT_TEMPLATES_PATH or cwd/templates/post-steps/audit.
|
|
3
4
|
*/
|
|
4
5
|
export type AuditTemplates = {
|
|
5
|
-
system: string;
|
|
6
|
-
user: string;
|
|
7
6
|
feedbackInjection: string;
|
|
8
|
-
synthesis: string;
|
|
9
7
|
};
|
|
10
8
|
export declare function loadAuditTemplates(): Promise<AuditTemplates>;
|
|
11
|
-
export declare function renderAuditSystem(template: string, data: {
|
|
12
|
-
customGuidelines?: string;
|
|
13
|
-
}): string;
|
|
14
|
-
export declare function renderAuditUser(template: string, data: {
|
|
15
|
-
originalInput: string;
|
|
16
|
-
promptContext: string;
|
|
17
|
-
candidateOutput: string;
|
|
18
|
-
mustChecks: {
|
|
19
|
-
check: string;
|
|
20
|
-
weight: number;
|
|
21
|
-
}[];
|
|
22
|
-
shouldChecks: {
|
|
23
|
-
check: string;
|
|
24
|
-
weight: number;
|
|
25
|
-
}[];
|
|
26
|
-
previousFeedback?: string;
|
|
27
|
-
previousCycle?: number;
|
|
28
|
-
}): string;
|
|
29
9
|
export declare function renderFeedbackInjection(template: string, data: {
|
|
30
10
|
cycle: number;
|
|
31
11
|
maxCycles: number;
|
|
@@ -36,37 +16,4 @@ export declare function renderFeedbackInjection(template: string, data: {
|
|
|
36
16
|
}[];
|
|
37
17
|
overallFeedback: string;
|
|
38
18
|
}): string;
|
|
39
|
-
export declare function renderSynthesis(template: string, data: {
|
|
40
|
-
originalInput: string;
|
|
41
|
-
promptContext: string;
|
|
42
|
-
candidateA: {
|
|
43
|
-
cycle: number;
|
|
44
|
-
score: number;
|
|
45
|
-
output: string;
|
|
46
|
-
failedChecks: {
|
|
47
|
-
kind: string;
|
|
48
|
-
check: string;
|
|
49
|
-
feedback: string;
|
|
50
|
-
}[];
|
|
51
|
-
passedChecks: {
|
|
52
|
-
kind: string;
|
|
53
|
-
check: string;
|
|
54
|
-
}[];
|
|
55
|
-
};
|
|
56
|
-
candidateB: {
|
|
57
|
-
cycle: number;
|
|
58
|
-
score: number;
|
|
59
|
-
output: string;
|
|
60
|
-
failedChecks: {
|
|
61
|
-
kind: string;
|
|
62
|
-
check: string;
|
|
63
|
-
feedback: string;
|
|
64
|
-
}[];
|
|
65
|
-
passedChecks: {
|
|
66
|
-
kind: string;
|
|
67
|
-
check: string;
|
|
68
|
-
}[];
|
|
69
|
-
};
|
|
70
|
-
customSynthesisGuidelines?: string;
|
|
71
|
-
}): string;
|
|
72
19
|
//# sourceMappingURL=loadAuditTemplates.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"loadAuditTemplates.d.ts","sourceRoot":"","sources":["../../../src/post-steps/audit/loadAuditTemplates.ts"],"names":[],"mappings":"AAAA
|
|
1
|
+
{"version":3,"file":"loadAuditTemplates.d.ts","sourceRoot":"","sources":["../../../src/post-steps/audit/loadAuditTemplates.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAQH,MAAM,MAAM,cAAc,GAAG;IAC3B,iBAAiB,EAAE,MAAM,CAAC;CAC3B,CAAC;AAMF,wBAAsB,kBAAkB,IAAI,OAAO,CAAC,cAAc,CAAC,CASlE;AAED,wBAAgB,uBAAuB,CAAC,QAAQ,EAAE,MAAM,EAAE,IAAI,EAAE;IAC9D,KAAK,EAAE,MAAM,CAAC;IACd,SAAS,EAAE,MAAM,CAAC;IAClB,YAAY,EAAE;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,KAAK,EAAE,MAAM,CAAC;QAAC,QAAQ,EAAE,MAAM,CAAA;KAAE,EAAE,CAAC;IAClE,eAAe,EAAE,MAAM,CAAC;CACzB,GAAG,MAAM,CAGT"}
|
|
@@ -1,62 +1,27 @@
|
|
|
1
1
|
/**
|
|
2
|
-
* Load audit POST step
|
|
2
|
+
* Load audit POST step feedback-injection template from disk.
|
|
3
|
+
* Base path = AUDIT_TEMPLATES_PATH or cwd/templates/post-steps/audit.
|
|
3
4
|
*/
|
|
4
5
|
import { readFile } from "fs/promises";
|
|
5
6
|
import { join } from "path";
|
|
6
7
|
import Handlebars from "handlebars";
|
|
7
|
-
const DEFAULT_SYSTEM = `You are a strict quality auditor. Evaluate every check, output structured markdown with "### Checks" and "### Overall feedback" sections.`;
|
|
8
|
-
const DEFAULT_USER = `Original input: {{originalInput}}\n\nContext: {{promptContext}}\n\nCandidate: {{candidateOutput}}\n\nMust: {{mustChecks}}\nShould: {{shouldChecks}}`;
|
|
9
8
|
const DEFAULT_FEEDBACK = `## Audit Feedback (cycle {{cycle}} of {{maxCycles}})\n\n{{overallFeedback}}\n\nFailed: {{failedChecks}}`;
|
|
10
|
-
const DEFAULT_SYNTHESIS = `Merge the best of two candidates. Output only the merged task output.`;
|
|
11
9
|
function getBasePath() {
|
|
12
10
|
return process.env.AUDIT_TEMPLATES_PATH ?? process.cwd();
|
|
13
11
|
}
|
|
14
12
|
export async function loadAuditTemplates() {
|
|
15
13
|
const base = join(getBasePath(), "templates", "post-steps", "audit");
|
|
16
|
-
let system = DEFAULT_SYSTEM;
|
|
17
|
-
let user = DEFAULT_USER;
|
|
18
14
|
let feedbackInjection = DEFAULT_FEEDBACK;
|
|
19
|
-
let synthesis = DEFAULT_SYNTHESIS;
|
|
20
|
-
try {
|
|
21
|
-
system = await readFile(join(base, "system.md"), "utf-8");
|
|
22
|
-
}
|
|
23
|
-
catch {
|
|
24
|
-
/* use default */
|
|
25
|
-
}
|
|
26
|
-
try {
|
|
27
|
-
user = await readFile(join(base, "user.txt"), "utf-8");
|
|
28
|
-
}
|
|
29
|
-
catch {
|
|
30
|
-
/* use default */
|
|
31
|
-
}
|
|
32
15
|
try {
|
|
33
16
|
feedbackInjection = await readFile(join(base, "feedback-injection.md"), "utf-8");
|
|
34
17
|
}
|
|
35
18
|
catch {
|
|
36
19
|
/* use default */
|
|
37
20
|
}
|
|
38
|
-
|
|
39
|
-
synthesis = await readFile(join(base, "synthesis.md"), "utf-8");
|
|
40
|
-
}
|
|
41
|
-
catch {
|
|
42
|
-
/* use default */
|
|
43
|
-
}
|
|
44
|
-
return { system, user, feedbackInjection, synthesis };
|
|
45
|
-
}
|
|
46
|
-
export function renderAuditSystem(template, data) {
|
|
47
|
-
const t = Handlebars.compile(template);
|
|
48
|
-
return t(data);
|
|
49
|
-
}
|
|
50
|
-
export function renderAuditUser(template, data) {
|
|
51
|
-
const t = Handlebars.compile(template);
|
|
52
|
-
return t(data);
|
|
21
|
+
return { feedbackInjection };
|
|
53
22
|
}
|
|
54
23
|
export function renderFeedbackInjection(template, data) {
|
|
55
24
|
const t = Handlebars.compile(template);
|
|
56
25
|
return t(data);
|
|
57
26
|
}
|
|
58
|
-
export function renderSynthesis(template, data) {
|
|
59
|
-
const t = Handlebars.compile(template);
|
|
60
|
-
return t(data);
|
|
61
|
-
}
|
|
62
27
|
//# sourceMappingURL=loadAuditTemplates.js.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"loadAuditTemplates.js","sourceRoot":"","sources":["../../../src/post-steps/audit/loadAuditTemplates.ts"],"names":[],"mappings":"AAAA
|
|
1
|
+
{"version":3,"file":"loadAuditTemplates.js","sourceRoot":"","sources":["../../../src/post-steps/audit/loadAuditTemplates.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAEH,OAAO,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AACvC,OAAO,EAAE,IAAI,EAAE,MAAM,MAAM,CAAC;AAC5B,OAAO,UAAU,MAAM,YAAY,CAAC;AAEpC,MAAM,gBAAgB,GAAG,yGAAyG,CAAC;AAMnI,SAAS,WAAW;IAClB,OAAO,OAAO,CAAC,GAAG,CAAC,oBAAoB,IAAI,OAAO,CAAC,GAAG,EAAE,CAAC;AAC3D,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,kBAAkB;IACtC,MAAM,IAAI,GAAG,IAAI,CAAC,WAAW,EAAE,EAAE,WAAW,EAAE,YAAY,EAAE,OAAO,CAAC,CAAC;IACrE,IAAI,iBAAiB,GAAG,gBAAgB,CAAC;IACzC,IAAI,CAAC;QACH,iBAAiB,GAAG,MAAM,QAAQ,CAAC,IAAI,CAAC,IAAI,EAAE,uBAAuB,CAAC,EAAE,OAAO,CAAC,CAAC;IACnF,CAAC;IAAC,MAAM,CAAC;QACP,iBAAiB;IACnB,CAAC;IACD,OAAO,EAAE,iBAAiB,EAAE,CAAC;AAC/B,CAAC;AAED,MAAM,UAAU,uBAAuB,CAAC,QAAgB,EAAE,IAKzD;IACC,MAAM,CAAC,GAAG,UAAU,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAC;IACvC,OAAO,CAAC,CAAC,IAAI,CAAC,CAAC;AACjB,CAAC"}
|
|
@@ -0,0 +1,8 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* Normalize FuncX `post/audit-checklist` run() output to {@link AuditLLMOutput}.
|
|
3
|
+
* Accepts markdown (preferred) or JSON `{ checks, overall_feedback }` from schema validation.
|
|
4
|
+
*/
|
|
5
|
+
import type { AuditLLMOutput } from "../../types/task-types.js";
|
|
6
|
+
/** Parse audit checklist output from FuncX `run()` value or markdown text. */
|
|
7
|
+
export declare function parseAuditFuncxOutput(value: unknown): AuditLLMOutput;
|
|
8
|
+
//# sourceMappingURL=parseAuditFuncxOutput.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"parseAuditFuncxOutput.d.ts","sourceRoot":"","sources":["../../../src/post-steps/audit/parseAuditFuncxOutput.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAEH,OAAO,KAAK,EAAoB,cAAc,EAAE,MAAM,2BAA2B,CAAC;AAgClF,8EAA8E;AAC9E,wBAAgB,qBAAqB,CAAC,KAAK,EAAE,OAAO,GAAG,cAAc,CAwBpE"}
|
|
@@ -0,0 +1,62 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* Normalize FuncX `post/audit-checklist` run() output to {@link AuditLLMOutput}.
|
|
3
|
+
* Accepts markdown (preferred) or JSON `{ checks, overall_feedback }` from schema validation.
|
|
4
|
+
*/
|
|
5
|
+
import { parseAuditOutputFromMarkdown } from "./parseAuditOutput.js";
|
|
6
|
+
function isRecord(v) {
|
|
7
|
+
return v !== null && typeof v === "object" && !Array.isArray(v);
|
|
8
|
+
}
|
|
9
|
+
function normalizePassed(v) {
|
|
10
|
+
if (typeof v === "boolean")
|
|
11
|
+
return v;
|
|
12
|
+
if (typeof v === "string")
|
|
13
|
+
return v.toLowerCase() === "yes" || v.toLowerCase() === "true";
|
|
14
|
+
return false;
|
|
15
|
+
}
|
|
16
|
+
function parseChecksFromJson(checks) {
|
|
17
|
+
if (!Array.isArray(checks))
|
|
18
|
+
return [];
|
|
19
|
+
const out = [];
|
|
20
|
+
for (const item of checks) {
|
|
21
|
+
if (!isRecord(item))
|
|
22
|
+
continue;
|
|
23
|
+
const check = typeof item.check === "string" ? item.check.trim() : "";
|
|
24
|
+
const kindRaw = typeof item.kind === "string" ? item.kind.toLowerCase() : "should";
|
|
25
|
+
const kind = kindRaw === "must" ? "must" : "should";
|
|
26
|
+
if (!check)
|
|
27
|
+
continue;
|
|
28
|
+
out.push({
|
|
29
|
+
check,
|
|
30
|
+
kind,
|
|
31
|
+
passed: normalizePassed(item.passed),
|
|
32
|
+
feedback: typeof item.feedback === "string" ? item.feedback.trim() : "",
|
|
33
|
+
});
|
|
34
|
+
}
|
|
35
|
+
return out;
|
|
36
|
+
}
|
|
37
|
+
/** Parse audit checklist output from FuncX `run()` value or markdown text. */
|
|
38
|
+
export function parseAuditFuncxOutput(value) {
|
|
39
|
+
if (typeof value === "string") {
|
|
40
|
+
return parseAuditOutputFromMarkdown(value);
|
|
41
|
+
}
|
|
42
|
+
if (!isRecord(value)) {
|
|
43
|
+
return { checks: [], overallFeedback: "" };
|
|
44
|
+
}
|
|
45
|
+
if (typeof value.markdown === "string") {
|
|
46
|
+
return parseAuditOutputFromMarkdown(value.markdown);
|
|
47
|
+
}
|
|
48
|
+
if (typeof value.text === "string") {
|
|
49
|
+
return parseAuditOutputFromMarkdown(value.text);
|
|
50
|
+
}
|
|
51
|
+
const checks = parseChecksFromJson(value.checks);
|
|
52
|
+
const overallFeedback = typeof value.overall_feedback === "string"
|
|
53
|
+
? value.overall_feedback.trim()
|
|
54
|
+
: typeof value.overallFeedback === "string"
|
|
55
|
+
? value.overallFeedback.trim()
|
|
56
|
+
: "";
|
|
57
|
+
if (checks.length > 0 || overallFeedback) {
|
|
58
|
+
return { checks, overallFeedback };
|
|
59
|
+
}
|
|
60
|
+
return parseAuditOutputFromMarkdown(JSON.stringify(value, null, 2));
|
|
61
|
+
}
|
|
62
|
+
//# sourceMappingURL=parseAuditFuncxOutput.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"parseAuditFuncxOutput.js","sourceRoot":"","sources":["../../../src/post-steps/audit/parseAuditFuncxOutput.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAGH,OAAO,EAAE,4BAA4B,EAAE,MAAM,uBAAuB,CAAC;AAErE,SAAS,QAAQ,CAAC,CAAU;IAC1B,OAAO,CAAC,KAAK,IAAI,IAAI,OAAO,CAAC,KAAK,QAAQ,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC;AAClE,CAAC;AAED,SAAS,eAAe,CAAC,CAAU;IACjC,IAAI,OAAO,CAAC,KAAK,SAAS;QAAE,OAAO,CAAC,CAAC;IACrC,IAAI,OAAO,CAAC,KAAK,QAAQ;QAAE,OAAO,CAAC,CAAC,WAAW,EAAE,KAAK,KAAK,IAAI,CAAC,CAAC,WAAW,EAAE,KAAK,MAAM,CAAC;IAC1F,OAAO,KAAK,CAAC;AACf,CAAC;AAED,SAAS,mBAAmB,CAAC,MAAe;IAC1C,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,MAAM,CAAC;QAAE,OAAO,EAAE,CAAC;IACtC,MAAM,GAAG,GAAuB,EAAE,CAAC;IACnC,KAAK,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;QAC1B,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC;YAAE,SAAS;QAC9B,MAAM,KAAK,GAAG,OAAO,IAAI,CAAC,KAAK,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACtE,MAAM,OAAO,GAAG,OAAO,IAAI,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC,CAAC,QAAQ,CAAC;QACnF,MAAM,IAAI,GAAG,OAAO,KAAK,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,QAAQ,CAAC;QACpD,IAAI,CAAC,KAAK;YAAE,SAAS;QACrB,GAAG,CAAC,IAAI,CAAC;YACP,KAAK;YACL,IAAI;YACJ,MAAM,EAAE,eAAe,CAAC,IAAI,CAAC,MAAM,CAAC;YACpC,QAAQ,EAAE,OAAO,IAAI,CAAC,QAAQ,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE;SACxE,CAAC,CAAC;IACL,CAAC;IACD,OAAO,GAAG,CAAC;AACb,CAAC;AAED,8EAA8E;AAC9E,MAAM,UAAU,qBAAqB,CAAC,KAAc;IAClD,IAAI,OAAO,KAAK,KAAK,QAAQ,EAAE,CAAC;QAC9B,OAAO,4BAA4B,CAAC,KAAK,CAAC,CAAC;IAC7C,CAAC;IACD,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,EAAE,CAAC;QACrB,OAAO,EAAE,MAAM,EAAE,EAAE,EAAE,eAAe,EAAE,EAAE,EAAE,CAAC;IAC7C,CAAC;IACD,IAAI,OAAO,KAAK,CAAC,QAAQ,KAAK,QAAQ,EAAE,CAAC;QACvC,OAAO,4BAA4B,CAAC,KAAK,CAAC,QAAQ,CAAC,CAAC;IACtD,CAAC;IACD,IAAI,OAAO,KAAK,CAAC,IAAI,KAAK,QAAQ,EAAE,CAAC;QACnC,OAAO,4BAA4B,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;IAClD,CAAC;IACD,MAAM,MAAM,GAAG,mBAAmB,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;IACjD,MAAM,eAAe,GACnB,OAAO,KAAK,CAAC,gBAAgB,KAAK,QAAQ;QACxC,CAAC,CAAC,KAAK,CAAC,gBAAgB,CAAC,IAAI,EAAE;QAC/B,CAAC,CAAC,OAAO,KAAK,CAAC,eAAe,KAAK,QAAQ;YACzC,CAAC,CAAC,KAAK,CAAC,eAAe,CAAC,IAAI,EAAE;YAC9B,CAAC,CAAC,EAAE,CAAC;IACX,IAAI,MAAM,CAAC,MAAM,GAAG,CAAC,IAAI,eAAe,EAAE,CAAC;QACzC,OAAO,EAAE,MAAM,EAAE,eAAe,EAAE,CAAC;IACrC,CAAC;IACD,OAAO,4BAA4B,CAAC,IAAI,CAAC,SAAS,CAAC,KAAK,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC;AACtE,CAAC"}
|
|
@@ -8,4 +8,6 @@ import type { AuditLLMOutput } from "../../types/task-types.js";
|
|
|
8
8
|
* Expects sections: "### Checks" with blocks of Check/Kind/Passed/Feedback, then "### Overall feedback".
|
|
9
9
|
*/
|
|
10
10
|
export declare function parseAuditOutputFromMarkdown(text: string): AuditLLMOutput;
|
|
11
|
+
/** Parse FuncX `post/audit-checklist` run() output (markdown or JSON) into {@link AuditLLMOutput}. */
|
|
12
|
+
export declare function parseAuditChecklistOutput(value: unknown): AuditLLMOutput;
|
|
11
13
|
//# sourceMappingURL=parseAuditOutput.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"parseAuditOutput.d.ts","sourceRoot":"","sources":["../../../src/post-steps/audit/parseAuditOutput.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAEH,OAAO,KAAK,EAAoB,cAAc,EAAE,MAAM,2BAA2B,CAAC;AAElF;;;GAGG;AACH,wBAAgB,4BAA4B,CAAC,IAAI,EAAE,MAAM,GAAG,cAAc,CA2CzE"}
|
|
1
|
+
{"version":3,"file":"parseAuditOutput.d.ts","sourceRoot":"","sources":["../../../src/post-steps/audit/parseAuditOutput.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAEH,OAAO,KAAK,EAAoB,cAAc,EAAE,MAAM,2BAA2B,CAAC;AAElF;;;GAGG;AACH,wBAAgB,4BAA4B,CAAC,IAAI,EAAE,MAAM,GAAG,cAAc,CA2CzE;AA+BD,sGAAsG;AACtG,wBAAgB,yBAAyB,CAAC,KAAK,EAAE,OAAO,GAAG,cAAc,CAwBxE"}
|
|
@@ -47,4 +47,60 @@ export function parseAuditOutputFromMarkdown(text) {
|
|
|
47
47
|
}
|
|
48
48
|
return { checks, overallFeedback };
|
|
49
49
|
}
|
|
50
|
+
function isRecord(v) {
|
|
51
|
+
return v !== null && typeof v === "object" && !Array.isArray(v);
|
|
52
|
+
}
|
|
53
|
+
function normalizePassed(v) {
|
|
54
|
+
if (typeof v === "boolean")
|
|
55
|
+
return v;
|
|
56
|
+
if (typeof v === "string")
|
|
57
|
+
return v.toLowerCase() === "yes" || v.toLowerCase() === "true";
|
|
58
|
+
return false;
|
|
59
|
+
}
|
|
60
|
+
function parseChecksFromJson(checks) {
|
|
61
|
+
if (!Array.isArray(checks))
|
|
62
|
+
return [];
|
|
63
|
+
const out = [];
|
|
64
|
+
for (const item of checks) {
|
|
65
|
+
if (!isRecord(item))
|
|
66
|
+
continue;
|
|
67
|
+
const check = typeof item.check === "string" ? item.check.trim() : "";
|
|
68
|
+
const kindRaw = typeof item.kind === "string" ? item.kind.toLowerCase() : "should";
|
|
69
|
+
const kind = kindRaw === "must" ? "must" : "should";
|
|
70
|
+
if (!check)
|
|
71
|
+
continue;
|
|
72
|
+
out.push({
|
|
73
|
+
check,
|
|
74
|
+
kind,
|
|
75
|
+
passed: normalizePassed(item.passed),
|
|
76
|
+
feedback: typeof item.feedback === "string" ? item.feedback.trim() : "",
|
|
77
|
+
});
|
|
78
|
+
}
|
|
79
|
+
return out;
|
|
80
|
+
}
|
|
81
|
+
/** Parse FuncX `post/audit-checklist` run() output (markdown or JSON) into {@link AuditLLMOutput}. */
|
|
82
|
+
export function parseAuditChecklistOutput(value) {
|
|
83
|
+
if (typeof value === "string") {
|
|
84
|
+
return parseAuditOutputFromMarkdown(value);
|
|
85
|
+
}
|
|
86
|
+
if (!isRecord(value)) {
|
|
87
|
+
return { checks: [], overallFeedback: "" };
|
|
88
|
+
}
|
|
89
|
+
if (typeof value.markdown === "string") {
|
|
90
|
+
return parseAuditOutputFromMarkdown(value.markdown);
|
|
91
|
+
}
|
|
92
|
+
if (typeof value.text === "string") {
|
|
93
|
+
return parseAuditOutputFromMarkdown(value.text);
|
|
94
|
+
}
|
|
95
|
+
const checks = parseChecksFromJson(value.checks);
|
|
96
|
+
const overallFeedback = typeof value.overall_feedback === "string"
|
|
97
|
+
? value.overall_feedback.trim()
|
|
98
|
+
: typeof value.overallFeedback === "string"
|
|
99
|
+
? value.overallFeedback.trim()
|
|
100
|
+
: "";
|
|
101
|
+
if (checks.length > 0 || overallFeedback) {
|
|
102
|
+
return { checks, overallFeedback };
|
|
103
|
+
}
|
|
104
|
+
return parseAuditOutputFromMarkdown(JSON.stringify(value, null, 2));
|
|
105
|
+
}
|
|
50
106
|
//# sourceMappingURL=parseAuditOutput.js.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"parseAuditOutput.js","sourceRoot":"","sources":["../../../src/post-steps/audit/parseAuditOutput.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAIH;;;GAGG;AACH,MAAM,UAAU,4BAA4B,CAAC,IAAY;IACvD,MAAM,OAAO,GAAG,IAAI,CAAC,IAAI,EAAE,CAAC;IAC5B,MAAM,MAAM,GAAuB,EAAE,CAAC;IACtC,IAAI,eAAe,GAAG,EAAE,CAAC;IAEzB,MAAM,YAAY,GAAG,qDAAqD,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;IACzF,IAAI,YAAY,EAAE,CAAC;QACjB,eAAe,GAAG,YAAY,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;IAC3C,CAAC;IAED,MAAM,aAAa,GAAG,2CAA2C,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;IAChF,IAAI,aAAa,EAAE,CAAC;QAClB,MAAM,KAAK,GAAG,aAAa,CAAC,CAAC,CAAC,CAAC;QAC/B,MAAM,UAAU,GAAG,4KAA4K,CAAC;QAChM,IAAI,CAAyB,CAAC;QAC9B,OAAO,CAAC,CAAC,GAAG,UAAU,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,KAAK,IAAI,EAAE,CAAC;YAC7C,MAAM,CAAC,IAAI,CAAC;gBACV,KAAK,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE;gBAClB,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,EAAuB;gBAC7C,MAAM,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,EAAE,KAAK,KAAK;gBACpC,QAAQ,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE;aACtB,CAAC,CAAC;QACL,CAAC;QACD,IAAI,MAAM,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YACxB,MAAM,cAAc,GAAG,KAAK,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;YAC5C,KAAK,MAAM,CAAC,IAAI,cAAc,EAAE,CAAC;gBAC/B,MAAM,UAAU,GAAG,2BAA2B,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;gBACvD,MAAM,SAAS,GAAG,gCAAgC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;gBAC3D,MAAM,WAAW,GAAG,6BAA6B,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;gBAC1D,MAAM,aAAa,GAAG,+BAA+B,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;gBAC9D,IAAI,UAAU,IAAI,SAAS,EAAE,CAAC;oBAC5B,MAAM,CAAC,IAAI,CAAC;wBACV,KAAK,EAAE,UAAU,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE;wBAC3B,IAAI,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,WAAW,EAAuB;wBACrD,MAAM,EAAE,WAAW,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,CAAC,WAAW,EAAE,KAAK,KAAK,CAAC,CAAC,CAAC,KAAK;wBACpE,QAAQ,EAAE,aAAa,CAAC,CAAC,CAAC,aAAa,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE;qBACvD,CAAC,CAAC;gBACL,CAAC;YACH,CAAC;QACH,CAAC;IACH,CAAC;IAED,OAAO,EAAE,MAAM,EAAE,eAAe,EAAE,CAAC;AACrC,CAAC"}
|
|
1
|
+
{"version":3,"file":"parseAuditOutput.js","sourceRoot":"","sources":["../../../src/post-steps/audit/parseAuditOutput.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAIH;;;GAGG;AACH,MAAM,UAAU,4BAA4B,CAAC,IAAY;IACvD,MAAM,OAAO,GAAG,IAAI,CAAC,IAAI,EAAE,CAAC;IAC5B,MAAM,MAAM,GAAuB,EAAE,CAAC;IACtC,IAAI,eAAe,GAAG,EAAE,CAAC;IAEzB,MAAM,YAAY,GAAG,qDAAqD,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;IACzF,IAAI,YAAY,EAAE,CAAC;QACjB,eAAe,GAAG,YAAY,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;IAC3C,CAAC;IAED,MAAM,aAAa,GAAG,2CAA2C,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;IAChF,IAAI,aAAa,EAAE,CAAC;QAClB,MAAM,KAAK,GAAG,aAAa,CAAC,CAAC,CAAC,CAAC;QAC/B,MAAM,UAAU,GAAG,4KAA4K,CAAC;QAChM,IAAI,CAAyB,CAAC;QAC9B,OAAO,CAAC,CAAC,GAAG,UAAU,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,KAAK,IAAI,EAAE,CAAC;YAC7C,MAAM,CAAC,IAAI,CAAC;gBACV,KAAK,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE;gBAClB,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,EAAuB;gBAC7C,MAAM,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,WAAW,EAAE,KAAK,KAAK;gBACpC,QAAQ,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE;aACtB,CAAC,CAAC;QACL,CAAC;QACD,IAAI,MAAM,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YACxB,MAAM,cAAc,GAAG,KAAK,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;YAC5C,KAAK,MAAM,CAAC,IAAI,cAAc,EAAE,CAAC;gBAC/B,MAAM,UAAU,GAAG,2BAA2B,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;gBACvD,MAAM,SAAS,GAAG,gCAAgC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;gBAC3D,MAAM,WAAW,GAAG,6BAA6B,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;gBAC1D,MAAM,aAAa,GAAG,+BAA+B,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;gBAC9D,IAAI,UAAU,IAAI,SAAS,EAAE,CAAC;oBAC5B,MAAM,CAAC,IAAI,CAAC;wBACV,KAAK,EAAE,UAAU,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE;wBAC3B,IAAI,EAAE,SAAS,CAAC,CAAC,CAAC,CAAC,WAAW,EAAuB;wBACrD,MAAM,EAAE,WAAW,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,CAAC,WAAW,EAAE,KAAK,KAAK,CAAC,CAAC,CAAC,KAAK;wBACpE,QAAQ,EAAE,aAAa,CAAC,CAAC,CAAC,aAAa,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE;qBACvD,CAAC,CAAC;gBACL,CAAC;YACH,CAAC;QACH,CAAC;IACH,CAAC;IAED,OAAO,EAAE,MAAM,EAAE,eAAe,EAAE,CAAC;AACrC,CAAC;AAED,SAAS,QAAQ,CAAC,CAAU;IAC1B,OAAO,CAAC,KAAK,IAAI,IAAI,OAAO,CAAC,KAAK,QAAQ,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC;AAClE,CAAC;AAED,SAAS,eAAe,CAAC,CAAU;IACjC,IAAI,OAAO,CAAC,KAAK,SAAS;QAAE,OAAO,CAAC,CAAC;IACrC,IAAI,OAAO,CAAC,KAAK,QAAQ;QAAE,OAAO,CAAC,CAAC,WAAW,EAAE,KAAK,KAAK,IAAI,CAAC,CAAC,WAAW,EAAE,KAAK,MAAM,CAAC;IAC1F,OAAO,KAAK,CAAC;AACf,CAAC;AAED,SAAS,mBAAmB,CAAC,MAAe;IAC1C,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,MAAM,CAAC;QAAE,OAAO,EAAE,CAAC;IACtC,MAAM,GAAG,GAAuB,EAAE,CAAC;IACnC,KAAK,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;QAC1B,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC;YAAE,SAAS;QAC9B,MAAM,KAAK,GAAG,OAAO,IAAI,CAAC,KAAK,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACtE,MAAM,OAAO,GAAG,OAAO,IAAI,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC,CAAC,CAAC,QAAQ,CAAC;QACnF,MAAM,IAAI,GAAG,OAAO,KAAK,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,QAAQ,CAAC;QACpD,IAAI,CAAC,KAAK;YAAE,SAAS;QACrB,GAAG,CAAC,IAAI,CAAC;YACP,KAAK;YACL,IAAI;YACJ,MAAM,EAAE,eAAe,CAAC,IAAI,CAAC,MAAM,CAAC;YACpC,QAAQ,EAAE,OAAO,IAAI,CAAC,QAAQ,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE;SACxE,CAAC,CAAC;IACL,CAAC;IACD,OAAO,GAAG,CAAC;AACb,CAAC;AAED,sGAAsG;AACtG,MAAM,UAAU,yBAAyB,CAAC,KAAc;IACtD,IAAI,OAAO,KAAK,KAAK,QAAQ,EAAE,CAAC;QAC9B,OAAO,4BAA4B,CAAC,KAAK,CAAC,CAAC;IAC7C,CAAC;IACD,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,EAAE,CAAC;QACrB,OAAO,EAAE,MAAM,EAAE,EAAE,EAAE,eAAe,EAAE,EAAE,EAAE,CAAC;IAC7C,CAAC;IACD,IAAI,OAAO,KAAK,CAAC,QAAQ,KAAK,QAAQ,EAAE,CAAC;QACvC,OAAO,4BAA4B,CAAC,KAAK,CAAC,QAAQ,CAAC,CAAC;IACtD,CAAC;IACD,IAAI,OAAO,KAAK,CAAC,IAAI,KAAK,QAAQ,EAAE,CAAC;QACnC,OAAO,4BAA4B,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;IAClD,CAAC;IACD,MAAM,MAAM,GAAG,mBAAmB,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;IACjD,MAAM,eAAe,GACnB,OAAO,KAAK,CAAC,gBAAgB,KAAK,QAAQ;QACxC,CAAC,CAAC,KAAK,CAAC,gBAAgB,CAAC,IAAI,EAAE;QAC/B,CAAC,CAAC,OAAO,KAAK,CAAC,eAAe,KAAK,QAAQ;YACzC,CAAC,CAAC,KAAK,CAAC,eAAe,CAAC,IAAI,EAAE;YAC9B,CAAC,CAAC,EAAE,CAAC;IACX,IAAI,MAAM,CAAC,MAAM,GAAG,CAAC,IAAI,eAAe,EAAE,CAAC;QACzC,OAAO,EAAE,MAAM,EAAE,eAAe,EAAE,CAAC;IACrC,CAAC;IACD,OAAO,4BAA4B,CAAC,IAAI,CAAC,SAAS,CAAC,KAAK,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC;AACtE,CAAC"}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"runAudit.d.ts","sourceRoot":"","sources":["../../../src/post-steps/audit/runAudit.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,KAAK,EAAE,cAAc,EAAE,eAAe,EAAE,MAAM,sBAAsB,CAAC;AAC5E,OAAO,KAAK,EACV,WAAW,EAEX,qBAAqB,EAErB,gBAAgB,EACjB,MAAM,2BAA2B,CAAC;AACnC,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,mCAAmC,CAAC;
|
|
1
|
+
{"version":3,"file":"runAudit.d.ts","sourceRoot":"","sources":["../../../src/post-steps/audit/runAudit.ts"],"names":[],"mappings":"AAAA;;GAEG;AAEH,OAAO,KAAK,EAAE,cAAc,EAAE,eAAe,EAAE,MAAM,sBAAsB,CAAC;AAC5E,OAAO,KAAK,EACV,WAAW,EAEX,qBAAqB,EAErB,gBAAgB,EACjB,MAAM,2BAA2B,CAAC;AACnC,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,mCAAmC,CAAC;AAe7E,MAAM,MAAM,uBAAuB,CAAC,OAAO,GAAG,GAAG,IAAI;IACnD,OAAO,EAAE,cAAc,CAAC;IACxB,UAAU,EAAE,eAAe,CAAC,OAAO,CAAC,CAAC;IACrC,eAAe,EAAE,MAAM,CAAC;IACxB,MAAM,EAAE,WAAW,CAAC;IACpB,SAAS,EAAE,CAAC,eAAe,EAAE,MAAM,KAAK,OAAO,CAAC,eAAe,CAAC,OAAO,CAAC,CAAC,CAAC;IAC1E,iEAAiE;IACjE,cAAc,CAAC,EAAE,mBAAmB,GAAG,IAAI,CAAC;CAC7C,CAAC;AAEF,MAAM,MAAM,sBAAsB,GAAG;IACnC,UAAU,EAAE,MAAM,CAAC;IACnB,QAAQ,EAAE,qBAAqB,CAAC;IAChC,iBAAiB,EAAE,gBAAgB,EAAE,CAAC;CACvC,CAAC;AA2CF,wBAAsB,gBAAgB,CAAC,OAAO,GAAG,GAAG,EAClD,GAAG,EAAE,uBAAuB,CAAC,OAAO,CAAC,GACpC,OAAO,CAAC,sBAAsB,CAAC,CAmXjC"}
|
|
@@ -2,11 +2,9 @@
|
|
|
2
2
|
* Audit POST step: quality-gate loop with re-runs and optional synthesis.
|
|
3
3
|
*/
|
|
4
4
|
import { llmCallEnvPrefix, resolveLlmCall, } from "../resolvePostStepConfig.js";
|
|
5
|
-
import { loadAuditTemplates,
|
|
6
|
-
import {
|
|
7
|
-
import {
|
|
8
|
-
import { runPostStepLlmCall } from "../../internal/runPostStepLlmCall.js";
|
|
9
|
-
import { buildXynthesisWorkScopeIdentity } from "../../internal/buildXynthesisWorkScopeIdentity.js";
|
|
5
|
+
import { loadAuditTemplates, renderFeedbackInjection } from "./loadAuditTemplates.js";
|
|
6
|
+
import { parseAuditChecklistOutput } from "./parseAuditOutput.js";
|
|
7
|
+
import { runAuditChecklistFuncxCall, runAuditMergeFuncxCall } from "./runAuditFuncxCall.js";
|
|
10
8
|
import { createNxCache } from "nx-cache";
|
|
11
9
|
import { createHash } from "crypto";
|
|
12
10
|
import { getPostActionModelFromSlots } from "../../utils/routeModelConfigSlots.js";
|
|
@@ -81,65 +79,28 @@ export async function runAuditPostStep(ctx) {
|
|
|
81
79
|
// pushed here. Surfaced on `metadata.llmCalls` (see end of function). Outside trace mode the array
|
|
82
80
|
// remains empty and is omitted from metadata.
|
|
83
81
|
const llmCalls = [];
|
|
84
|
-
const observationSink = traceCollector
|
|
85
|
-
? (obs) => {
|
|
86
|
-
llmCalls.push(obs);
|
|
87
|
-
}
|
|
88
|
-
: undefined;
|
|
89
82
|
let currentOutput = typeof mainResult.rawText === "string"
|
|
90
83
|
? mainResult.rawText
|
|
91
84
|
: JSON.stringify(mainResult.parsed ?? mainResult.rawText ?? "", null, 2);
|
|
92
85
|
const originalInputStr = typeof request.input === "string" ? request.input : JSON.stringify(request.input ?? {}, null, 2);
|
|
93
|
-
const baseTraceMeta = () => ({
|
|
94
|
-
jobId: request.jobId,
|
|
95
|
-
taskId: request.taskId,
|
|
96
|
-
skillKey: request.skillKey,
|
|
97
|
-
phase: "audit",
|
|
98
|
-
postStepType: "audit",
|
|
99
|
-
});
|
|
100
|
-
const workScopeIdentity = buildXynthesisWorkScopeIdentity({
|
|
101
|
-
jobId: request.jobId,
|
|
102
|
-
taskId: request.taskId,
|
|
103
|
-
agentId: request.agentId,
|
|
104
|
-
graphId: request.graphId,
|
|
105
|
-
nodeId: request.nodeId,
|
|
106
|
-
prevNodeId: request.prevNodeId,
|
|
107
|
-
coreSkillId: request.coreSkillId,
|
|
108
|
-
masterSkillId: request.masterSkillId,
|
|
109
|
-
identity: request.identity,
|
|
110
|
-
});
|
|
111
|
-
const makeTraceTask = (opts) => ({
|
|
112
|
-
taskType: "post-execution",
|
|
113
|
-
details: opts.details,
|
|
114
|
-
modelUsed: opts.modelUsed,
|
|
115
|
-
metadata: {
|
|
116
|
-
...baseTraceMeta(),
|
|
117
|
-
step: { phase: "post", type: "audit", stepId: opts.stepId },
|
|
118
|
-
},
|
|
119
|
-
});
|
|
120
86
|
const callAudit = async (params) => {
|
|
121
|
-
const {
|
|
87
|
+
const { funcxChecklistInput, llmCallOverride, stepId } = params;
|
|
122
88
|
const effective = { ...auditLlmCall, ...(llmCallOverride ?? {}) };
|
|
123
|
-
|
|
124
|
-
|
|
125
|
-
|
|
126
|
-
|
|
89
|
+
const { text } = await runAuditChecklistFuncxCall({
|
|
90
|
+
originalInput: funcxChecklistInput.originalInput,
|
|
91
|
+
promptContext: funcxChecklistInput.promptContext,
|
|
92
|
+
candidateOutput: funcxChecklistInput.candidateOutput,
|
|
93
|
+
mustChecks,
|
|
94
|
+
shouldChecks,
|
|
95
|
+
previousFeedback: funcxChecklistInput.previousFeedback,
|
|
96
|
+
previousCycle: funcxChecklistInput.previousCycle,
|
|
97
|
+
customAuditGuidelines: config.customAuditGuidelines,
|
|
127
98
|
jobId: postStepJobId,
|
|
128
99
|
taskId: request.taskId,
|
|
129
100
|
agentId: "audit-evaluator",
|
|
130
|
-
|
|
131
|
-
|
|
132
|
-
|
|
133
|
-
stepId,
|
|
134
|
-
details: "audit evaluator call",
|
|
135
|
-
modelUsed: effective.model ?? null,
|
|
136
|
-
})
|
|
137
|
-
: undefined,
|
|
138
|
-
observationSink,
|
|
139
|
-
stage: stage ?? "post-audit",
|
|
140
|
-
stepId,
|
|
141
|
-
workScopeIdentity,
|
|
142
|
-
});
|
|
101
|
+
}, effective);
|
|
102
|
+
void stepId;
|
|
103
|
+
return text;
|
|
143
104
|
};
|
|
144
105
|
let auditCallFailures = 0;
|
|
145
106
|
/** Prompt context MAIN used for `currentOutput` on this audit cycle (matches `reRunMain` override after prior failures). */
|
|
@@ -163,18 +124,13 @@ export async function runAuditPostStep(ctx) {
|
|
|
163
124
|
else {
|
|
164
125
|
rawAuditText = await callAudit({
|
|
165
126
|
stepId: `audit-evaluator-cycle-${cycle}`,
|
|
166
|
-
|
|
167
|
-
customGuidelines: config.customAuditGuidelines,
|
|
168
|
-
}),
|
|
169
|
-
userPrompt: renderAuditUser(templates.user, {
|
|
127
|
+
funcxChecklistInput: {
|
|
170
128
|
originalInput: originalInputStr,
|
|
171
129
|
promptContext: mainContextForEvaluator,
|
|
172
130
|
candidateOutput: currentOutput,
|
|
173
|
-
mustChecks,
|
|
174
|
-
shouldChecks,
|
|
175
131
|
previousFeedback: cycle > 1 ? auditResults[auditResults.length - 1]?.overallFeedback : undefined,
|
|
176
132
|
previousCycle: cycle > 1 ? cycle - 1 : undefined,
|
|
177
|
-
}
|
|
133
|
+
},
|
|
178
134
|
});
|
|
179
135
|
auditCache.write(cacheKey, rawAuditText, { ttlMs: 300_000 });
|
|
180
136
|
}
|
|
@@ -182,18 +138,13 @@ export async function runAuditPostStep(ctx) {
|
|
|
182
138
|
else {
|
|
183
139
|
rawAuditText = await callAudit({
|
|
184
140
|
stepId: `audit-evaluator-cycle-${cycle}`,
|
|
185
|
-
|
|
186
|
-
customGuidelines: config.customAuditGuidelines,
|
|
187
|
-
}),
|
|
188
|
-
userPrompt: renderAuditUser(templates.user, {
|
|
141
|
+
funcxChecklistInput: {
|
|
189
142
|
originalInput: originalInputStr,
|
|
190
143
|
promptContext: mainContextForEvaluator,
|
|
191
144
|
candidateOutput: currentOutput,
|
|
192
|
-
mustChecks,
|
|
193
|
-
shouldChecks,
|
|
194
145
|
previousFeedback: cycle > 1 ? auditResults[auditResults.length - 1]?.overallFeedback : undefined,
|
|
195
146
|
previousCycle: cycle > 1 ? cycle - 1 : undefined,
|
|
196
|
-
}
|
|
147
|
+
},
|
|
197
148
|
});
|
|
198
149
|
}
|
|
199
150
|
}
|
|
@@ -227,7 +178,7 @@ export async function runAuditPostStep(ctx) {
|
|
|
227
178
|
}
|
|
228
179
|
let parsed;
|
|
229
180
|
try {
|
|
230
|
-
parsed =
|
|
181
|
+
parsed = parseAuditChecklistOutput(rawAuditText);
|
|
231
182
|
}
|
|
232
183
|
catch {
|
|
233
184
|
parsed = { checks: [], overallFeedback: "Parse failed." };
|
|
@@ -278,50 +229,42 @@ export async function runAuditPostStep(ctx) {
|
|
|
278
229
|
if (selectionStrategy === "synthesis" && candidates.length >= 2) {
|
|
279
230
|
const sorted = [...candidates].sort((a, b) => b.weightedScore - a.weightedScore);
|
|
280
231
|
const [a, b] = sorted.slice(0, 2);
|
|
281
|
-
const
|
|
282
|
-
|
|
283
|
-
|
|
284
|
-
|
|
285
|
-
|
|
286
|
-
|
|
287
|
-
|
|
288
|
-
|
|
289
|
-
|
|
290
|
-
|
|
291
|
-
|
|
292
|
-
|
|
293
|
-
|
|
294
|
-
|
|
295
|
-
|
|
296
|
-
|
|
297
|
-
|
|
298
|
-
|
|
299
|
-
|
|
232
|
+
const candidateAEnvelope = {
|
|
233
|
+
cycle: a.cycle,
|
|
234
|
+
score: a.weightedScore,
|
|
235
|
+
output: a.output,
|
|
236
|
+
failedChecks: a.auditResult.checks
|
|
237
|
+
.filter((c) => !c.passed)
|
|
238
|
+
.map((c) => ({ kind: c.kind, check: c.check, feedback: c.feedback })),
|
|
239
|
+
passedChecks: a.auditResult.checks
|
|
240
|
+
.filter((c) => c.passed)
|
|
241
|
+
.map((c) => ({ kind: c.kind, check: c.check })),
|
|
242
|
+
};
|
|
243
|
+
const candidateBEnvelope = {
|
|
244
|
+
cycle: b.cycle,
|
|
245
|
+
score: b.weightedScore,
|
|
246
|
+
output: b.output,
|
|
247
|
+
failedChecks: b.auditResult.checks
|
|
248
|
+
.filter((c) => !c.passed)
|
|
249
|
+
.map((c) => ({ kind: c.kind, check: c.check, feedback: c.feedback })),
|
|
250
|
+
passedChecks: b.auditResult.checks
|
|
251
|
+
.filter((c) => c.passed)
|
|
252
|
+
.map((c) => ({ kind: c.kind, check: c.check })),
|
|
253
|
+
};
|
|
300
254
|
let synthesizedOutput;
|
|
301
255
|
let mergeFailed = false;
|
|
302
256
|
let mergeError = null;
|
|
303
257
|
try {
|
|
304
|
-
|
|
305
|
-
|
|
306
|
-
|
|
307
|
-
|
|
308
|
-
|
|
309
|
-
|
|
258
|
+
synthesizedOutput = await runAuditMergeFuncxCall({
|
|
259
|
+
originalInput: originalInputStr,
|
|
260
|
+
promptContext: contextMarkdown,
|
|
261
|
+
candidateA: candidateAEnvelope,
|
|
262
|
+
candidateB: candidateBEnvelope,
|
|
263
|
+
customSynthesisGuidelines: config.customSynthesisGuidelines,
|
|
310
264
|
jobId: postStepJobId,
|
|
311
265
|
taskId: request.taskId,
|
|
312
266
|
agentId: "audit-synthesis-merge",
|
|
313
|
-
|
|
314
|
-
traceCollector,
|
|
315
|
-
traceTask: traceCollector
|
|
316
|
-
? makeTraceTask({
|
|
317
|
-
stepId: "audit-synthesis-merge",
|
|
318
|
-
details: "audit synthesis merge",
|
|
319
|
-
modelUsed: synthLlmCall.model ?? null,
|
|
320
|
-
})
|
|
321
|
-
: undefined,
|
|
322
|
-
observationSink,
|
|
323
|
-
});
|
|
324
|
-
synthesizedOutput = text.trim();
|
|
267
|
+
}, synthLlmCall);
|
|
325
268
|
}
|
|
326
269
|
catch (err) {
|
|
327
270
|
// Previously the merge failure was silently swallowed; now we still fall back to the best
|
|
@@ -343,16 +286,13 @@ export async function runAuditPostStep(ctx) {
|
|
|
343
286
|
}
|
|
344
287
|
const synthAuditText = await callAudit({
|
|
345
288
|
stepId: "audit-evaluator-synth-candidate",
|
|
346
|
-
|
|
347
|
-
userPrompt: renderAuditUser(templates.user, {
|
|
289
|
+
funcxChecklistInput: {
|
|
348
290
|
originalInput: originalInputStr,
|
|
349
291
|
promptContext: contextMarkdown,
|
|
350
292
|
candidateOutput: synthesizedOutput,
|
|
351
|
-
|
|
352
|
-
shouldChecks,
|
|
353
|
-
}),
|
|
293
|
+
},
|
|
354
294
|
});
|
|
355
|
-
const synthParsed =
|
|
295
|
+
const synthParsed = parseAuditChecklistOutput(synthAuditText);
|
|
356
296
|
const synthScore = computeWeightedScore(synthParsed.checks, gateway);
|
|
357
297
|
const synthResult = { ...synthParsed, weightedScore: synthScore };
|
|
358
298
|
auditResults.push(synthResult);
|