@tangle-network/agent-eval 0.30.0 → 0.31.1

package/dist/traces.js

@@ -17,13 +17,23 @@ import {
   TraceNotFoundError,
   analyzeTraces,
   buildTraceAnalystTools,
+  buildTraceInsightContext,
+  buildTraceInsightPrompt,
   createReplayFetch,
+  defaultTraceInsightPanel,
+  describeTraceInsightScope,
+  domainEvidencePattern,
   exportRunAsOtlp,
+  inferDomainKeywords,
   iterateRawCalls,
+  planTraceInsightQuestions,
   redactString,
   redactValue,
-  traceAnalystFunctionGroup
-} from "./chunk-UW4NOOZI.js";
+  scoreTraceInsightReadiness,
+  tokenizeDomainWords,
+  traceAnalystFunctionGroup,
+  traceAnalystOnRunComplete
+} from "./chunk-HIO4UIS5.js";
 import {
   aggregateLlm,
   argHash,
@@ -47,7 +57,7 @@ import {
   RunIntegrityError,
   assertRunCaptured,
   throwIfRunIncomplete
-} from "./chunk-KTGTIOFD.js";
+} from "./chunk-UBPIXOC4.js";
 import {
   FileSystemRawProviderSink,
   InMemoryRawProviderSink,
@@ -60,304 +70,8 @@ import {
   llmSpanFromProvider
 } from "./chunk-TVVP3ZZQ.js";
 import "./chunk-VSMTAMNK.js";
-import "./chunk-NG236HPC.js";
+import "./chunk-QYJT52YW.js";
 import "./chunk-PZ5AY32C.js";
-
-// src/trace-analyst/hook.ts
-var DEFAULT_QUESTION = "Summarise what happened in this run. Surface any failure modes, surprising findings, or evidence that the run's verdict is wrong.";
-function traceAnalystOnRunComplete(opts) {
-  return async (ctx) => {
-    if (opts.shouldRun && !opts.shouldRun(ctx)) return;
-    const source = opts.analyze.source;
-    if (source === void 0) {
-      await ctx.store.appendEvent({
-        eventId: `analyst-skip-${ctx.runId}`,
-        runId: ctx.runId,
-        kind: "log",
-        timestamp: Date.now(),
-        payload: { source: "trace_analyst_hook", reason: "no source configured" }
-      });
-      return;
-    }
-    const result = await analyzeTraces({ question: opts.question ?? DEFAULT_QUESTION }, {
-      ...opts.analyze,
-      source
-    });
-    if (opts.save) await opts.save(result, ctx);
-    if (opts.gateOn && !opts.gateOn(result, ctx)) {
-      await ctx.store.appendEvent({
-        eventId: `analyst-gate-${ctx.runId}`,
-        runId: ctx.runId,
-        kind: "log",
-        timestamp: Date.now(),
-        payload: {
-          source: "trace_analyst_hook",
-          reason: "analyst_gate_failed",
-          findings: result.findings
-        }
-      });
-    }
-  };
-}
-
-// src/trace-analyst/insights.ts
-var DOMAIN_STOP_WORDS = /* @__PURE__ */ new Set([
-  "and",
-  "advanced",
-  "app",
-  "build",
-  "create",
-  "easy",
-  "expert",
-  "extreme",
-  "for",
-  "from",
-  "hard",
-  "implementation",
-  "integrate",
-  "medium",
-  "project",
-  "task",
-  "the",
-  "this",
-  "with",
-  "workflow"
-]);
-function tokenizeDomainWords(value) {
-  return [...value.matchAll(/[A-Za-z][A-Za-z0-9.+#-]{2,}/g)].map((match) => match[0].toLowerCase()).filter((word) => !DOMAIN_STOP_WORDS.has(word));
-}
-function inferDomainKeywords(suite) {
-  const suiteWords = new Set(tokenizeDomainWords(`${suite.name} ${suite.collectionId ?? ""}`));
-  const source = [
-    suite.name,
-    suite.collectionId ?? "",
-    ...suite.tasks.flatMap((task) => [
-      task.id,
-      task.name,
-      task.prompt ?? "",
-      task.difficulty ?? "",
-      ...task.tags ?? [],
-      ...task.gaps ?? []
-    ])
-  ].join(" ");
-  const counts = /* @__PURE__ */ new Map();
-  for (const word of tokenizeDomainWords(source)) counts.set(word, (counts.get(word) ?? 0) + 1);
-  return [...counts.entries()].filter(([word, count]) => count >= 2 || suiteWords.has(word)).sort((a, b) => b[1] - a[1] || a[0].localeCompare(b[0])).map(([word]) => word).slice(0, 18);
-}
-function domainEvidencePattern(keywords) {
-  const escaped = keywords.filter((keyword) => keyword.length >= 3).map((keyword) => keyword.replace(/[.*+?^${}()|[\]\\]/g, "\\$&"));
-  return escaped.length > 0 ? new RegExp(`(?<![A-Za-z0-9])(?:${escaped.join("|")})(?![A-Za-z0-9])`, "i") : /(?<![A-Za-z0-9])(?:sdk|api|css|dns|xml|provider|client|service|integration|webhook|transaction|auth|oauth|graphql|rest)(?![A-Za-z0-9])/i;
-}
-function describeTraceInsightScope(suite) {
-  const taskLabel = suite.tasks.length === 1 ? "1 implementation task" : `${suite.tasks.length} implementation tasks`;
-  const tags = /* @__PURE__ */ new Map();
-  for (const task of suite.tasks) {
-    for (const tag of task.tags ?? []) tags.set(tag, (tags.get(tag) ?? 0) + 1);
-  }
-  const topTags = [...tags.entries()].sort((a, b) => b[1] - a[1] || a[0].localeCompare(b[0])).slice(0, 8).map(([tag]) => tag);
-  if (topTags.length > 0) return `${taskLabel} across ${topTags.join(", ")}.`;
-  const difficulties = [
-    ...new Set(
-      suite.tasks.map((task) => task.difficulty).filter((value) => Boolean(value))
-    )
-  ].join(", ");
-  return `${taskLabel} across ${difficulties || "the selected benchmark scope"}.`;
-}
-function planTraceInsightQuestions(input) {
-  const hasFailures = input.suite.tasks.some((task) => task.outcome && task.outcome !== "satisfied");
-  const hasMultipleShots = input.suite.tasks.some(
-    (task) => (task.gaps ?? []).some((gap) => /shot|review|retry|continue/i.test(gap))
-  );
-  const questions = [
-    {
-      id: "execution-path",
-      question: "What did the worker actually do before the first meaningful implementation edit?",
-      why: "Separates grounded execution from polished but shallow output."
-    },
-    {
-      id: "research-grounding",
-      question: "Did the worker inspect docs, source, examples, or package references before committing to an implementation path?",
-      why: "Identifies whether failures came from weak retrieval, weak examples, or premature coding."
-    },
-    {
-      id: "domain-proof",
-      question: "Which tasks produced executable domain proof versus UI copy, placeholders, or inferred behavior?",
-      why: "Keeps product-quality claims tied to concrete evidence."
-    },
-    {
-      id: "root-cause",
-      question: "For each major failure cluster, is the likely root cause prompt/scaffold, docs/examples, SDK/API ergonomics, evaluator, runtime, or model behavior?",
-      why: "Turns trace observations into actionable ownership."
-    },
-    {
-      id: "evidence-quality",
-      question: "Which external-facing claims are directly supported by trace ids, span ids, verifier findings, reviewer notes, or generated code?",
-      why: "Prevents unsupported customer-report conclusions."
-    }
-  ];
-  if (hasMultipleShots) {
-    questions.push({
-      id: "reviewer-lift",
-      question: "Where did reviewer feedback improve score, stall, or regress across shots?",
-      why: "Shows whether the driver loop is learning or merely repeating work."
-    });
-  }
-  if (hasFailures) {
-    questions.push({
-      id: "optimization-targets",
-      question: "Which prompt, evaluator, scaffold, or workflow changes should feed the next GEPA/autoresearch optimization run?",
-      why: "Connects benchmark evidence to the optimization loop."
-    });
-  }
-  return questions;
-}
-function buildTraceInsightContext(input) {
-  return {
-    suite: input.suite,
-    scope: describeTraceInsightScope(input.suite),
-    keywords: inferDomainKeywords(input.suite),
-    questions: planTraceInsightQuestions(input),
-    panel: defaultTraceInsightPanel(),
-    findings: input.findings ?? [],
-    agent: input.agent ?? null,
-    totals: input.totals ?? null
-  };
-}
-function scoreTraceInsightReadiness(context) {
-  const failedTasks = context.suite.tasks.filter(
-    (task) => task.outcome && task.outcome !== "satisfied"
-  );
-  const findingTaskIds = new Set(context.findings.flatMap((finding) => finding.taskIds));
-  const failedTasksWithFindings = failedTasks.filter((task) => findingTaskIds.has(task.id));
-  const tasksWithGaps = context.suite.tasks.filter((task) => (task.gaps ?? []).length > 0);
-  const gates = [
-    {
-      id: "domain-context",
-      label: "Domain context inferred",
-      passed: context.keywords.length > 0,
-      severity: "high",
-      detail: context.keywords.length > 0 ? `${context.keywords.length} domain terms inferred: ${context.keywords.slice(0, 8).join(", ")}` : "No domain terms were inferred from suite, tasks, prompts, tags, or gaps."
-    },
-    {
-      id: "panel-coverage",
-      label: "Analyst panel planned",
-      passed: context.panel.length >= 4 && context.questions.length >= 5,
-      severity: "high",
-      detail: `${context.panel.length} panel roles and ${context.questions.length} investigation questions planned.`
-    },
-    {
-      id: "failure-coverage",
-      label: "Failures mapped to findings",
-      passed: failedTasks.length === 0 || failedTasksWithFindings.length / failedTasks.length >= 0.5,
-      severity: "critical",
-      detail: failedTasks.length === 0 ? "No failed tasks in suite." : `${failedTasksWithFindings.length}/${failedTasks.length} failed tasks appear in finding clusters.`
-    },
-    {
-      id: "gap-evidence",
-      label: "Task gaps captured",
-      passed: failedTasks.length === 0 || tasksWithGaps.length / failedTasks.length >= 0.5,
-      severity: "medium",
-      detail: `${tasksWithGaps.length} tasks include explicit evaluator or analyst gaps.`
-    }
-  ];
-  const penalty = gates.reduce((sum, gate) => {
-    if (gate.passed) return sum;
-    if (gate.severity === "critical") return sum + 35;
-    if (gate.severity === "high") return sum + 20;
-    if (gate.severity === "medium") return sum + 10;
-    return sum + 5;
-  }, 0);
-  const score = Math.max(0, Math.min(1, 1 - penalty / 100));
-  return {
-    score,
-    grade: score >= 0.9 ? "external-ready" : score >= 0.7 ? "internal-review" : "raw-analysis",
-    gates
-  };
-}
-function defaultTraceInsightPanel() {
-  return [
-    {
-      id: "trace-forensics",
-      name: "Trace Forensics",
-      responsibility: "Reconstruct what the worker did in order, including research, edits, reviewer interventions, verifier feedback, and stop reason."
-    },
-    {
-      id: "root-cause",
-      name: "Root Cause",
-      responsibility: "Map failures to prompt/scaffold, docs/examples, SDK/API/product ergonomics, evaluator, runtime, or model behavior."
-    },
-    {
-      id: "optimization",
-      name: "Optimization",
-      responsibility: "Identify prompt, reviewer, evaluator, scaffold, and GEPA/autoresearch changes that should be tested next."
-    },
-    {
-      id: "external-evidence",
-      name: "External Evidence",
-      responsibility: "Separate customer-safe claims from internal harness findings and reject conclusions without task, trace, span, code, reviewer, or verifier evidence."
-    }
-  ];
-}
-function buildTraceInsightPrompt(input) {
-  const context = buildTraceInsightContext(input);
-  const maxRepresentativeTraces = input.maxRepresentativeTraces ?? 6;
-  return `Analyze this benchmark run and produce evidence-backed trace intelligence.
-
-Audience:
-- internal AI/product leadership
-- possible customer-facing report for ${input.suite.name}
-
-Investigation plan:
-${context.questions.map((item, index) => `${index + 1}. ${item.question} (${item.why})`).join("\n")}
-
-Analyst panel:
-${context.panel.map((role) => `- ${role.name}: ${role.responsibility}`).join("\n")}
-
-If the task branches are independent, use subagents for the panel roles above and aggregate their findings. Do not run a panel role unless its answer will change the final report.
-
-Required output:
-1. Executive verdict: what this run proves and does not prove.
-2. The investigation questions you answered and the evidence used.
-3. Failure taxonomy: agent prompting, evaluator/harness, docs/examples, SDK/API/product integration, infra.
-4. Evidence-backed examples with trace ids/task ids and concrete verifier findings.
-5. Highest-ROI fixes for the benchmark harness, prompt/GEPA optimization, and customer-facing product/docs surface.
-6. What is safe for an external report versus what must stay internal.
-7. One rerun plan that would validate lift after optimization.
-
-Budget:
-- Inspect the dataset overview, the failure summary, and at most ${maxRepresentativeTraces} representative traces.
-- Prefer traces named in the failure summary over broad exploration.
-- Do not do exhaustive trace sweeps.
-- Return the final report as soon as the taxonomy and examples are supported.
-
-Run summary:
-${JSON.stringify(
-    {
-      suite: input.suite.name,
-      scope: context.scope,
-      inferredKeywords: context.keywords,
-      agent: context.agent,
-      totals: context.totals,
-      findings: context.findings.map((finding) => ({
-        kind: finding.kind,
-        severity: finding.severity,
-        taskCount: finding.taskIds.length,
-        proposedFixClass: finding.proposedFixClass
-      })),
-      failures: input.suite.tasks.filter((task) => task.outcome && task.outcome !== "satisfied").map((task) => ({
-        task: task.id,
-        difficulty: task.difficulty,
-        outcome: task.outcome,
-        score: task.score,
-        gaps: task.gaps ?? []
-      }))
-    },
-    null,
-    2
-  )}
-
-Use the trace tools. Do not invent facts. Cite task ids. Separate customer-facing claims from internal harness/model findings.`;
-}
 export {
   DEFAULT_REDACTION_RULES,
   DEFAULT_TRACE_ANALYST_BUDGETS,

package/dist/traces.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/trace-analyst/hook.ts","../src/trace-analyst/insights.ts"],"sourcesContent":["/**\n * Trace-analyst auto-execution hook.\n *\n * Wires `analyzeTraces` into a `TraceEmitter`'s `onRunComplete` so a\n * direct matrix run produces an analysis artifact without an out-of-band\n * step. Designed for the case where the consumer reports \"the analyst\n * never ran\" — the cause is almost always orchestration, not the analyst.\n *\n * Usage:\n *\n *   const emitter = new TraceEmitter(store, {\n *     onRunComplete: [traceAnalystOnRunComplete({ analyze: opts, save })],\n *   })\n *\n * Hooks are best-effort by default — they never crash the underlying run.\n * The caller decides whether to gate the run on the analysis result via\n * the `gateOn` callback.\n */\n\nimport type { RunCompleteHook, RunCompleteHookContext } from '../trace/emitter'\nimport { type AnalyzeTracesOptions, type AnalyzeTracesResult, analyzeTraces } from './analyst'\n\nexport interface TraceAnalystHookOptions {\n  /**\n   * Options forwarded to `analyzeTraces`. The hook supplies the question\n   * if you don't pass one — defaulting to a launch-grade prompt that asks\n   * for failure modes, surprising findings, and a recommendation.\n   */\n  analyze: Omit<AnalyzeTracesOptions, 'source'> & { source?: AnalyzeTracesOptions['source'] }\n  /**\n   * Override the question. The default is intentionally generic:\n   * \"Summarise what happened in this run, surface any failure modes,\n   *  surprising findings, or evidence the verdict is wrong.\"\n   */\n  question?: string\n  /**\n   * Persist the result. The hook calls this with the analysis output and\n   * the run context. Common implementations write to a TraceAnalysisStore\n   * or append to a per-run JSONL.\n   */\n  save?: (result: AnalyzeTracesResult, ctx: RunCompleteHookContext) => Promise<void>\n  /**\n   * Predicate gating execution per run. Default: every completed run.\n   * Use to skip aborted runs, debug runs, or runs without LLM activity.\n   */\n  shouldRun?: (ctx: RunCompleteHookContext) => boolean\n  /**\n   * Optional gate: if set and returns false, the hook records the failure\n   * as a log event on the run instead of staying quiet. The caller can\n   * then trigger downstream alerts off `analyst_gate_failed` log events.\n   */\n  gateOn?: (result: AnalyzeTracesResult, ctx: RunCompleteHookContext) => boolean\n}\n\nconst DEFAULT_QUESTION =\n  \"Summarise what happened in this run. Surface any failure modes, surprising findings, or evidence that the run's verdict is wrong.\"\n\nexport function traceAnalystOnRunComplete(opts: TraceAnalystHookOptions): RunCompleteHook {\n  return async (ctx: RunCompleteHookContext) => {\n    if (opts.shouldRun && !opts.shouldRun(ctx)) return\n    const source = opts.analyze.source\n    if (source === undefined) {\n      // The analyst needs a source. If the caller didn't supply one we don't\n      // run — but we do leave a breadcrumb so the absence is visible.\n      await ctx.store.appendEvent({\n        eventId: `analyst-skip-${ctx.runId}`,\n        runId: ctx.runId,\n        kind: 'log',\n        timestamp: Date.now(),\n        payload: { source: 'trace_analyst_hook', reason: 'no source configured' },\n      })\n      return\n    }\n    const result = await analyzeTraces({ question: opts.question ?? DEFAULT_QUESTION }, {\n      ...opts.analyze,\n      source,\n    } as AnalyzeTracesOptions)\n    if (opts.save) await opts.save(result, ctx)\n    if (opts.gateOn && !opts.gateOn(result, ctx)) {\n      await ctx.store.appendEvent({\n        eventId: `analyst-gate-${ctx.runId}`,\n        runId: ctx.runId,\n        kind: 'log',\n        timestamp: Date.now(),\n        payload: {\n          source: 'trace_analyst_hook',\n          reason: 'analyst_gate_failed',\n          findings: result.findings,\n        },\n      })\n    }\n  }\n}\n","export interface TraceInsightTask {\n  id: string\n  name: string\n  prompt?: string\n  difficulty?: string\n  tags?: string[]\n  outcome?: string\n  score?: number\n  gaps?: string[]\n}\n\nexport interface TraceInsightSuite {\n  name: string\n  collectionId?: string\n  tasks: TraceInsightTask[]\n}\n\nexport interface TraceInsightFinding {\n  kind: string\n  severity?: string\n  taskIds: string[]\n  evidence?: string\n  proposedFixClass?: string\n}\n\nexport interface TraceInsightQuestion {\n  id: string\n  question: string\n  why: string\n}\n\nexport interface TraceInsightPanelRole {\n  id: string\n  name: string\n  responsibility: string\n}\n\nexport interface TraceInsightPromptInput {\n  suite: TraceInsightSuite\n  findings?: TraceInsightFinding[]\n  agent?: Record<string, unknown>\n  totals?: Record<string, unknown>\n  maxRepresentativeTraces?: number\n}\n\nexport interface TraceInsightContext {\n  suite: TraceInsightSuite\n  scope: string\n  keywords: string[]\n  questions: TraceInsightQuestion[]\n  panel: TraceInsightPanelRole[]\n  findings: TraceInsightFinding[]\n  agent: Record<string, unknown> | null\n  totals: Record<string, unknown> | null\n}\n\nexport interface TraceInsightQualityGate {\n  id: string\n  label: string\n  passed: boolean\n  severity: 'critical' | 'high' | 'medium' | 'low'\n  detail: string\n}\n\nexport interface TraceInsightReadiness {\n  score: number\n  grade: 'external-ready' | 'internal-review' | 'raw-analysis'\n  gates: TraceInsightQualityGate[]\n}\n\nconst DOMAIN_STOP_WORDS = new Set([\n  'and',\n  'advanced',\n  'app',\n  'build',\n  'create',\n  'easy',\n  'expert',\n  'extreme',\n  'for',\n  'from',\n  'hard',\n  'implementation',\n  'integrate',\n  'medium',\n  'project',\n  'task',\n  'the',\n  'this',\n  'with',\n  'workflow',\n])\n\nexport function tokenizeDomainWords(value: string): string[] {\n  return [...value.matchAll(/[A-Za-z][A-Za-z0-9.+#-]{2,}/g)]\n    .map((match) => match[0].toLowerCase())\n    .filter((word) => !DOMAIN_STOP_WORDS.has(word))\n}\n\nexport function inferDomainKeywords(suite: TraceInsightSuite): string[] {\n  const suiteWords = new Set(tokenizeDomainWords(`${suite.name} ${suite.collectionId ?? ''}`))\n  const source = [\n    suite.name,\n    suite.collectionId ?? '',\n    ...suite.tasks.flatMap((task) => [\n      task.id,\n      task.name,\n      task.prompt ?? '',\n      task.difficulty ?? '',\n      ...(task.tags ?? []),\n      ...(task.gaps ?? []),\n    ]),\n  ].join(' ')\n  const counts = new Map<string, number>()\n  for (const word of tokenizeDomainWords(source)) counts.set(word, (counts.get(word) ?? 0) + 1)\n  return [...counts.entries()]\n    .filter(([word, count]) => count >= 2 || suiteWords.has(word))\n    .sort((a, b) => b[1] - a[1] || a[0].localeCompare(b[0]))\n    .map(([word]) => word)\n    .slice(0, 18)\n}\n\nexport function domainEvidencePattern(keywords: string[]): RegExp {\n  const escaped = keywords\n    .filter((keyword) => keyword.length >= 3)\n    .map((keyword) => keyword.replace(/[.*+?^${}()|[\\]\\\\]/g, '\\\\$&'))\n  return escaped.length > 0\n    ? new RegExp(`(?<![A-Za-z0-9])(?:${escaped.join('|')})(?![A-Za-z0-9])`, 'i')\n    : /(?<![A-Za-z0-9])(?:sdk|api|css|dns|xml|provider|client|service|integration|webhook|transaction|auth|oauth|graphql|rest)(?![A-Za-z0-9])/i\n}\n\nexport function describeTraceInsightScope(suite: TraceInsightSuite): string {\n  const taskLabel =\n    suite.tasks.length === 1\n      ? '1 implementation task'\n      : `${suite.tasks.length} implementation tasks`\n  const tags = new Map<string, number>()\n  for (const task of suite.tasks) {\n    for (const tag of task.tags ?? []) tags.set(tag, (tags.get(tag) ?? 0) + 1)\n  }\n  const topTags = [...tags.entries()]\n    .sort((a, b) => b[1] - a[1] || a[0].localeCompare(b[0]))\n    .slice(0, 8)\n    .map(([tag]) => tag)\n  if (topTags.length > 0) return `${taskLabel} across ${topTags.join(', ')}.`\n  const difficulties = [\n    ...new Set(\n      suite.tasks.map((task) => task.difficulty).filter((value): value is string => Boolean(value)),\n    ),\n  ].join(', ')\n  return `${taskLabel} across ${difficulties || 'the selected benchmark scope'}.`\n}\n\nexport function planTraceInsightQuestions(input: TraceInsightPromptInput): TraceInsightQuestion[] {\n  const hasFailures = input.suite.tasks.some((task) => task.outcome && task.outcome !== 'satisfied')\n  const hasMultipleShots = input.suite.tasks.some((task) =>\n    (task.gaps ?? []).some((gap) => /shot|review|retry|continue/i.test(gap)),\n  )\n  const questions: TraceInsightQuestion[] = [\n    {\n      id: 'execution-path',\n      question: 'What did the worker actually do before the first meaningful implementation edit?',\n      why: 'Separates grounded execution from polished but shallow output.',\n    },\n    {\n      id: 'research-grounding',\n      question:\n        'Did the worker inspect docs, source, examples, or package references before committing to an implementation path?',\n      why: 'Identifies whether failures came from weak retrieval, weak examples, or premature coding.',\n    },\n    {\n      id: 'domain-proof',\n      question:\n        'Which tasks produced executable domain proof versus UI copy, placeholders, or inferred behavior?',\n      why: 'Keeps product-quality claims tied to concrete evidence.',\n    },\n    {\n      id: 'root-cause',\n      question:\n        'For each major failure cluster, is the likely root cause prompt/scaffold, docs/examples, SDK/API ergonomics, evaluator, runtime, or model behavior?',\n      why: 'Turns trace observations into actionable ownership.',\n    },\n    {\n      id: 'evidence-quality',\n      question:\n        'Which external-facing claims are directly supported by trace ids, span ids, verifier findings, reviewer notes, or generated code?',\n      why: 'Prevents unsupported customer-report conclusions.',\n    },\n  ]\n  if (hasMultipleShots) {\n    questions.push({\n      id: 'reviewer-lift',\n      question: 'Where did reviewer feedback improve score, stall, or regress across shots?',\n      why: 'Shows whether the driver loop is learning or merely repeating work.',\n    })\n  }\n  if (hasFailures) {\n    questions.push({\n      id: 'optimization-targets',\n      question:\n        'Which prompt, evaluator, scaffold, or workflow changes should feed the next GEPA/autoresearch optimization run?',\n      why: 'Connects benchmark evidence to the optimization loop.',\n    })\n  }\n  return questions\n}\n\nexport function buildTraceInsightContext(input: TraceInsightPromptInput): TraceInsightContext {\n  return {\n    suite: input.suite,\n    scope: describeTraceInsightScope(input.suite),\n    keywords: inferDomainKeywords(input.suite),\n    questions: planTraceInsightQuestions(input),\n    panel: defaultTraceInsightPanel(),\n    findings: input.findings ?? [],\n    agent: input.agent ?? null,\n    totals: input.totals ?? null,\n  }\n}\n\nexport function scoreTraceInsightReadiness(context: TraceInsightContext): TraceInsightReadiness {\n  const failedTasks = context.suite.tasks.filter(\n    (task) => task.outcome && task.outcome !== 'satisfied',\n  )\n  const findingTaskIds = new Set(context.findings.flatMap((finding) => finding.taskIds))\n  const failedTasksWithFindings = failedTasks.filter((task) => findingTaskIds.has(task.id))\n  const tasksWithGaps = context.suite.tasks.filter((task) => (task.gaps ?? []).length > 0)\n  const gates: TraceInsightQualityGate[] = [\n    {\n      id: 'domain-context',\n      label: 'Domain context inferred',\n      passed: context.keywords.length > 0,\n      severity: 'high',\n      detail:\n        context.keywords.length > 0\n          ? `${context.keywords.length} domain terms inferred: ${context.keywords.slice(0, 8).join(', ')}`\n          : 'No domain terms were inferred from suite, tasks, prompts, tags, or gaps.',\n    },\n    {\n      id: 'panel-coverage',\n      label: 'Analyst panel planned',\n      passed: context.panel.length >= 4 && context.questions.length >= 5,\n      severity: 'high',\n      detail: `${context.panel.length} panel roles and ${context.questions.length} investigation questions planned.`,\n    },\n    {\n      id: 'failure-coverage',\n      label: 'Failures mapped to findings',\n      passed:\n        failedTasks.length === 0 || failedTasksWithFindings.length / failedTasks.length >= 0.5,\n      severity: 'critical',\n      detail:\n        failedTasks.length === 0\n          ? 'No failed tasks in suite.'\n          : `${failedTasksWithFindings.length}/${failedTasks.length} failed tasks appear in finding clusters.`,\n    },\n    {\n      id: 'gap-evidence',\n      label: 'Task gaps captured',\n      passed: failedTasks.length === 0 || tasksWithGaps.length / failedTasks.length >= 0.5,\n      severity: 'medium',\n      detail: `${tasksWithGaps.length} tasks include explicit evaluator or analyst gaps.`,\n    },\n  ]\n  const penalty = gates.reduce((sum, gate) => {\n    if (gate.passed) return sum\n    if (gate.severity === 'critical') return sum + 35\n    if (gate.severity === 'high') return sum + 20\n    if (gate.severity === 'medium') return sum + 10\n    return sum + 5\n  }, 0)\n  const score = Math.max(0, Math.min(1, 1 - penalty / 100))\n  return {\n    score,\n    grade: score >= 0.9 ? 'external-ready' : score >= 0.7 ? 'internal-review' : 'raw-analysis',\n    gates,\n  }\n}\n\nexport function defaultTraceInsightPanel(): TraceInsightPanelRole[] {\n  return [\n    {\n      id: 'trace-forensics',\n      name: 'Trace Forensics',\n      responsibility:\n        'Reconstruct what the worker did in order, including research, edits, reviewer interventions, verifier feedback, and stop reason.',\n    },\n    {\n      id: 'root-cause',\n      name: 'Root Cause',\n      responsibility:\n        'Map failures to prompt/scaffold, docs/examples, SDK/API/product ergonomics, evaluator, runtime, or model behavior.',\n    },\n    {\n      id: 'optimization',\n      name: 'Optimization',\n      responsibility:\n        'Identify prompt, reviewer, evaluator, scaffold, and GEPA/autoresearch changes that should be tested next.',\n    },\n    {\n      id: 'external-evidence',\n      name: 'External Evidence',\n      responsibility:\n        'Separate customer-safe claims from internal harness findings and reject conclusions without task, trace, span, code, reviewer, or verifier evidence.',\n    },\n  ]\n}\n\nexport function buildTraceInsightPrompt(input: TraceInsightPromptInput): string {\n  const context = buildTraceInsightContext(input)\n  const maxRepresentativeTraces = input.maxRepresentativeTraces ?? 6\n  return `Analyze this benchmark run and produce evidence-backed trace intelligence.\n\nAudience:\n- internal AI/product leadership\n- possible customer-facing report for ${input.suite.name}\n\nInvestigation plan:\n${context.questions.map((item, index) => `${index + 1}. ${item.question} (${item.why})`).join('\\n')}\n\nAnalyst panel:\n${context.panel.map((role) => `- ${role.name}: ${role.responsibility}`).join('\\n')}\n\nIf the task branches are independent, use subagents for the panel roles above and aggregate their findings. Do not run a panel role unless its answer will change the final report.\n\nRequired output:\n1. Executive verdict: what this run proves and does not prove.\n2. The investigation questions you answered and the evidence used.\n3. Failure taxonomy: agent prompting, evaluator/harness, docs/examples, SDK/API/product integration, infra.\n4. Evidence-backed examples with trace ids/task ids and concrete verifier findings.\n5. Highest-ROI fixes for the benchmark harness, prompt/GEPA optimization, and customer-facing product/docs surface.\n6. What is safe for an external report versus what must stay internal.\n7. One rerun plan that would validate lift after optimization.\n\nBudget:\n- Inspect the dataset overview, the failure summary, and at most ${maxRepresentativeTraces} representative traces.\n- Prefer traces named in the failure summary over broad exploration.\n- Do not do exhaustive trace sweeps.\n- Return the final report as soon as the taxonomy and examples are supported.\n\nRun summary:\n${JSON.stringify(\n  {\n    suite: input.suite.name,\n    scope: context.scope,\n    inferredKeywords: context.keywords,\n    agent: context.agent,\n    totals: context.totals,\n    findings: context.findings.map((finding) => ({\n      kind: finding.kind,\n      severity: finding.severity,\n      taskCount: finding.taskIds.length,\n      proposedFixClass: finding.proposedFixClass,\n    })),\n    failures: input.suite.tasks\n      .filter((task) => task.outcome && task.outcome !== 'satisfied')\n      .map((task) => ({\n        task: task.id,\n        difficulty: task.difficulty,\n        outcome: task.outcome,\n        score: task.score,\n        gaps: task.gaps ?? [],\n      })),\n  },\n  null,\n  2,\n)}\n\nUse the trace tools. Do not invent facts. Cite task ids. Separate customer-facing claims from internal harness/model findings.`\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAsDA,IAAM,mBACJ;AAEK,SAAS,0BAA0B,MAAgD;AACxF,SAAO,OAAO,QAAgC;AAC5C,QAAI,KAAK,aAAa,CAAC,KAAK,UAAU,GAAG,EAAG;AAC5C,UAAM,SAAS,KAAK,QAAQ;AAC5B,QAAI,WAAW,QAAW;AAGxB,YAAM,IAAI,MAAM,YAAY;AAAA,QAC1B,SAAS,gBAAgB,IAAI,KAAK;AAAA,QAClC,OAAO,IAAI;AAAA,QACX,MAAM;AAAA,QACN,WAAW,KAAK,IAAI;AAAA,QACpB,SAAS,EAAE,QAAQ,sBAAsB,QAAQ,uBAAuB;AAAA,MAC1E,CAAC;AACD;AAAA,IACF;AACA,UAAM,SAAS,MAAM,cAAc,EAAE,UAAU,KAAK,YAAY,iBAAiB,GAAG;AAAA,MAClF,GAAG,KAAK;AAAA,MACR;AAAA,IACF,CAAyB;AACzB,QAAI,KAAK,KAAM,OAAM,KAAK,KAAK,QAAQ,GAAG;AAC1C,QAAI,KAAK,UAAU,CAAC,KAAK,OAAO,QAAQ,GAAG,GAAG;AAC5C,YAAM,IAAI,MAAM,YAAY;AAAA,QAC1B,SAAS,gBAAgB,IAAI,KAAK;AAAA,QAClC,OAAO,IAAI;AAAA,QACX,MAAM;AAAA,QACN,WAAW,KAAK,IAAI;AAAA,QACpB,SAAS;AAAA,UACP,QAAQ;AAAA,UACR,QAAQ;AAAA,UACR,UAAU,OAAO;AAAA,QACnB;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AACF;;;ACtBA,IAAM,oBAAoB,oBAAI,IAAI;AAAA,EAChC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF,CAAC;AAEM,SAAS,oBAAoB,OAAyB;AAC3D,SAAO,CAAC,GAAG,MAAM,SAAS,8BAA8B,CAAC,EACtD,IAAI,CAAC,UAAU,MAAM,CAAC,EAAE,YAAY,CAAC,EACrC,OAAO,CAAC,SAAS,CAAC,kBAAkB,IAAI,IAAI,CAAC;AAClD;AAEO,SAAS,oBAAoB,OAAoC;AACtE,QAAM,aAAa,IAAI,IAAI,oBAAoB,GAAG,MAAM,IAAI,IAAI,MAAM,gBAAgB,EAAE,EAAE,CAAC;AAC3F,QAAM,SAAS;AAAA,IACb,MAAM;AAAA,IACN,MAAM,gBAAgB;AAAA,IACtB,GAAG,MAAM,MAAM,QAAQ,CAAC,SAAS;AAAA,MAC/B,KAAK;AAAA,MACL,KAAK;AAAA,MACL,KAAK,UAAU;AAAA,MACf,KAAK,cAAc;AAAA,MACnB,GAAI,KAAK,QAAQ,CAAC;AAAA,MAClB,GAAI,KAAK,QAAQ,CAAC;AAAA,IACpB,CAAC;AAAA,EACH,EAAE,KAAK,GAAG;AACV,QAAM,SAAS,oBAAI,IAAoB;AACvC,aAAW,QAAQ,oBAAoB,MAAM,EAAG,QAAO,IAAI,OAAO,OAAO,IAAI,IAAI,KAAK,KAAK,CAAC;AAC5F,SAAO,CAAC,GAAG,OAAO,QAAQ,CAAC,EACxB,OAAO,CAAC,CAAC,MAAM,KAAK,MAAM,SAAS,KAAK,WAAW,IAAI,IAAI,CAAC,EAC5D,KAAK,CAAC,GAAG,MAAM,EAAE,CAAC,IAAI,EAAE,CAAC,KAAK,EAAE,CAAC,EAAE,cAAc,EAAE,CAAC,CAAC,CAAC,EACtD,IAAI,CAAC,CAAC,IAAI,MAAM,IAAI,EACpB,MAAM,GAAG,EAAE;AAChB;AAEO,SAAS,sBAAsB,UAA4B;AAChE,QAAM,UAAU,SACb,OAAO,CAAC,YAAY,QAAQ,UAAU,CAAC,EACvC,IAAI,CAAC,YAAY,QAAQ,QAAQ,uBAAuB,MAAM,CAAC;AAClE,SAAO,QAAQ,SAAS,IACpB,IAAI,OAAO,sBAAsB,QAAQ,KAAK,GAAG,CAAC,oBAAoB,GAAG,IACzE;AACN;AAEO,SAAS,0BAA0B,OAAkC;AAC1E,QAAM,YACJ,MAAM,MAAM,WAAW,IACnB,0BACA,GAAG,MAAM,MAAM,MAAM;AAC3B,QAAM,OAAO,oBAAI,IAAoB;AACrC,aAAW,QAAQ,MAAM,OAAO;AAC9B,eAAW,OAAO,KAAK,QAAQ,CAAC,EAAG,MAAK,IAAI,MAAM,KAAK,IAAI,GAAG,KAAK,KAAK,CAAC;AAAA,EAC3E;AACA,QAAM,UAAU,CAAC,GAAG,KAAK,QAAQ,CAAC,EAC/B,KAAK,CAAC,GAAG,MAAM,EAAE,CAAC,IAAI,EAAE,CAAC,KAAK,EAAE,CAAC,EAAE,cAAc,EAAE,CAAC,CAAC,CAAC,EACtD,MAAM,GAAG,CAAC,EACV,IAAI,CAAC,CAAC,GAAG,MAAM,GAAG;AACrB,MAAI,QAAQ,SAAS,EAAG,QAAO,GAAG,SAAS,WAAW,QAAQ,KAAK,IAAI,CAAC;AACxE,QAAM,eAAe;AAAA,IACnB,GAAG,IAAI;AAAA,MACL,MAAM,MAAM,IAAI,CAAC,SAAS,KAAK,UAAU,EAAE,OAAO,CAAC,UAA2B,QAAQ,KAAK,CAAC;AAAA,IAC9F;AAAA,EACF,EAAE,KAAK,IAAI;AACX,SAAO,GAAG,SAAS,WAAW,gBAAgB,8BAA8B;AAC9E;AAEO,SAAS,0BAA0B,OAAwD;AAChG,QAAM,cAAc,MAAM,MAAM,MAAM,KAAK,CAAC,SAAS,KAAK,WAAW,KAAK,YAAY,WAAW;AACjG,QAAM,mBAAmB,MAAM,MAAM,MAAM;AAAA,IAAK,CAAC,UAC9C,KAAK,QAAQ,CAAC,GAAG,KAAK,CAAC,QAAQ,8BAA8B,KAAK,GAAG,CAAC;AAAA,EACzE;AACA,QAAM,YAAoC;AAAA,IACxC;AAAA,MACE,IAAI;AAAA,MACJ,UAAU;AAAA,MACV,KAAK;AAAA,IACP;AAAA,IACA;AAAA,MACE,IAAI;AAAA,MACJ,UACE;AAAA,MACF,KAAK;AAAA,IACP;AAAA,IACA;AAAA,MACE,IAAI;AAAA,MACJ,UACE;AAAA,MACF,KAAK;AAAA,IACP;AAAA,IACA;AAAA,MACE,IAAI;AAAA,MACJ,UACE;AAAA,MACF,KAAK;AAAA,IACP;AAAA,IACA;AAAA,MACE,IAAI;AAAA,MACJ,UACE;AAAA,MACF,KAAK;AAAA,IACP;AAAA,EACF;AACA,MAAI,kBAAkB;AACpB,cAAU,KAAK;AAAA,MACb,IAAI;AAAA,MACJ,UAAU;AAAA,MACV,KAAK;AAAA,IACP,CAAC;AAAA,EACH;AACA,MAAI,aAAa;AACf,cAAU,KAAK;AAAA,MACb,IAAI;AAAA,MACJ,UACE;AAAA,MACF,KAAK;AAAA,IACP,CAAC;AAAA,EACH;AACA,SAAO;AACT;AAEO,SAAS,yBAAyB,OAAqD;AAC5F,SAAO;AAAA,IACL,OAAO,MAAM;AAAA,IACb,OAAO,0BAA0B,MAAM,KAAK;AAAA,IAC5C,UAAU,oBAAoB,MAAM,KAAK;AAAA,IACzC,WAAW,0BAA0B,KAAK;AAAA,IAC1C,OAAO,yBAAyB;AAAA,IAChC,UAAU,MAAM,YAAY,CAAC;AAAA,IAC7B,OAAO,MAAM,SAAS;AAAA,IACtB,QAAQ,MAAM,UAAU;AAAA,EAC1B;AACF;AAEO,SAAS,2BAA2B,SAAqD;AAC9F,QAAM,cAAc,QAAQ,MAAM,MAAM;AAAA,IACtC,CAAC,SAAS,KAAK,WAAW,KAAK,YAAY;AAAA,EAC7C;AACA,QAAM,iBAAiB,IAAI,IAAI,QAAQ,SAAS,QAAQ,CAAC,YAAY,QAAQ,OAAO,CAAC;AACrF,QAAM,0BAA0B,YAAY,OAAO,CAAC,SAAS,eAAe,IAAI,KAAK,EAAE,CAAC;AACxF,QAAM,gBAAgB,QAAQ,MAAM,MAAM,OAAO,CAAC,UAAU,KAAK,QAAQ,CAAC,GAAG,SAAS,CAAC;AACvF,QAAM,QAAmC;AAAA,IACvC;AAAA,MACE,IAAI;AAAA,MACJ,OAAO;AAAA,MACP,QAAQ,QAAQ,SAAS,SAAS;AAAA,MAClC,UAAU;AAAA,MACV,QACE,QAAQ,SAAS,SAAS,IACtB,GAAG,QAAQ,SAAS,MAAM,2BAA2B,QAAQ,SAAS,MAAM,GAAG,CAAC,EAAE,KAAK,IAAI,CAAC,KAC5F;AAAA,IACR;AAAA,IACA;AAAA,MACE,IAAI;AAAA,MACJ,OAAO;AAAA,MACP,QAAQ,QAAQ,MAAM,UAAU,KAAK,QAAQ,UAAU,UAAU;AAAA,MACjE,UAAU;AAAA,MACV,QAAQ,GAAG,QAAQ,MAAM,MAAM,oBAAoB,QAAQ,UAAU,MAAM;AAAA,IAC7E;AAAA,IACA;AAAA,MACE,IAAI;AAAA,MACJ,OAAO;AAAA,MACP,QACE,YAAY,WAAW,KAAK,wBAAwB,SAAS,YAAY,UAAU;AAAA,MACrF,UAAU;AAAA,MACV,QACE,YAAY,WAAW,IACnB,8BACA,GAAG,wBAAwB,MAAM,IAAI,YAAY,MAAM;AAAA,IAC/D;AAAA,IACA;AAAA,MACE,IAAI;AAAA,MACJ,OAAO;AAAA,MACP,QAAQ,YAAY,WAAW,KAAK,cAAc,SAAS,YAAY,UAAU;AAAA,MACjF,UAAU;AAAA,MACV,QAAQ,GAAG,cAAc,MAAM;AAAA,IACjC;AAAA,EACF;AACA,QAAM,UAAU,MAAM,OAAO,CAAC,KAAK,SAAS;AAC1C,QAAI,KAAK,OAAQ,QAAO;AACxB,QAAI,KAAK,aAAa,WAAY,QAAO,MAAM;AAC/C,QAAI,KAAK,aAAa,OAAQ,QAAO,MAAM;AAC3C,QAAI,KAAK,aAAa,SAAU,QAAO,MAAM;AAC7C,WAAO,MAAM;AAAA,EACf,GAAG,CAAC;AACJ,QAAM,QAAQ,KAAK,IAAI,GAAG,KAAK,IAAI,GAAG,IAAI,UAAU,GAAG,CAAC;AACxD,SAAO;AAAA,IACL;AAAA,IACA,OAAO,SAAS,MAAM,mBAAmB,SAAS,MAAM,oBAAoB;AAAA,IAC5E;AAAA,EACF;AACF;AAEO,SAAS,2BAAoD;AAClE,SAAO;AAAA,IACL;AAAA,MACE,IAAI;AAAA,MACJ,MAAM;AAAA,MACN,gBACE;AAAA,IACJ;AAAA,IACA;AAAA,MACE,IAAI;AAAA,MACJ,MAAM;AAAA,MACN,gBACE;AAAA,IACJ;AAAA,IACA;AAAA,MACE,IAAI;AAAA,MACJ,MAAM;AAAA,MACN,gBACE;AAAA,IACJ;AAAA,IACA;AAAA,MACE,IAAI;AAAA,MACJ,MAAM;AAAA,MACN,gBACE;AAAA,IACJ;AAAA,EACF;AACF;AAEO,SAAS,wBAAwB,OAAwC;AAC9E,QAAM,UAAU,yBAAyB,KAAK;AAC9C,QAAM,0BAA0B,MAAM,2BAA2B;AACjE,SAAO;AAAA;AAAA;AAAA;AAAA,wCAI+B,MAAM,MAAM,IAAI;AAAA;AAAA;AAAA,EAGtD,QAAQ,UAAU,IAAI,CAAC,MAAM,UAAU,GAAG,QAAQ,CAAC,KAAK,KAAK,QAAQ,KAAK,KAAK,GAAG,GAAG,EAAE,KAAK,IAAI,CAAC;AAAA;AAAA;AAAA,EAGjG,QAAQ,MAAM,IAAI,CAAC,SAAS,KAAK,KAAK,IAAI,KAAK,KAAK,cAAc,EAAE,EAAE,KAAK,IAAI,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,mEAcf,uBAAuB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAMxF,KAAK;AAAA,IACL;AAAA,MACE,OAAO,MAAM,MAAM;AAAA,MACnB,OAAO,QAAQ;AAAA,MACf,kBAAkB,QAAQ;AAAA,MAC1B,OAAO,QAAQ;AAAA,MACf,QAAQ,QAAQ;AAAA,MAChB,UAAU,QAAQ,SAAS,IAAI,CAAC,aAAa;AAAA,QAC3C,MAAM,QAAQ;AAAA,QACd,UAAU,QAAQ;AAAA,QAClB,WAAW,QAAQ,QAAQ;AAAA,QAC3B,kBAAkB,QAAQ;AAAA,MAC5B,EAAE;AAAA,MACF,UAAU,MAAM,MAAM,MACnB,OAAO,CAAC,SAAS,KAAK,WAAW,KAAK,YAAY,WAAW,EAC7D,IAAI,CAAC,UAAU;AAAA,QACd,MAAM,KAAK;AAAA,QACX,YAAY,KAAK;AAAA,QACjB,SAAS,KAAK;AAAA,QACd,OAAO,KAAK;AAAA,QACZ,MAAM,KAAK,QAAQ,CAAC;AAAA,MACtB,EAAE;AAAA,IACN;AAAA,IACA;AAAA,IACA;AAAA,EACF,CAAC;AAAA;AAAA;AAGD;","names":[]}
+{"version":3,"sources":[],"sourcesContent":[],"mappings":"","names":[]}

package/dist/{trajectory-BFmveYZt.d.ts → trajectory-CnoBo-JY.d.ts}

@@ -1,4 +1,4 @@
-import { S as Span, b as TraceEvent, T as TraceStore } from './store-BP5be6s7.js';
+import { S as Span, b as TraceEvent, T as TraceStore } from './store-Db2Bv8Cf.js';
 
 /**
  * Trajectory — ordered, structured view over a run's spans.

package/dist/wire/index.d.ts

@@ -1,14 +1,14 @@
-import { n as FeedbackTrajectoryStore } from '../feedback-trajectory-j0nJFgC6.js';
-import { T as TraceStore } from '../store-BP5be6s7.js';
+import { n as FeedbackTrajectoryStore } from '../feedback-trajectory-iATEAHmc.js';
+import { T as TraceStore } from '../store-Db2Bv8Cf.js';
 import { z } from 'zod';
 import { OpenAPIObject } from 'openapi3-ts/oas31';
 import * as hono_types from 'hono/types';
 import { ServerType } from '@hono/node-server';
 import { Hono } from 'hono';
-import '../control-runtime-BRdQ0wrx.js';
-import '../emitter-BqjeOvJh.js';
-import '../dataset-CiK_3LDr.js';
-import '../errors-BZ9sTdz7.js';
+import '../control-runtime-BZ_lVLYW.js';
+import '../emitter-DP_cSSiw.js';
+import '../dataset-ueRVTUoY.js';
+import '../errors-mje_cKOs.js';
 
 declare const RubricDimensionSchema: z.ZodObject<{
     id: z.ZodString;

package/dist/wire/index.js

@@ -33,10 +33,10 @@ import {
   runRpcBatch,
   runRpcOnce,
   startServer
-} from "../chunk-XFZCM5Z3.js";
-import "../chunk-4S4BM3QQ.js";
+} from "../chunk-SMSGXM74.js";
+import "../chunk-M6RZ5LJN.js";
 import "../chunk-PC4UYEBM.js";
-import "../chunk-NG236HPC.js";
+import "../chunk-QYJT52YW.js";
 import "../chunk-PZ5AY32C.js";
 export {
   BUILTIN_RUBRICS,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tangle-network/agent-eval",
-  "version": "0.30.0",
+  "version": "0.31.1",
   "description": "Substrate for self-improving agents: traces, verifiable rewards, preferences, GEPA / reflective mutation, auto-research, replay, sequential anytime-valid stats, and release gates.",
   "homepage": "https://github.com/tangle-network/agent-eval#readme",
   "repository": {
@@ -109,17 +109,6 @@
   "publishConfig": {
     "access": "public"
   },
-  "scripts": {
-    "build": "tsup && pnpm openapi",
-    "dev": "tsup --watch",
-    "prepare": "pnpm build",
-    "test": "vitest run",
-    "test:watch": "vitest",
-    "typecheck": "tsc --noEmit",
-    "lint": "biome check src",
-    "format": "biome format --write src",
-    "openapi": "node dist/cli.js openapi --out dist/openapi.json"
-  },
   "dependencies": {
     "@asteasolutions/zod-to-openapi": "^8.5.0",
     "@ax-llm/ax": "^19.0.25",
@@ -136,16 +125,18 @@
     "typescript": "^5.7.0",
     "vitest": "^3.0.0"
   },
-  "pnpm": {
-    "minimumReleaseAge": 4320,
-    "minimumReleaseAgeExclude": [],
-    "overrides": {
-      "postcss@<8.5.10": "^8.5.10"
-    }
-  },
   "engines": {
     "node": ">=20"
   },
   "license": "MIT",
-  "packageManager": "pnpm@10.22.0"
-}
+  "scripts": {
+    "build": "tsup && pnpm openapi",
+    "dev": "tsup --watch",
+    "test": "vitest run",
+    "test:watch": "vitest",
+    "typecheck": "tsc --noEmit",
+    "lint": "biome check src",
+    "format": "biome format --write src",
+    "openapi": "node dist/cli.js openapi --out dist/openapi.json"
+  }
+}

package/dist/chunk-NG236HPC.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../src/errors.ts"],"sourcesContent":["/**\n * Error taxonomy for `@tangle-network/agent-eval`.\n *\n * Every error this package throws as part of its *public contract* extends\n * `AgentEvalError`. Consumers can pattern-match by `instanceof <Subclass>` or\n * by the stable string `code` carried on the base class.\n *\n * The codes are stable across minor versions; new codes can be added, but\n * existing codes never change meaning. New subclasses are non-breaking.\n *\n * Internal invariant guards (`throw new Error('this should never happen')`)\n * remain plain `Error`s on purpose — they're programmer-mistake assertions,\n * not consumer-catchable contract failures.\n */\n\nexport type AgentEvalErrorCode =\n  | 'validation'\n  | 'not_found'\n  | 'config'\n  | 'capture_integrity'\n  | 'judge'\n  | 'verification'\n  | 'replay'\n\nexport class AgentEvalError extends Error {\n  /** Stable string code. Survives minification; safe to switch on. */\n  readonly code: AgentEvalErrorCode\n\n  constructor(code: AgentEvalErrorCode, message: string, options?: { cause?: unknown }) {\n    super(message, options)\n    this.name = this.constructor.name\n    this.code = code\n  }\n}\n\n/** Caller passed invalid arguments (out of range, mutually-exclusive options, bad shape). */\nexport class ValidationError extends AgentEvalError {\n  constructor(message: string, options?: { cause?: unknown }) {\n    super('validation', message, options)\n  }\n}\n\n/** A named resource (run, span, rubric, scenario, dataset row, route) does not exist. */\nexport class NotFoundError extends AgentEvalError {\n  constructor(message: string, options?: { cause?: unknown }) {\n    super('not_found', message, options)\n  }\n}\n\n/** Configuration missing or malformed (`HOME` unset, required image not supplied, env var absent). */\nexport class ConfigError extends AgentEvalError {\n  constructor(message: string, options?: { cause?: unknown }) {\n    super('config', message, options)\n  }\n}\n\n/**\n * A run is missing the artifacts a launch-grade check requires:\n * raw HTTP capture absent, no LLM spans, route assertion failed, run-end\n * assertion tripped. Block ship on this; do not catch and move on.\n */\nexport class CaptureIntegrityError extends AgentEvalError {\n  constructor(message: string, options?: { cause?: unknown }) {\n    super('capture_integrity', message, options)\n  }\n}\n\n/** A judge call failed in a way that's not retryable: schema parse failure, bad rubric, conflicting dimensions. */\nexport class JudgeError extends AgentEvalError {\n  constructor(message: string, options?: { cause?: unknown }) {\n    super('judge', message, options)\n  }\n}\n\n/** A verifier signalled a hard failure (compile, test, schema) — distinct from a low judge score. */\nexport class VerificationError extends AgentEvalError {\n  constructor(message: string, options?: { cause?: unknown }) {\n    super('verification', message, options)\n  }\n}\n\n/** Replay cache cannot satisfy a request: miss with no fallback, sink lacks list(), unsupported URL. */\nexport class ReplayError extends AgentEvalError {\n  constructor(message: string, options?: { cause?: unknown }) {\n    super('replay', message, options)\n  }\n}\n"],"mappings":";AAwBO,IAAM,iBAAN,cAA6B,MAAM;AAAA;AAAA,EAE/B;AAAA,EAET,YAAY,MAA0B,SAAiB,SAA+B;AACpF,UAAM,SAAS,OAAO;AACtB,SAAK,OAAO,KAAK,YAAY;AAC7B,SAAK,OAAO;AAAA,EACd;AACF;AAGO,IAAM,kBAAN,cAA8B,eAAe;AAAA,EAClD,YAAY,SAAiB,SAA+B;AAC1D,UAAM,cAAc,SAAS,OAAO;AAAA,EACtC;AACF;AAGO,IAAM,gBAAN,cAA4B,eAAe;AAAA,EAChD,YAAY,SAAiB,SAA+B;AAC1D,UAAM,aAAa,SAAS,OAAO;AAAA,EACrC;AACF;AAGO,IAAM,cAAN,cAA0B,eAAe;AAAA,EAC9C,YAAY,SAAiB,SAA+B;AAC1D,UAAM,UAAU,SAAS,OAAO;AAAA,EAClC;AACF;AAOO,IAAM,wBAAN,cAAoC,eAAe;AAAA,EACxD,YAAY,SAAiB,SAA+B;AAC1D,UAAM,qBAAqB,SAAS,OAAO;AAAA,EAC7C;AACF;AAGO,IAAM,aAAN,cAAyB,eAAe;AAAA,EAC7C,YAAY,SAAiB,SAA+B;AAC1D,UAAM,SAAS,SAAS,OAAO;AAAA,EACjC;AACF;AAGO,IAAM,oBAAN,cAAgC,eAAe;AAAA,EACpD,YAAY,SAAiB,SAA+B;AAC1D,UAAM,gBAAgB,SAAS,OAAO;AAAA,EACxC;AACF;AAGO,IAAM,cAAN,cAA0B,eAAe;AAAA,EAC9C,YAAY,SAAiB,SAA+B;AAC1D,UAAM,UAAU,SAAS,OAAO;AAAA,EAClC;AACF;","names":[]}

package/dist/chunk-NLMNWKVM.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../src/run-record.ts"],"sourcesContent":["/**\n * Paper-grade RunRecord schema + runtime validator.\n *\n * Every run that participates in a promotion gate, paper table, or\n * researcher loop SHOULD be recorded as a `RunRecord`. The mandatory\n * fields are exactly those the paper \"Two Loops, Three Roles\" requires\n * for reproducibility: who/what/when/cost/seed/hash, plus the search vs\n * holdout split tag and either a `searchScore` or a `holdoutScore`.\n *\n * This is intentionally NOT a replacement for the rich `Run` /\n * `ProposeReviewReport` / `ScenarioResult` types already in the\n * package. Those are runtime structures with full provenance. A\n * `RunRecord` is the analysis-time projection — the JSON-friendly\n * row you'd put in a parquet file or paste into a notebook.\n *\n * Validate at the boundary:\n *\n *   const rec = validateRunRecord(rawJson)         // throws on missing\n *   const ok  = isRunRecord(rawJson)               // boolean check\n *   const rec = parseRunRecordSafe(rawJson)        // { ok, value | error }\n *\n * The validator runs in pure TS — zod is intentionally NOT a\n * dependency. Round-trip tested in `tests/run-record.test.ts`.\n */\n\n/** Search/dev/holdout split tag. 'search' is the paper-grade alias for the\n *  combined train+test pool that the optimizer is allowed to read. */\nexport type RunSplitTag = 'search' | 'dev' | 'holdout'\n\nexport interface RunTokenUsage {\n  input: number\n  output: number\n  cached?: number\n}\n\nexport interface RunJudgeMetadata {\n  model: string\n  promptVersion: string\n  /** [0,1] confidence the judge declared. Constant judge confidence\n   *  across many runs is a fallback signal (see `canary.ts`). */\n  confidence: number\n  /** True if the judge degraded to a fallback path (rules-only,\n   *  prior-call cache, etc.). The canary uses this to alert. */\n  fallback: boolean\n}\n\nexport interface RunOutcome {\n  /** Score on the search/optimization split. Optional because a\n   *  holdout-only evaluation only fills `holdoutScore`. */\n  searchScore?: number\n  /** Score on the held-out split. Optional because a search-only run\n   *  only fills `searchScore`. At least one must be present. */\n  holdoutScore?: number\n  /** Bag of any other metric the run produced — judge dimensions,\n   *  pass/fail counters, latency stats, etc. Numeric only — keeps\n   *  reporters honest. */\n  raw: Record<string, number>\n}\n\n/**\n * Mandatory paper-grade fields for a single evaluation run. Optional\n * fields are extension points; mandatory fields throw if missing.\n *\n * Hash discipline:\n *   - `promptHash` is the sha256 of the EFFECTIVE prompt sent to the\n *     model (after any steering bundle merge).\n *   - `configHash` is the sha256 of the effective run config (model,\n *     temperature, tools, judges, splits). The pair (promptHash,\n *     configHash) uniquely identifies an experimental cell.\n *\n * Model snapshot discipline:\n *   - `model` MUST encode a snapshot version. Bare aliases like\n *     `claude-sonnet-4` or `gpt-4o` are banned — they remap silently.\n *     Use `claude-sonnet-4-6@2025-04-15` or `gpt-4o-2024-11-20`.\n */\nexport interface RunRecord {\n  /** UUID for the run. */\n  runId: string\n  /** Logical experiment grouping (a treatment vs a baseline within\n   *  the same sweep should share `experimentId`). */\n  experimentId: string\n  /** Stable identifier for the candidate (variant) being run. The\n   *  promotion gate compares two `candidateId`s on matched items. */\n  candidateId: string\n  /** RNG seed for the run. Always recorded — silent re-seeding is\n   *  the most common cause of non-reproducible numbers. */\n  seed: number\n  /** Model identifier WITH snapshot version. */\n  model: string\n  /** sha256 of the effective prompt (post-steering). */\n  promptHash: string\n  /** sha256 of the effective config. */\n  configHash: string\n  /** Git SHA the harness was run from. */\n  commitSha: string\n  /** End-to-end wall-clock duration in milliseconds. */\n  wallMs: number\n  /** Time spent queued before execution started, if known. */\n  queueMs?: number\n  /** Total USD cost. Mandatory — runs without a cost number are\n   *  unbounded by definition and must not be admitted into the gate. */\n  costUsd: number\n  /** Token usage breakdown. */\n  tokenUsage: RunTokenUsage\n  /** Judge-side metadata, if a judge was used. */\n  judgeMetadata?: RunJudgeMetadata\n  /** Per-split scores + raw bag. */\n  outcome: RunOutcome\n  /** Categorical failure tag, when the run failed and the harness\n   *  classified it. Free-form string; standard tags live in\n   *  `failure-taxonomy.ts`. */\n  failureMode?: string\n  /** Which split this run was drawn from. */\n  splitTag: RunSplitTag\n  /**\n   * Stable scenario identifier the run was scored against. Optional for\n   * backwards compatibility, but **strongly recommended**: every primitive\n   * that pairs runs by scenario (preferences, paired stats, BT tournament)\n   * keys on this. The campaign artifact populates it canonically; legacy\n   * runs without it fall back to inference from `outcome.raw.scenario_id`\n   * or `experimentId`.\n   */\n  scenarioId?: string\n}\n\n// ── Validation ───────────────────────────────────────────────────────\n\nconst MANDATORY_TOP_LEVEL = [\n  'runId',\n  'experimentId',\n  'candidateId',\n  'seed',\n  'model',\n  'promptHash',\n  'configHash',\n  'commitSha',\n  'wallMs',\n  'costUsd',\n  'tokenUsage',\n  'outcome',\n  'splitTag',\n] as const\n\nimport { ValidationError } from './errors'\n\nconst SPLIT_TAGS: ReadonlyArray<RunSplitTag> = ['search', 'dev', 'holdout']\n\nexport class RunRecordValidationError extends ValidationError {\n  readonly path: string\n  constructor(message: string, path = '') {\n    super(path ? `${message} (at ${path})` : message)\n    this.path = path\n  }\n}\n\n/**\n * Strict validator. Throws `RunRecordValidationError` on the first\n * missing or wrongly-typed field. Returns the input cast to\n * `RunRecord` on success — the validator does not coerce.\n */\nexport function validateRunRecord(input: unknown): RunRecord {\n  if (input === null || typeof input !== 'object') {\n    throw new RunRecordValidationError('expected object')\n  }\n  const obj = input as Record<string, unknown>\n\n  for (const key of MANDATORY_TOP_LEVEL) {\n    if (!(key in obj)) {\n      throw new RunRecordValidationError(`missing mandatory field \"${key}\"`)\n    }\n  }\n\n  expectString(obj.runId, 'runId')\n  expectString(obj.experimentId, 'experimentId')\n  expectString(obj.candidateId, 'candidateId')\n  expectFiniteNumber(obj.seed, 'seed')\n  expectString(obj.model, 'model')\n  expectString(obj.promptHash, 'promptHash')\n  expectString(obj.configHash, 'configHash')\n  expectString(obj.commitSha, 'commitSha')\n  expectFiniteNumber(obj.wallMs, 'wallMs')\n  if (obj.queueMs !== undefined) expectFiniteNumber(obj.queueMs, 'queueMs')\n  expectFiniteNumber(obj.costUsd, 'costUsd')\n\n  // Snapshot discipline: bare model aliases are not paper-grade.\n  if (!modelHasSnapshot(obj.model as string)) {\n    throw new RunRecordValidationError(\n      `model \"${obj.model}\" lacks a snapshot version (use 'name@YYYY-MM-DD' or 'name-YYYYMMDD')`,\n      'model',\n    )\n  }\n\n  // Token usage.\n  const tu = obj.tokenUsage\n  if (tu === null || typeof tu !== 'object') {\n    throw new RunRecordValidationError('tokenUsage must be an object', 'tokenUsage')\n  }\n  const tuRec = tu as Record<string, unknown>\n  expectFiniteNumber(tuRec.input, 'tokenUsage.input')\n  expectFiniteNumber(tuRec.output, 'tokenUsage.output')\n  if (tuRec.cached !== undefined) expectFiniteNumber(tuRec.cached, 'tokenUsage.cached')\n\n  // Judge metadata, optional.\n  if (obj.judgeMetadata !== undefined) {\n    const jm = obj.judgeMetadata\n    if (jm === null || typeof jm !== 'object') {\n      throw new RunRecordValidationError('judgeMetadata must be an object', 'judgeMetadata')\n    }\n    const jmRec = jm as Record<string, unknown>\n    expectString(jmRec.model, 'judgeMetadata.model')\n    expectString(jmRec.promptVersion, 'judgeMetadata.promptVersion')\n    expectFiniteNumber(jmRec.confidence, 'judgeMetadata.confidence')\n    if (typeof jmRec.fallback !== 'boolean') {\n      throw new RunRecordValidationError(\n        'judgeMetadata.fallback must be boolean',\n        'judgeMetadata.fallback',\n      )\n    }\n  }\n\n  // Outcome.\n  const out = obj.outcome\n  if (out === null || typeof out !== 'object') {\n    throw new RunRecordValidationError('outcome must be an object', 'outcome')\n  }\n  const outRec = out as Record<string, unknown>\n  if (outRec.searchScore !== undefined)\n    expectFiniteNumber(outRec.searchScore, 'outcome.searchScore')\n  if (outRec.holdoutScore !== undefined)\n    expectFiniteNumber(outRec.holdoutScore, 'outcome.holdoutScore')\n  if (outRec.searchScore === undefined && outRec.holdoutScore === undefined) {\n    throw new RunRecordValidationError(\n      'outcome must define searchScore or holdoutScore (or both)',\n      'outcome',\n    )\n  }\n  const raw = outRec.raw\n  if (raw === null || typeof raw !== 'object') {\n    throw new RunRecordValidationError('outcome.raw must be an object', 'outcome.raw')\n  }\n  for (const [k, v] of Object.entries(raw as Record<string, unknown>)) {\n    expectFiniteNumber(v, `outcome.raw.${k}`)\n  }\n\n  // Failure mode optional.\n  if (obj.failureMode !== undefined) expectString(obj.failureMode, 'failureMode')\n\n  // Split tag.\n  if (typeof obj.splitTag !== 'string' || !SPLIT_TAGS.includes(obj.splitTag as RunSplitTag)) {\n    throw new RunRecordValidationError(\n      `splitTag must be one of ${SPLIT_TAGS.join(', ')}, got ${String(obj.splitTag)}`,\n      'splitTag',\n    )\n  }\n\n  return input as RunRecord\n}\n\n/** Boolean validator — convenience for filtering arrays. */\nexport function isRunRecord(input: unknown): input is RunRecord {\n  try {\n    validateRunRecord(input)\n    return true\n  } catch {\n    return false\n  }\n}\n\n/** Non-throwing validator — returns a discriminated union. */\nexport function parseRunRecordSafe(\n  input: unknown,\n): { ok: true; value: RunRecord } | { ok: false; error: RunRecordValidationError } {\n  try {\n    return { ok: true, value: validateRunRecord(input) }\n  } catch (e) {\n    if (e instanceof RunRecordValidationError) return { ok: false, error: e }\n    throw e\n  }\n}\n\n/** Round-trip helper — `JSON.parse(JSON.stringify(record))` then validate. */\nexport function roundTripRunRecord(record: RunRecord): RunRecord {\n  const json = JSON.stringify(record)\n  return validateRunRecord(JSON.parse(json))\n}\n\n// ── Internals ────────────────────────────────────────────────────────\n\nfunction expectString(value: unknown, path: string): void {\n  if (typeof value !== 'string' || value.length === 0) {\n    throw new RunRecordValidationError(`expected non-empty string`, path)\n  }\n}\n\nfunction expectFiniteNumber(value: unknown, path: string): void {\n  if (typeof value !== 'number' || !Number.isFinite(value)) {\n    throw new RunRecordValidationError(`expected finite number`, path)\n  }\n}\n\n/**\n * Heuristic snapshot check. Accepts:\n *   - `name@YYYY-MM-DD` (Anthropic style: `claude-sonnet-4-6@2025-04-15`)\n *   - `name-YYYYMMDD`   (OpenAI style: `gpt-4o-2024-11-20`)\n *   - `name@<arbitrary-token>` (allow opaque snapshots like `@v3`)\n *   - explicit `:date-...` Vertex-style tags\n *\n * Rejects bare aliases like `claude-sonnet-4` or `gpt-4o` that remap\n * silently as providers ship new snapshots.\n */\nfunction modelHasSnapshot(model: string): boolean {\n  if (model.includes('@')) return true\n  if (/-\\d{8}$/.test(model)) return true\n  if (/-\\d{4}-\\d{2}-\\d{2}$/.test(model)) return true\n  if (/:date-/.test(model)) return true\n  return false\n}\n"],"mappings":";;;;;AA+HA,IAAM,sBAAsB;AAAA,EAC1B;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAIA,IAAM,aAAyC,CAAC,UAAU,OAAO,SAAS;AAEnE,IAAM,2BAAN,cAAuC,gBAAgB;AAAA,EACnD;AAAA,EACT,YAAY,SAAiB,OAAO,IAAI;AACtC,UAAM,OAAO,GAAG,OAAO,QAAQ,IAAI,MAAM,OAAO;AAChD,SAAK,OAAO;AAAA,EACd;AACF;AAOO,SAAS,kBAAkB,OAA2B;AAC3D,MAAI,UAAU,QAAQ,OAAO,UAAU,UAAU;AAC/C,UAAM,IAAI,yBAAyB,iBAAiB;AAAA,EACtD;AACA,QAAM,MAAM;AAEZ,aAAW,OAAO,qBAAqB;AACrC,QAAI,EAAE,OAAO,MAAM;AACjB,YAAM,IAAI,yBAAyB,4BAA4B,GAAG,GAAG;AAAA,IACvE;AAAA,EACF;AAEA,eAAa,IAAI,OAAO,OAAO;AAC/B,eAAa,IAAI,cAAc,cAAc;AAC7C,eAAa,IAAI,aAAa,aAAa;AAC3C,qBAAmB,IAAI,MAAM,MAAM;AACnC,eAAa,IAAI,OAAO,OAAO;AAC/B,eAAa,IAAI,YAAY,YAAY;AACzC,eAAa,IAAI,YAAY,YAAY;AACzC,eAAa,IAAI,WAAW,WAAW;AACvC,qBAAmB,IAAI,QAAQ,QAAQ;AACvC,MAAI,IAAI,YAAY,OAAW,oBAAmB,IAAI,SAAS,SAAS;AACxE,qBAAmB,IAAI,SAAS,SAAS;AAGzC,MAAI,CAAC,iBAAiB,IAAI,KAAe,GAAG;AAC1C,UAAM,IAAI;AAAA,MACR,UAAU,IAAI,KAAK;AAAA,MACnB;AAAA,IACF;AAAA,EACF;AAGA,QAAM,KAAK,IAAI;AACf,MAAI,OAAO,QAAQ,OAAO,OAAO,UAAU;AACzC,UAAM,IAAI,yBAAyB,gCAAgC,YAAY;AAAA,EACjF;AACA,QAAM,QAAQ;AACd,qBAAmB,MAAM,OAAO,kBAAkB;AAClD,qBAAmB,MAAM,QAAQ,mBAAmB;AACpD,MAAI,MAAM,WAAW,OAAW,oBAAmB,MAAM,QAAQ,mBAAmB;AAGpF,MAAI,IAAI,kBAAkB,QAAW;AACnC,UAAM,KAAK,IAAI;AACf,QAAI,OAAO,QAAQ,OAAO,OAAO,UAAU;AACzC,YAAM,IAAI,yBAAyB,mCAAmC,eAAe;AAAA,IACvF;AACA,UAAM,QAAQ;AACd,iBAAa,MAAM,OAAO,qBAAqB;AAC/C,iBAAa,MAAM,eAAe,6BAA6B;AAC/D,uBAAmB,MAAM,YAAY,0BAA0B;AAC/D,QAAI,OAAO,MAAM,aAAa,WAAW;AACvC,YAAM,IAAI;AAAA,QACR;AAAA,QACA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAGA,QAAM,MAAM,IAAI;AAChB,MAAI,QAAQ,QAAQ,OAAO,QAAQ,UAAU;AAC3C,UAAM,IAAI,yBAAyB,6BAA6B,SAAS;AAAA,EAC3E;AACA,QAAM,SAAS;AACf,MAAI,OAAO,gBAAgB;AACzB,uBAAmB,OAAO,aAAa,qBAAqB;AAC9D,MAAI,OAAO,iBAAiB;AAC1B,uBAAmB,OAAO,cAAc,sBAAsB;AAChE,MAAI,OAAO,gBAAgB,UAAa,OAAO,iBAAiB,QAAW;AACzE,UAAM,IAAI;AAAA,MACR;AAAA,MACA;AAAA,IACF;AAAA,EACF;AACA,QAAM,MAAM,OAAO;AACnB,MAAI,QAAQ,QAAQ,OAAO,QAAQ,UAAU;AAC3C,UAAM,IAAI,yBAAyB,iCAAiC,aAAa;AAAA,EACnF;AACA,aAAW,CAAC,GAAG,CAAC,KAAK,OAAO,QAAQ,GAA8B,GAAG;AACnE,uBAAmB,GAAG,eAAe,CAAC,EAAE;AAAA,EAC1C;AAGA,MAAI,IAAI,gBAAgB,OAAW,cAAa,IAAI,aAAa,aAAa;AAG9E,MAAI,OAAO,IAAI,aAAa,YAAY,CAAC,WAAW,SAAS,IAAI,QAAuB,GAAG;AACzF,UAAM,IAAI;AAAA,MACR,2BAA2B,WAAW,KAAK,IAAI,CAAC,SAAS,OAAO,IAAI,QAAQ,CAAC;AAAA,MAC7E;AAAA,IACF;AAAA,EACF;AAEA,SAAO;AACT;AAGO,SAAS,YAAY,OAAoC;AAC9D,MAAI;AACF,sBAAkB,KAAK;AACvB,WAAO;AAAA,EACT,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAGO,SAAS,mBACd,OACiF;AACjF,MAAI;AACF,WAAO,EAAE,IAAI,MAAM,OAAO,kBAAkB,KAAK,EAAE;AAAA,EACrD,SAAS,GAAG;AACV,QAAI,aAAa,yBAA0B,QAAO,EAAE,IAAI,OAAO,OAAO,EAAE;AACxE,UAAM;AAAA,EACR;AACF;AAGO,SAAS,mBAAmB,QAA8B;AAC/D,QAAM,OAAO,KAAK,UAAU,MAAM;AAClC,SAAO,kBAAkB,KAAK,MAAM,IAAI,CAAC;AAC3C;AAIA,SAAS,aAAa,OAAgB,MAAoB;AACxD,MAAI,OAAO,UAAU,YAAY,MAAM,WAAW,GAAG;AACnD,UAAM,IAAI,yBAAyB,6BAA6B,IAAI;AAAA,EACtE;AACF;AAEA,SAAS,mBAAmB,OAAgB,MAAoB;AAC9D,MAAI,OAAO,UAAU,YAAY,CAAC,OAAO,SAAS,KAAK,GAAG;AACxD,UAAM,IAAI,yBAAyB,0BAA0B,IAAI;AAAA,EACnE;AACF;AAYA,SAAS,iBAAiB,OAAwB;AAChD,MAAI,MAAM,SAAS,GAAG,EAAG,QAAO;AAChC,MAAI,UAAU,KAAK,KAAK,EAAG,QAAO;AAClC,MAAI,sBAAsB,KAAK,KAAK,EAAG,QAAO;AAC9C,MAAI,SAAS,KAAK,KAAK,EAAG,QAAO;AACjC,SAAO;AACT;","names":[]}

package/dist/chunk-RUI6SIHY.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../src/eval-campaign.ts"],"sourcesContent":["/**\n * EvalCampaign — opinionated matrix runner that wires the four\n * capture-integrity directives by construction.\n *\n * The canonical benchmark shape — matrix runner → for each\n * (variant, scenario, seed) → start a TraceEmitter → call LLMs → end the\n * run → analyze — has a bug class at the integration boundary: raw\n * events not captured, route silently wrong, integrity not asserted,\n * analyst never run. The directives in `SKILL.md § Capture integrity`\n * are the mitigations.\n *\n * `EvalCampaign` is the structural fix — consumers don't wire the\n * integrity surface themselves; the campaign owns it. Specifically:\n *\n *   - calls `assertLlmRoute` once at preflight before any work runs\n *   - constructs a per-run `TraceStore` and `RawProviderSink` via factories\n *   - constructs the `TraceEmitter` with `onRunComplete: [analyst hook]`\n *   - hands the runner an `LlmClientOptions` pre-wired with the sink and\n *     trace context — the runner can't accidentally call an LLM without\n *     capturing the raw HTTP envelope\n *   - calls `assertRunCaptured` after every `endRun` and routes failures\n *     through a configurable policy (`throw` / `mark_failed` / `log`)\n *   - assembles per-run `RunRecord`s and runs `researchReport` at the end\n *     so the campaign artifact is launch-decision-grade by default\n *   - embeds the campaign fingerprint (a SHA-256 over the canonicalised\n *     run set) and optional `preregistrationHash` in the report\n *\n * The runner contract is intentionally narrow: produce a `CampaignRunOutcome`\n * given a fully-wired `CampaignRunContext`. Everything orchestration-shaped\n * lives in the campaign. This is the inversion-of-control point — consumers\n * stop writing matrix runners and start writing scenario-runners.\n *\n * Out of scope for v1 (tracked in `docs/research-report-methodology.md`):\n *\n *   - Distributed/cluster execution (concurrency is local async)\n *   - Adaptive sampling / sequential interim looks\n *   - Resume from partial state across crashes\n *   - LLM-call retry beyond what `LlmClient` already does\n */\n\nimport { assertLlmRoute, type LlmClientOptions, type LlmRouteRequirements } from './llm-client'\nimport { canonicalize, hashJson } from './pre-registration'\nimport type {\n  RunJudgeMetadata,\n  RunOutcome,\n  RunRecord,\n  RunSplitTag,\n  RunTokenUsage,\n} from './run-record'\nimport { type ResearchReport, type ResearchReportOptions, researchReport } from './summary-report'\nimport type { RunCompleteHook } from './trace/emitter'\nimport { TraceEmitter } from './trace/emitter'\nimport {\n  assertRunCaptured,\n  RunIntegrityError,\n  type RunIntegrityExpectations,\n  type RunIntegrityReport,\n} from './trace/integrity'\nimport { FileSystemRawProviderSink, type RawProviderSink } from './trace/raw-provider-sink'\nimport type { TraceStore } from './trace/store'\n\n// ── Public types ─────────────────────────────────────────────────────────\n\nexport interface CampaignVariant<V> {\n  id: string\n  payload: V\n}\n\nexport interface CampaignScenario {\n  scenarioId: string\n  /** Free-form metadata propagated to runs and reports. */\n  tags?: Record<string, string>\n}\n\nexport interface CampaignRunContext<V> {\n  /** Stable run id. The campaign generates this; the runner does not. */\n  runId: string\n  /** Logical experiment id (campaignId by default; overridable per-run via opts). */\n  experimentId: string\n  variant: V\n  variantId: string\n  scenarioId: string\n  scenarioTags: Record<string, string>\n  seed: number\n  splitTag: RunSplitTag\n  /**\n   * The TraceEmitter for this run, with `onRunComplete` hooks pre-wired\n   * (analyst auto-execution if configured, plus integrity check). The\n   * runner MUST call `emitter.startRun` before doing any work and either\n   * `emitter.endRun` or `emitter.abortRun` before returning.\n   */\n  emitter: TraceEmitter\n  store: TraceStore\n  rawSink: RawProviderSink\n  /**\n   * Pre-wired LLM client options — `rawSink` and `traceContext` are populated\n   * so any `callLlm(req, ctx.llmOpts)` automatically captures raw HTTP. The\n   * runner can spread additional fields if needed.\n   */\n  llmOpts: LlmClientOptions\n}\n\nexport interface CampaignRunOutcome {\n  /** Did the run pass? Mirrors `RunOutcome.pass` semantics. */\n  pass: boolean\n  /** Score for the run on its split. Maps to `searchScore` or `holdoutScore`. */\n  score: number\n  /** Mandatory cost in USD. Use 0 + raw.cost_unknown=1 only if truly unknown. */\n  costUsd: number\n  tokenUsage: RunTokenUsage\n  /** Snapshot model id (e.g. `claude-sonnet-4-6@2025-04-15`). */\n  model: string\n  /** sha256 of the effective prompt sent to the model. */\n  promptHash: string\n  /** sha256 of the effective config (model, temperature, tools, judges, splits). */\n  configHash: string\n  /** Optional extra numeric metrics to land in `outcome.raw`. */\n  raw?: Record<string, number>\n  /** Optional failure-taxonomy tag if the run failed. */\n  failureMode?: string\n  /** Optional judge metadata when a judge was used. */\n  judgeMetadata?: RunJudgeMetadata\n}\n\nexport type CampaignRunner<V> = (ctx: CampaignRunContext<V>) => Promise<CampaignRunOutcome>\n\nexport type CampaignIntegrityPolicy = 'throw' | 'mark_failed' | 'log'\n\nexport interface EvalCampaignOptions<V> {\n  /**\n   * Stable id for the campaign. Used as the default `experimentId` on\n   * every run, and folded into the campaign fingerprint.\n   */\n  campaignId: string\n  variants: CampaignVariant<V>[]\n  scenarios: CampaignScenario[]\n  /** Default `[0, 1, 2]`. */\n  seeds?: number[]\n  /** Default `'holdout'` — the split that anchors a launch decision. */\n  splitTag?: RunSplitTag\n  /** Git SHA the campaign is run against. Mandatory; `RunRecord` rejects unset. */\n  commitSha: string\n  /**\n   * LLM client config. Augmented per-run with `rawSink` and `traceContext`\n   * before being passed to the runner. The campaign asserts this config\n   * matches `routeRequirements` once at preflight.\n   */\n  llmOpts: LlmClientOptions\n  /**\n   * Default `{ requireExplicitBaseUrl: true, requireAuth: true }` — fail\n   * loud if the campaign would silently fall back to the public router or\n   * run unauthenticated. Override with an empty object to disable.\n   */\n  routeRequirements?: LlmRouteRequirements\n  /**\n   * Per-run TraceStore factory. Common shape: a fresh store per run keyed\n   * on `runId`. Implementations that share a store across the campaign\n   * are valid — the campaign only writes through `emitter`.\n   */\n  storeFactory: (params: CampaignFactoryParams) => TraceStore\n  /**\n   * Per-run RawProviderSink factory. Defaults to `FileSystemRawProviderSink`\n   * rooted at `${workDir}/raw-events/${runId}` if `workDir` is supplied;\n   * otherwise required. Forensic capture is non-negotiable in a campaign\n   * run — pass `NoopRawProviderSink` explicitly if you want to opt out.\n   */\n  rawSinkFactory?: (params: CampaignFactoryParams) => RawProviderSink\n  /**\n   * Filesystem root for default `rawSinkFactory`. Ignored if\n   * `rawSinkFactory` is supplied.\n   */\n  workDir?: string\n  /**\n   * Extra `onRunComplete` hooks the campaign appends (after its own\n   * integrity-check hook). Pass `traceAnalystOnRunComplete(...)` here.\n   */\n  onRunComplete?: RunCompleteHook[]\n  /**\n   * Per-run integrity expectations. Defaults to:\n   *   `{ llmSpansMin: 1, requireRawCoverageOfLlmSpans: true, requireOutcome: true }`.\n   * Override (e.g. `{ llmSpansMin: 0 }`) for runs that don't call LLMs.\n   */\n  integrity?: RunIntegrityExpectations\n  /** Behaviour when integrity fails. Default `'mark_failed'`. */\n  onIntegrityFailure?: CampaignIntegrityPolicy\n  /**\n   * Per-run runner. Receives a fully-wired context; produces an outcome\n   * the campaign converts into a `RunRecord`.\n   */\n  runner: CampaignRunner<V>\n  /**\n   * If set, the campaign computes `researchReport` at the end. `comparator`\n   * is a `variantId`. Other fields are forwarded verbatim.\n   */\n  report?: { comparator?: string } & Omit<\n    ResearchReportOptions,\n    'comparator' | 'preregistrationHash' | 'generatedAt'\n  >\n  /**\n   * Hash of a signed `HypothesisManifest` (see `pre-registration.ts`).\n   * Embedded in the campaign fingerprint and the research report.\n   */\n  preregistrationHash?: string\n  /** Local concurrency. Default `1` (sequential). */\n  concurrency?: number\n  /**\n   * Override the time source. Tests pass a mock to make wallMs deterministic.\n   */\n  now?: () => number\n  /** Override the runId generator. Tests pin this. */\n  runId?: (params: CampaignFactoryParams) => string\n}\n\nexport interface CampaignFactoryParams {\n  campaignId: string\n  runId: string\n  variantId: string\n  scenarioId: string\n  seed: number\n}\n\nexport interface FailedRun {\n  runId: string\n  variantId: string\n  scenarioId: string\n  seed: number\n  reason: string\n  error?: string\n}\n\nexport interface EvalCampaignResult {\n  campaignId: string\n  /** SHA-256 over canonicalised `(variantIds, scenarioIds, seeds, comparator, splitTag, baseUrl, provider, preregistrationHash)`. */\n  campaignFingerprint: string\n  preregistrationHash: string | null\n  /** Successful runs only. Failed runs land in `failedRuns`. */\n  runs: RunRecord[]\n  /** Integrity reports for every successful run. */\n  integrityReports: RunIntegrityReport[]\n  failedRuns: FailedRun[]\n  /** Computed when `report` is set on options. */\n  report?: ResearchReport\n  startedAt: string\n  endedAt: string\n}\n\n// ── Implementation ───────────────────────────────────────────────────────\n\nconst DEFAULT_INTEGRITY: RunIntegrityExpectations = {\n  llmSpansMin: 1,\n  requireRawCoverageOfLlmSpans: true,\n  requireOutcome: true,\n}\n\nconst DEFAULT_ROUTE: LlmRouteRequirements = {\n  requireExplicitBaseUrl: true,\n  requireAuth: true,\n}\n\nexport async function runEvalCampaign<V>(\n  opts: EvalCampaignOptions<V>,\n): Promise<EvalCampaignResult> {\n  // ── Preflight ──────────────────────────────────────────────────────\n  assertLlmRoute(opts.llmOpts, opts.routeRequirements ?? DEFAULT_ROUTE)\n\n  if (opts.variants.length === 0) {\n    throw new Error('runEvalCampaign: variants must be non-empty.')\n  }\n  if (opts.scenarios.length === 0) {\n    throw new Error('runEvalCampaign: scenarios must be non-empty.')\n  }\n  const variantIds = new Set<string>()\n  for (const v of opts.variants) {\n    if (variantIds.has(v.id)) {\n      throw new Error(`runEvalCampaign: duplicate variant id \"${v.id}\".`)\n    }\n    variantIds.add(v.id)\n  }\n  const scenarioIds = new Set<string>()\n  for (const s of opts.scenarios) {\n    if (scenarioIds.has(s.scenarioId)) {\n      throw new Error(`runEvalCampaign: duplicate scenarioId \"${s.scenarioId}\".`)\n    }\n    scenarioIds.add(s.scenarioId)\n  }\n  if (opts.report?.comparator && !variantIds.has(opts.report.comparator)) {\n    throw new Error(\n      `runEvalCampaign: report.comparator \"${opts.report.comparator}\" is not a configured variantId.`,\n    )\n  }\n  if (!opts.commitSha) {\n    throw new Error('runEvalCampaign: commitSha is required (every RunRecord needs it).')\n  }\n\n  const seeds = opts.seeds ?? [0, 1, 2]\n  const splitTag: RunSplitTag = opts.splitTag ?? 'holdout'\n  const concurrency = Math.max(1, opts.concurrency ?? 1)\n  const integrity = { ...DEFAULT_INTEGRITY, ...(opts.integrity ?? {}) }\n  const onIntegrityFailure: CampaignIntegrityPolicy = opts.onIntegrityFailure ?? 'mark_failed'\n  const now = opts.now ?? (() => Date.now())\n  const baseUrl = (opts.llmOpts.baseUrl ?? '').replace(/\\/+$/, '')\n  const provider = opts.llmOpts.provider ?? null\n  const preregistrationHash = opts.preregistrationHash ?? null\n\n  const rawSinkFactory = opts.rawSinkFactory ?? defaultRawSinkFactory(opts.workDir)\n\n  // ── Fingerprint ────────────────────────────────────────────────────\n  const campaignFingerprint = await hashJson(\n    canonicalize({\n      campaignId: opts.campaignId,\n      variants: opts.variants.map((v) => v.id).sort(),\n      scenarios: opts.scenarios.map((s) => s.scenarioId).sort(),\n      seeds: [...seeds].sort((a, b) => a - b),\n      splitTag,\n      comparator: opts.report?.comparator ?? null,\n      baseUrl,\n      provider,\n      preregistrationHash,\n    }),\n  )\n\n  // ── Plan the matrix ────────────────────────────────────────────────\n  type Cell = { variant: CampaignVariant<V>; scenario: CampaignScenario; seed: number }\n  const cells: Cell[] = []\n  for (const variant of opts.variants) {\n    for (const scenario of opts.scenarios) {\n      for (const seed of seeds) {\n        cells.push({ variant, scenario, seed })\n      }\n    }\n  }\n\n  const startedAt = new Date(now()).toISOString()\n  const runs: RunRecord[] = []\n  const integrityReports: RunIntegrityReport[] = []\n  const failedRuns: FailedRun[] = []\n\n  // ── Execute (bounded-concurrency worker pool) ──────────────────────\n  let cursor = 0\n  async function worker(): Promise<void> {\n    while (true) {\n      const i = cursor++\n      if (i >= cells.length) return\n      const cell = cells[i]!\n      try {\n        const result = await runOneCell(cell)\n        runs.push(result.record)\n        integrityReports.push(result.integrity)\n      } catch (err) {\n        if (err instanceof CellExecutionError) {\n          failedRuns.push(err.failed)\n          if (err.integrity) integrityReports.push(err.integrity)\n        } else {\n          // Genuine bug — not a runner failure, not an integrity failure.\n          // Surface it; don't silently mask.\n          throw err\n        }\n      }\n    }\n  }\n\n  async function runOneCell(\n    cell: Cell,\n  ): Promise<{ record: RunRecord; integrity: RunIntegrityReport }> {\n    const runId = (opts.runId ?? defaultRunId)({\n      campaignId: opts.campaignId,\n      runId: '', // unused by default generator\n      variantId: cell.variant.id,\n      scenarioId: cell.scenario.scenarioId,\n      seed: cell.seed,\n    })\n    const factoryParams: CampaignFactoryParams = {\n      campaignId: opts.campaignId,\n      runId,\n      variantId: cell.variant.id,\n      scenarioId: cell.scenario.scenarioId,\n      seed: cell.seed,\n    }\n    const store = opts.storeFactory(factoryParams)\n    const rawSink = rawSinkFactory(factoryParams)\n\n    const emitter = new TraceEmitter(store, {\n      runId,\n      now: opts.now,\n      onRunComplete: opts.onRunComplete,\n    })\n\n    const llmOpts: LlmClientOptions = {\n      ...opts.llmOpts,\n      rawSink,\n      traceContext: { runId },\n    }\n\n    const ctx: CampaignRunContext<V> = {\n      runId,\n      experimentId: opts.campaignId,\n      variant: cell.variant.payload,\n      variantId: cell.variant.id,\n      scenarioId: cell.scenario.scenarioId,\n      scenarioTags: cell.scenario.tags ?? {},\n      seed: cell.seed,\n      splitTag,\n      emitter,\n      store,\n      rawSink,\n      llmOpts,\n    }\n\n    const wallStart = now()\n    let outcome: CampaignRunOutcome\n    try {\n      outcome = await opts.runner(ctx)\n    } catch (err) {\n      const message = err instanceof Error ? err.message : String(err)\n      // The runner threw mid-execution; give it a chance to have aborted.\n      try {\n        await emitter.abortRun(message)\n      } catch {\n        // Already aborted/ended; ignore.\n      }\n      throw new CellExecutionError({\n        runId,\n        variantId: cell.variant.id,\n        scenarioId: cell.scenario.scenarioId,\n        seed: cell.seed,\n        reason: 'runner_threw',\n        error: message,\n      })\n    }\n    const wallMs = now() - wallStart\n\n    const integrityReport = await assertRunCaptured(store, runId, { ...integrity, rawSink })\n    if (!integrityReport.ok) {\n      switch (onIntegrityFailure) {\n        case 'throw':\n          throw new RunIntegrityError(integrityReport)\n        case 'mark_failed':\n          throw new CellExecutionError(\n            {\n              runId,\n              variantId: cell.variant.id,\n              scenarioId: cell.scenario.scenarioId,\n              seed: cell.seed,\n              reason: 'integrity_failed',\n              error: integrityReport.issues.map((i) => i.code).join(', '),\n            },\n            integrityReport,\n          )\n        case 'log':\n          // Caller wants the run admitted with a flagged report; fall through.\n          break\n      }\n    }\n\n    const recordOutcome: RunOutcome = {\n      raw: outcome.raw ?? {},\n    }\n    if (splitTag === 'holdout') recordOutcome.holdoutScore = outcome.score\n    else recordOutcome.searchScore = outcome.score\n\n    const record: RunRecord = {\n      runId,\n      experimentId: opts.campaignId,\n      candidateId: cell.variant.id,\n      seed: cell.seed,\n      model: outcome.model,\n      promptHash: outcome.promptHash,\n      configHash: outcome.configHash,\n      commitSha: opts.commitSha,\n      wallMs,\n      costUsd: outcome.costUsd,\n      tokenUsage: outcome.tokenUsage,\n      judgeMetadata: outcome.judgeMetadata,\n      outcome: recordOutcome,\n      failureMode: outcome.failureMode,\n      splitTag,\n      scenarioId: cell.scenario.scenarioId,\n    }\n    return { record, integrity: integrityReport }\n  }\n\n  const workers = Array.from({ length: Math.min(concurrency, cells.length) }, () => worker())\n  await Promise.all(workers)\n\n  // ── Optional research report ───────────────────────────────────────\n  let report: ResearchReport | undefined\n  if (opts.report) {\n    const reportOpts: ResearchReportOptions = {\n      ...opts.report,\n      comparator: opts.report.comparator,\n      split: splitTag === 'dev' ? 'search' : splitTag,\n      generatedAt: new Date(now()).toISOString(),\n      preregistrationHash: preregistrationHash ?? undefined,\n    }\n    report = await researchReport(runs, reportOpts)\n  }\n\n  const endedAt = new Date(now()).toISOString()\n\n  return {\n    campaignId: opts.campaignId,\n    campaignFingerprint,\n    preregistrationHash,\n    runs,\n    integrityReports,\n    failedRuns,\n    report,\n    startedAt,\n    endedAt,\n  }\n}\n\n// ── Internal ─────────────────────────────────────────────────────────────\n\nclass CellExecutionError extends Error {\n  readonly failed: FailedRun\n  readonly integrity?: RunIntegrityReport\n  constructor(failed: FailedRun, integrity?: RunIntegrityReport) {\n    super(`cell ${failed.variantId}/${failed.scenarioId}@${failed.seed} failed: ${failed.reason}`)\n    this.failed = failed\n    this.integrity = integrity\n  }\n}\n\nfunction defaultRawSinkFactory(workDir: string | undefined) {\n  return (params: CampaignFactoryParams): RawProviderSink => {\n    if (!workDir) {\n      throw new Error(\n        'runEvalCampaign: rawSinkFactory not supplied and workDir not set. Pass either to enable raw provider capture, or pass `new NoopRawProviderSink()` via rawSinkFactory to opt out explicitly.',\n      )\n    }\n    return new FileSystemRawProviderSink({\n      dir: `${workDir}/raw-events/${params.runId}`,\n    })\n  }\n}\n\nfunction defaultRunId(params: CampaignFactoryParams): string {\n  // Stable across re-runs: fingerprint of (campaignId, variantId, scenarioId, seed).\n  // Caller can override via opts.runId for non-deterministic IDs.\n  const base = `${params.campaignId}::${params.variantId}::${params.scenarioId}::${params.seed}`\n  // Lightweight hex: we don't need crypto-grade here, just stability + uniqueness.\n  let h1 = 0x811c9dc5\n  let h2 = 0x12345678\n  for (let i = 0; i < base.length; i++) {\n    const c = base.charCodeAt(i)\n    h1 = Math.imul(h1 ^ c, 0x01000193) >>> 0\n    h2 = Math.imul(h2 ^ c, 0x9e3779b1) >>> 0\n  }\n  return `run-${h1.toString(16).padStart(8, '0')}${h2.toString(16).padStart(8, '0')}`\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;AAwPA,IAAM,oBAA8C;AAAA,EAClD,aAAa;AAAA,EACb,8BAA8B;AAAA,EAC9B,gBAAgB;AAClB;AAEA,IAAM,gBAAsC;AAAA,EAC1C,wBAAwB;AAAA,EACxB,aAAa;AACf;AAEA,eAAsB,gBACpB,MAC6B;AAE7B,iBAAe,KAAK,SAAS,KAAK,qBAAqB,aAAa;AAEpE,MAAI,KAAK,SAAS,WAAW,GAAG;AAC9B,UAAM,IAAI,MAAM,8CAA8C;AAAA,EAChE;AACA,MAAI,KAAK,UAAU,WAAW,GAAG;AAC/B,UAAM,IAAI,MAAM,+CAA+C;AAAA,EACjE;AACA,QAAM,aAAa,oBAAI,IAAY;AACnC,aAAW,KAAK,KAAK,UAAU;AAC7B,QAAI,WAAW,IAAI,EAAE,EAAE,GAAG;AACxB,YAAM,IAAI,MAAM,0CAA0C,EAAE,EAAE,IAAI;AAAA,IACpE;AACA,eAAW,IAAI,EAAE,EAAE;AAAA,EACrB;AACA,QAAM,cAAc,oBAAI,IAAY;AACpC,aAAW,KAAK,KAAK,WAAW;AAC9B,QAAI,YAAY,IAAI,EAAE,UAAU,GAAG;AACjC,YAAM,IAAI,MAAM,0CAA0C,EAAE,UAAU,IAAI;AAAA,IAC5E;AACA,gBAAY,IAAI,EAAE,UAAU;AAAA,EAC9B;AACA,MAAI,KAAK,QAAQ,cAAc,CAAC,WAAW,IAAI,KAAK,OAAO,UAAU,GAAG;AACtE,UAAM,IAAI;AAAA,MACR,uCAAuC,KAAK,OAAO,UAAU;AAAA,IAC/D;AAAA,EACF;AACA,MAAI,CAAC,KAAK,WAAW;AACnB,UAAM,IAAI,MAAM,oEAAoE;AAAA,EACtF;AAEA,QAAM,QAAQ,KAAK,SAAS,CAAC,GAAG,GAAG,CAAC;AACpC,QAAM,WAAwB,KAAK,YAAY;AAC/C,QAAM,cAAc,KAAK,IAAI,GAAG,KAAK,eAAe,CAAC;AACrD,QAAM,YAAY,EAAE,GAAG,mBAAmB,GAAI,KAAK,aAAa,CAAC,EAAG;AACpE,QAAM,qBAA8C,KAAK,sBAAsB;AAC/E,QAAM,MAAM,KAAK,QAAQ,MAAM,KAAK,IAAI;AACxC,QAAM,WAAW,KAAK,QAAQ,WAAW,IAAI,QAAQ,QAAQ,EAAE;AAC/D,QAAM,WAAW,KAAK,QAAQ,YAAY;AAC1C,QAAM,sBAAsB,KAAK,uBAAuB;AAExD,QAAM,iBAAiB,KAAK,kBAAkB,sBAAsB,KAAK,OAAO;AAGhF,QAAM,sBAAsB,MAAM;AAAA,IAChC,aAAa;AAAA,MACX,YAAY,KAAK;AAAA,MACjB,UAAU,KAAK,SAAS,IAAI,CAAC,MAAM,EAAE,EAAE,EAAE,KAAK;AAAA,MAC9C,WAAW,KAAK,UAAU,IAAI,CAAC,MAAM,EAAE,UAAU,EAAE,KAAK;AAAA,MACxD,OAAO,CAAC,GAAG,KAAK,EAAE,KAAK,CAAC,GAAG,MAAM,IAAI,CAAC;AAAA,MACtC;AAAA,MACA,YAAY,KAAK,QAAQ,cAAc;AAAA,MACvC;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAIA,QAAM,QAAgB,CAAC;AACvB,aAAW,WAAW,KAAK,UAAU;AACnC,eAAW,YAAY,KAAK,WAAW;AACrC,iBAAW,QAAQ,OAAO;AACxB,cAAM,KAAK,EAAE,SAAS,UAAU,KAAK,CAAC;AAAA,MACxC;AAAA,IACF;AAAA,EACF;AAEA,QAAM,YAAY,IAAI,KAAK,IAAI,CAAC,EAAE,YAAY;AAC9C,QAAM,OAAoB,CAAC;AAC3B,QAAM,mBAAyC,CAAC;AAChD,QAAM,aAA0B,CAAC;AAGjC,MAAI,SAAS;AACb,iBAAe,SAAwB;AACrC,WAAO,MAAM;AACX,YAAM,IAAI;AACV,UAAI,KAAK,MAAM,OAAQ;AACvB,YAAM,OAAO,MAAM,CAAC;AACpB,UAAI;AACF,cAAM,SAAS,MAAM,WAAW,IAAI;AACpC,aAAK,KAAK,OAAO,MAAM;AACvB,yBAAiB,KAAK,OAAO,SAAS;AAAA,MACxC,SAAS,KAAK;AACZ,YAAI,eAAe,oBAAoB;AACrC,qBAAW,KAAK,IAAI,MAAM;AAC1B,cAAI,IAAI,UAAW,kBAAiB,KAAK,IAAI,SAAS;AAAA,QACxD,OAAO;AAGL,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAEA,iBAAe,WACb,MAC+D;AAC/D,UAAM,SAAS,KAAK,SAAS,cAAc;AAAA,MACzC,YAAY,KAAK;AAAA,MACjB,OAAO;AAAA;AAAA,MACP,WAAW,KAAK,QAAQ;AAAA,MACxB,YAAY,KAAK,SAAS;AAAA,MAC1B,MAAM,KAAK;AAAA,IACb,CAAC;AACD,UAAM,gBAAuC;AAAA,MAC3C,YAAY,KAAK;AAAA,MACjB;AAAA,MACA,WAAW,KAAK,QAAQ;AAAA,MACxB,YAAY,KAAK,SAAS;AAAA,MAC1B,MAAM,KAAK;AAAA,IACb;AACA,UAAM,QAAQ,KAAK,aAAa,aAAa;AAC7C,UAAM,UAAU,eAAe,aAAa;AAE5C,UAAM,UAAU,IAAI,aAAa,OAAO;AAAA,MACtC;AAAA,MACA,KAAK,KAAK;AAAA,MACV,eAAe,KAAK;AAAA,IACtB,CAAC;AAED,UAAM,UAA4B;AAAA,MAChC,GAAG,KAAK;AAAA,MACR;AAAA,MACA,cAAc,EAAE,MAAM;AAAA,IACxB;AAEA,UAAM,MAA6B;AAAA,MACjC;AAAA,MACA,cAAc,KAAK;AAAA,MACnB,SAAS,KAAK,QAAQ;AAAA,MACtB,WAAW,KAAK,QAAQ;AAAA,MACxB,YAAY,KAAK,SAAS;AAAA,MAC1B,cAAc,KAAK,SAAS,QAAQ,CAAC;AAAA,MACrC,MAAM,KAAK;AAAA,MACX;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAEA,UAAM,YAAY,IAAI;AACtB,QAAI;AACJ,QAAI;AACF,gBAAU,MAAM,KAAK,OAAO,GAAG;AAAA,IACjC,SAAS,KAAK;AACZ,YAAM,UAAU,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;AAE/D,UAAI;AACF,cAAM,QAAQ,SAAS,OAAO;AAAA,MAChC,QAAQ;AAAA,MAER;AACA,YAAM,IAAI,mBAAmB;AAAA,QAC3B;AAAA,QACA,WAAW,KAAK,QAAQ;AAAA,QACxB,YAAY,KAAK,SAAS;AAAA,QAC1B,MAAM,KAAK;AAAA,QACX,QAAQ;AAAA,QACR,OAAO;AAAA,MACT,CAAC;AAAA,IACH;AACA,UAAM,SAAS,IAAI,IAAI;AAEvB,UAAM,kBAAkB,MAAM,kBAAkB,OAAO,OAAO,EAAE,GAAG,WAAW,QAAQ,CAAC;AACvF,QAAI,CAAC,gBAAgB,IAAI;AACvB,cAAQ,oBAAoB;AAAA,QAC1B,KAAK;AACH,gBAAM,IAAI,kBAAkB,eAAe;AAAA,QAC7C,KAAK;AACH,gBAAM,IAAI;AAAA,YACR;AAAA,cACE;AAAA,cACA,WAAW,KAAK,QAAQ;AAAA,cACxB,YAAY,KAAK,SAAS;AAAA,cAC1B,MAAM,KAAK;AAAA,cACX,QAAQ;AAAA,cACR,OAAO,gBAAgB,OAAO,IAAI,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,IAAI;AAAA,YAC5D;AAAA,YACA;AAAA,UACF;AAAA,QACF,KAAK;AAEH;AAAA,MACJ;AAAA,IACF;AAEA,UAAM,gBAA4B;AAAA,MAChC,KAAK,QAAQ,OAAO,CAAC;AAAA,IACvB;AACA,QAAI,aAAa,UAAW,eAAc,eAAe,QAAQ;AAAA,QAC5D,eAAc,cAAc,QAAQ;AAEzC,UAAM,SAAoB;AAAA,MACxB;AAAA,MACA,cAAc,KAAK;AAAA,MACnB,aAAa,KAAK,QAAQ;AAAA,MAC1B,MAAM,KAAK;AAAA,MACX,OAAO,QAAQ;AAAA,MACf,YAAY,QAAQ;AAAA,MACpB,YAAY,QAAQ;AAAA,MACpB,WAAW,KAAK;AAAA,MAChB;AAAA,MACA,SAAS,QAAQ;AAAA,MACjB,YAAY,QAAQ;AAAA,MACpB,eAAe,QAAQ;AAAA,MACvB,SAAS;AAAA,MACT,aAAa,QAAQ;AAAA,MACrB;AAAA,MACA,YAAY,KAAK,SAAS;AAAA,IAC5B;AACA,WAAO,EAAE,QAAQ,WAAW,gBAAgB;AAAA,EAC9C;AAEA,QAAM,UAAU,MAAM,KAAK,EAAE,QAAQ,KAAK,IAAI,aAAa,MAAM,MAAM,EAAE,GAAG,MAAM,OAAO,CAAC;AAC1F,QAAM,QAAQ,IAAI,OAAO;AAGzB,MAAI;AACJ,MAAI,KAAK,QAAQ;AACf,UAAM,aAAoC;AAAA,MACxC,GAAG,KAAK;AAAA,MACR,YAAY,KAAK,OAAO;AAAA,MACxB,OAAO,aAAa,QAAQ,WAAW;AAAA,MACvC,aAAa,IAAI,KAAK,IAAI,CAAC,EAAE,YAAY;AAAA,MACzC,qBAAqB,uBAAuB;AAAA,IAC9C;AACA,aAAS,MAAM,eAAe,MAAM,UAAU;AAAA,EAChD;AAEA,QAAM,UAAU,IAAI,KAAK,IAAI,CAAC,EAAE,YAAY;AAE5C,SAAO;AAAA,IACL,YAAY,KAAK;AAAA,IACjB;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;AAIA,IAAM,qBAAN,cAAiC,MAAM;AAAA,EAC5B;AAAA,EACA;AAAA,EACT,YAAY,QAAmB,WAAgC;AAC7D,UAAM,QAAQ,OAAO,SAAS,IAAI,OAAO,UAAU,IAAI,OAAO,IAAI,YAAY,OAAO,MAAM,EAAE;AAC7F,SAAK,SAAS;AACd,SAAK,YAAY;AAAA,EACnB;AACF;AAEA,SAAS,sBAAsB,SAA6B;AAC1D,SAAO,CAAC,WAAmD;AACzD,QAAI,CAAC,SAAS;AACZ,YAAM,IAAI;AAAA,QACR;AAAA,MACF;AAAA,IACF;AACA,WAAO,IAAI,0BAA0B;AAAA,MACnC,KAAK,GAAG,OAAO,eAAe,OAAO,KAAK;AAAA,IAC5C,CAAC;AAAA,EACH;AACF;AAEA,SAAS,aAAa,QAAuC;AAG3D,QAAM,OAAO,GAAG,OAAO,UAAU,KAAK,OAAO,SAAS,KAAK,OAAO,UAAU,KAAK,OAAO,IAAI;AAE5F,MAAI,KAAK;AACT,MAAI,KAAK;AACT,WAAS,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK;AACpC,UAAM,IAAI,KAAK,WAAW,CAAC;AAC3B,SAAK,KAAK,KAAK,KAAK,GAAG,QAAU,MAAM;AACvC,SAAK,KAAK,KAAK,KAAK,GAAG,UAAU,MAAM;AAAA,EACzC;AACA,SAAO,OAAO,GAAG,SAAS,EAAE,EAAE,SAAS,GAAG,GAAG,CAAC,GAAG,GAAG,SAAS,EAAE,EAAE,SAAS,GAAG,GAAG,CAAC;AACnF;","names":[]}