@tangle-network/agent-eval 0.72.0 → 0.72.4

package/dist/analyst/index.js

@@ -0,0 +1,371 @@
+import {
+  FindingsStore,
+  RunCritic,
+  SEMANTIC_CONCEPT_JUDGE_VERSION,
+  SKILL_USAGE_ANALYST,
+  SkillUsageAnalyst,
+  behavioralAnalyst,
+  buildDefaultAnalystRegistry,
+  buildSkillUsageReport,
+  createAnalystAi,
+  createChatClient,
+  defaultIsMaterial,
+  deriveEfficiencyFindings,
+  diffFindings,
+  emitSkillUsageFindings,
+  runSemanticConceptJudge
+} from "../chunk-7W4SM7FD.js";
+import {
+  ANALYST_SEVERITIES,
+  AnalystRegistry,
+  DEFAULT_TRACE_ANALYST_KINDS,
+  FAILURE_MODE_KIND_SPEC,
+  FINDING_SUBJECT_GRAMMAR_PROMPT,
+  FINDING_SUBJECT_KINDS,
+  FindingSubjectStringSchema,
+  IMPROVEMENT_KIND_SPEC,
+  KIND_EXPECTED_SUBJECTS,
+  KNOWLEDGE_GAP_KIND_SPEC,
+  KNOWLEDGE_POISONING_KIND_SPEC,
+  RAW_FINDING_SCHEMA_PROMPT,
+  RawAnalystFindingSchema,
+  buildTraceToolsForGroup,
+  coerceJson,
+  coerceToFindingRows,
+  computeFindingId,
+  createTraceAnalystKind,
+  makeFinding,
+  parseFindingSubject,
+  parseRawFinding,
+  renderFindingSubject,
+  renderPriorFindings,
+  stripCodeFences,
+  structureFindings
+} from "../chunk-WYIHD6EB.js";
+import "../chunk-IHDHUN2X.js";
+import {
+  analyzeTraces
+} from "../chunk-VUINJM5M.js";
+import "../chunk-PC4UYEBM.js";
+import "../chunk-3BFEG2F6.js";
+import "../chunk-PZ5AY32C.js";
+
+// src/analyst/adapters.ts
+var ADAPTER_REV = "1";
+function liftSeverity(s) {
+  switch (s) {
+    case "critical":
+      return "critical";
+    case "major":
+      return "high";
+    case "minor":
+      return "medium";
+    case "info":
+      return "info";
+  }
+}
+function createTraceAnalystAdapter(opts) {
+  const id = opts.id ?? "trace-analyst";
+  const area = opts.area ?? "agent-reasoning";
+  return {
+    id,
+    description: "Runs the agent-eval trace analyst over an OTLP trace store and lifts its bulleted findings.",
+    inputKind: "trace-store",
+    cost: { kind: "llm", models: opts.model ? [opts.model] : void 0 },
+    version: `trace-analyst-${ADAPTER_REV}`,
+    async analyze(store, ctx) {
+      const out = [];
+      for (const question of opts.questions) {
+        if (ctx.signal?.aborted) break;
+        const result = await analyzeTraces(
+          { question },
+          { source: store, ai: opts.ai, model: opts.model, ...opts.extra }
+        );
+        const subject = ctx.tags?.subject ?? question.slice(0, 60);
+        if (result.findings.length === 0) {
+          out.push(
+            makeFinding({
+              analyst_id: id,
+              area,
+              subject,
+              claim: result.answer.slice(0, 200),
+              rationale: result.answer,
+              severity: "info",
+              confidence: 0.5,
+              evidence_refs: [],
+              metadata: {
+                actor_prompt_version: result.actorPromptVersion,
+                turns: result.turnCount
+              }
+            })
+          );
+          continue;
+        }
+        result.findings.forEach((claim, i) => {
+          out.push(
+            makeFinding({
+              analyst_id: id,
+              area,
+              subject,
+              claim,
+              rationale: i === 0 ? result.answer : void 0,
+              severity: "medium",
+              confidence: 0.6,
+              evidence_refs: [],
+              metadata: { question, turns: result.turnCount, finding_index: i }
+            })
+          );
+        });
+      }
+      return out;
+    }
+  };
+}
+function createVerifierAdapter(opts) {
+  const id = opts.id ?? "multi-layer-verifier";
+  const area = opts.area ?? "verification";
+  return {
+    id,
+    description: "Runs a MultiLayerVerifier and lifts each layer's findings into the analyst envelope.",
+    inputKind: "custom",
+    cost: { kind: "deterministic" },
+    version: `verifier-${ADAPTER_REV}`,
+    async analyze(env, ctx) {
+      const report = await opts.verifier.run({ env, ...opts.options });
+      const out = [];
+      for (const layer of report.layers) {
+        for (const finding of layer.findings) {
+          out.push(liftLayerFinding(id, area, layer.layer, finding));
+        }
+        if (layer.status === "fail" || layer.status === "error" || layer.status === "timeout") {
+          out.push(
+            makeFinding({
+              analyst_id: id,
+              area,
+              subject: layer.layer,
+              claim: `layer "${layer.layer}" ${layer.status}: ${layer.reason ?? "no reason given"}`,
+              severity: layer.status === "error" ? "high" : layer.status === "timeout" ? "medium" : "high",
+              confidence: 1,
+              evidence_refs: [],
+              metadata: {
+                layer_status: layer.status,
+                duration_ms: layer.durationMs,
+                score: layer.score,
+                diagnostics: layer.diagnostics
+              }
+            })
+          );
+        }
+      }
+      ctx.log?.("verifier complete", {
+        layers: report.layers.length,
+        blended: report.blendedScore,
+        all_pass: report.allPass
+      });
+      return out;
+    }
+  };
+}
+function liftLayerFinding(analyst_id, area, layer, f) {
+  return makeFinding({
+    analyst_id,
+    area,
+    subject: f.layer ?? layer,
+    claim: f.message,
+    severity: liftSeverity(f.severity),
+    confidence: 0.85,
+    evidence_refs: f.evidence ? [{ kind: "artifact", uri: "inline:evidence", excerpt: f.evidence }] : [],
+    metadata: f.detail
+  });
+}
+function createRunCriticAdapter(opts = {}) {
+  const id = opts.id ?? "run-critic";
+  const area = opts.area ?? "run-quality";
+  const critic = opts.critic ?? new RunCritic();
+  const threshold = opts.threshold ?? 0.5;
+  return {
+    id,
+    description: "Scores a single run across success / grounding / drift / tool-quality and surfaces below-threshold dimensions.",
+    inputKind: "custom",
+    cost: { kind: "deterministic" },
+    version: `run-critic-${ADAPTER_REV}`,
+    async analyze(trace) {
+      const score = critic.scoreTrace(trace);
+      const out = [];
+      const dims = [
+        ["success", "critical", "run did not complete successfully"],
+        ["goalProgress", "high", "goal progress is low"],
+        ["repoGroundedness", "high", "output is poorly grounded in the repository"],
+        ["toolUseQuality", "medium", "tool use quality is low"],
+        ["patchQuality", "medium", "no real patch/edit evidence"],
+        ["testReality", "high", "no real test/build evidence"],
+        ["finalGate", "critical", "final gate is blocking"]
+      ];
+      for (const [dim, sev, msg] of dims) {
+        const value = score[dim];
+        if (typeof value === "number" && value < threshold) {
+          out.push(
+            makeFinding({
+              analyst_id: id,
+              area,
+              subject: dim,
+              claim: msg,
+              rationale: `${dim}=${value.toFixed(2)} below threshold ${threshold}`,
+              severity: sev,
+              confidence: 1,
+              evidence_refs: [],
+              metadata: { dimension: dim, value, threshold, run_id: trace.run.runId }
+            })
+          );
+        }
+      }
+      if (score.driftPenalty > 1 - threshold) {
+        out.push(
+          makeFinding({
+            analyst_id: id,
+            area,
+            subject: "drift",
+            claim: "agent output drifted from repository signal",
+            rationale: `driftPenalty=${score.driftPenalty.toFixed(2)}`,
+            severity: "medium",
+            confidence: 0.9,
+            evidence_refs: [],
+            metadata: { drift_penalty: score.driftPenalty, notes: score.notes }
+          })
+        );
+      }
+      return out;
+    }
+  };
+}
+function createJudgeAdapter(opts) {
+  const id = opts.id ?? "judge";
+  const area = opts.area ?? "judge";
+  const threshold = opts.threshold ?? 6;
+  return {
+    id,
+    description: "Wraps an agent-eval JudgeFn into an analyst; below-threshold dimensions surface as findings.",
+    inputKind: "judge-input",
+    cost: opts.cost ?? { kind: "llm" },
+    version: `judge-${ADAPTER_REV}`,
+    async analyze(input) {
+      const scores = await opts.judge(opts.tcloud, input);
+      return scores.filter((s) => normalize10(s.score) < threshold).map((s) => liftJudgeScore(id, area, s));
+    }
+  };
+}
+function normalize10(s) {
+  return s <= 1 ? s * 10 : s;
+}
+function liftJudgeScore(analyst_id, area, s) {
+  const score10 = normalize10(s.score);
+  const severity = score10 < 3 ? "critical" : score10 < 5 ? "high" : score10 < 7 ? "medium" : "low";
+  return makeFinding({
+    analyst_id,
+    area,
+    subject: s.dimension,
+    claim: `${s.judgeName}/${s.dimension} scored ${score10.toFixed(1)}/10`,
+    rationale: s.reasoning,
+    severity,
+    confidence: 0.8,
+    evidence_refs: s.evidence ? [{ kind: "artifact", uri: "inline:evidence", excerpt: s.evidence }] : [],
+    metadata: { judge_name: s.judgeName, dimension: s.dimension, score_10: score10 }
+  });
+}
+function createSemanticConceptJudgeAdapter(opts = {}) {
+  const id = opts.id ?? "semantic-concept-judge";
+  const area = opts.area ?? "concept-coverage";
+  return {
+    id,
+    description: "Runs the semantic-concept judge and surfaces missing / weak concepts as findings.",
+    inputKind: "custom",
+    cost: { kind: "llm", models: opts.options?.model ? [opts.options.model] : void 0 },
+    version: `${SEMANTIC_CONCEPT_JUDGE_VERSION}-adapter-${ADAPTER_REV}`,
+    async analyze(input) {
+      const result = await runSemanticConceptJudge(input, opts.options);
+      if (!result.available) {
+        return [
+          makeFinding({
+            analyst_id: id,
+            area,
+            claim: "semantic-concept judge unavailable",
+            rationale: result.error,
+            severity: "info",
+            confidence: 1,
+            evidence_refs: [],
+            metadata: { reason: result.error }
+          })
+        ];
+      }
+      const out = [];
+      for (const f of result.findings) {
+        if (f.present && f.score >= 7) continue;
+        out.push(
+          makeFinding({
+            analyst_id: id,
+            area,
+            subject: f.concept,
+            claim: f.present ? `concept "${f.concept}" is weak (${f.score}/10)` : `concept "${f.concept}" is missing`,
+            rationale: f.evidence,
+            severity: liftSeverity(f.severity),
+            confidence: 0.85,
+            evidence_refs: [{ kind: "artifact", uri: "inline:evidence", excerpt: f.evidence }],
+            metadata: {
+              concept: f.concept,
+              present: f.present,
+              score_10: f.score,
+              cost_usd: result.costUsd ?? void 0
+            }
+          })
+        );
+      }
+      return out;
+    }
+  };
+}
+export {
+  ANALYST_SEVERITIES,
+  AnalystRegistry,
+  DEFAULT_TRACE_ANALYST_KINDS,
+  FAILURE_MODE_KIND_SPEC,
+  FINDING_SUBJECT_GRAMMAR_PROMPT,
+  FINDING_SUBJECT_KINDS,
+  FindingSubjectStringSchema,
+  FindingsStore,
+  IMPROVEMENT_KIND_SPEC,
+  KIND_EXPECTED_SUBJECTS,
+  KNOWLEDGE_GAP_KIND_SPEC,
+  KNOWLEDGE_POISONING_KIND_SPEC,
+  RAW_FINDING_SCHEMA_PROMPT,
+  RawAnalystFindingSchema,
+  SKILL_USAGE_ANALYST,
+  SkillUsageAnalyst,
+  behavioralAnalyst,
+  buildDefaultAnalystRegistry,
+  buildSkillUsageReport,
+  buildTraceToolsForGroup,
+  coerceJson,
+  coerceToFindingRows,
+  computeFindingId,
+  createAnalystAi,
+  createChatClient,
+  createJudgeAdapter,
+  createRunCriticAdapter,
+  createSemanticConceptJudgeAdapter,
+  createTraceAnalystAdapter,
+  createTraceAnalystKind,
+  createVerifierAdapter,
+  defaultIsMaterial,
+  deriveEfficiencyFindings,
+  diffFindings,
+  emitSkillUsageFindings,
+  liftSeverity,
+  makeFinding,
+  parseFindingSubject,
+  parseRawFinding,
+  renderFindingSubject,
+  renderPriorFindings,
+  stripCodeFences,
+  structureFindings
+};
+//# sourceMappingURL=index.js.map

package/dist/analyst/index.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/analyst/adapters.ts"],"sourcesContent":["/**\n * Adapter factories — lift each existing agent-eval primitive into the\n * Analyst contract without re-implementing it.\n *\n * Five primitives, five factories. Each one:\n *   - Builds an Analyst with a stable id (caller chooses; defaults\n *     given), a sensible default `inputKind`, a version derived from\n *     the wrapped primitive's version + an adapter revision, and an\n *     `analyze()` that calls the primitive and lifts its output to\n *     AnalystFinding[] using `makeFinding()`.\n *   - Maps severities: the existing `Severity` ('critical' | 'major' |\n *     'minor' | 'info') projects onto AnalystSeverity ('critical' |\n *     'high' | 'medium' | 'low' | 'info'); 'major' → 'high', 'minor' →\n *     'medium'. Domain analysts that want finer-grained mapping override.\n *\n * Adapters never own state. Calling the same factory twice with the\n * same primitive instance is safe.\n */\n\nimport type { AxAIService } from '@ax-llm/ax'\nimport type {\n  Finding as LayerFinding,\n  Severity as LayerSeverity,\n  MultiLayerVerifier,\n  VerifyOptions,\n} from '../multi-layer-verifier'\nimport { RunCritic, type RunTrace } from '../run-critic'\nimport {\n  runSemanticConceptJudge,\n  SEMANTIC_CONCEPT_JUDGE_VERSION,\n  type SemanticConceptJudgeInput,\n  type SemanticConceptJudgeOptions,\n} from '../semantic-concept-judge'\nimport { type AnalyzeTracesOptions, analyzeTraces } from '../trace-analyst/analyst'\nimport type { TraceAnalysisStore } from '../trace-analyst/store'\nimport type { JudgeFn, JudgeInput, JudgeScore, TCloud } from '../types'\nimport type { Analyst, AnalystFinding, AnalystSeverity } from './types'\nimport { makeFinding } from './types'\n\nconst ADAPTER_REV = '1'\n\n// ── Severity bridges ───────────────────────────────────────────────\n\nexport function liftSeverity(s: LayerSeverity): AnalystSeverity {\n  switch (s) {\n    case 'critical':\n      return 'critical'\n    case 'major':\n      return 'high'\n    case 'minor':\n      return 'medium'\n    case 'info':\n      return 'info'\n  }\n}\n\n// ── 1. analyzeTraces → Analyst ─────────────────────────────────────\n\nexport interface TraceAnalystAdapterOpts {\n  id?: string\n  area?: string\n  /** The natural-language question(s) put to the analyst. One finding per question. */\n  questions: string[]\n  /** Caller-provided AxAI service — same one trace-analyst.ts expects. */\n  ai: AxAIService\n  model?: string\n  /** Forwarded to analyzeTraces. */\n  extra?: Omit<AnalyzeTracesOptions, 'source' | 'ai' | 'model'>\n}\n\n/**\n * @deprecated Prefer `createTraceAnalystKind` + one of the failure /\n * improvement kinds from `./kinds`. This adapter wraps the legacy\n * `analyzeTraces` flow whose output is `findings:string[]` — every\n * bullet gets flat-defaulted severity `medium` / confidence `0.6`,\n * which loses the per-finding grading kinds provide via Ax structured\n * output + Zod validation. Kept for one minor while consumers migrate.\n */\nexport function createTraceAnalystAdapter(\n  opts: TraceAnalystAdapterOpts,\n): Analyst<TraceAnalysisStore> {\n  const id = opts.id ?? 'trace-analyst'\n  const area = opts.area ?? 'agent-reasoning'\n  return {\n    id,\n    description:\n      'Runs the agent-eval trace analyst over an OTLP trace store and lifts its bulleted findings.',\n    inputKind: 'trace-store',\n    cost: { kind: 'llm', models: opts.model ? [opts.model] : undefined },\n    version: `trace-analyst-${ADAPTER_REV}`,\n    async analyze(store, ctx) {\n      const out: AnalystFinding[] = []\n      for (const question of opts.questions) {\n        if (ctx.signal?.aborted) break\n        const result = await analyzeTraces(\n          { question },\n          { source: store, ai: opts.ai, model: opts.model, ...opts.extra },\n        )\n        const subject = ctx.tags?.subject ?? question.slice(0, 60)\n        // The responder produces a list of bullet strings. Each becomes\n        // one finding; the prose answer is attached as rationale on the\n        // first (so renderers that show only top-N still get context).\n        if (result.findings.length === 0) {\n          out.push(\n            makeFinding({\n              analyst_id: id,\n              area,\n              subject,\n              claim: result.answer.slice(0, 200),\n              rationale: result.answer,\n              severity: 'info',\n              confidence: 0.5,\n              evidence_refs: [],\n              metadata: {\n                actor_prompt_version: result.actorPromptVersion,\n                turns: result.turnCount,\n              },\n            }),\n          )\n          continue\n        }\n        result.findings.forEach((claim, i) => {\n          out.push(\n            makeFinding({\n              analyst_id: id,\n              area,\n              subject,\n              claim,\n              rationale: i === 0 ? result.answer : undefined,\n              severity: 'medium',\n              confidence: 0.6,\n              evidence_refs: [],\n              metadata: { question, turns: result.turnCount, finding_index: i },\n            }),\n          )\n        })\n      }\n      return out\n    },\n  }\n}\n\n// ── 2. MultiLayerVerifier → Analyst ─────────────────────────────────\n\nexport interface VerifierAdapterOpts<Env> {\n  id?: string\n  area?: string\n  verifier: MultiLayerVerifier<Env>\n  /**\n   * The verifier expects an `env` per run. Adapters take it from\n   * `AnalystRunInputs.custom[<id>]` via the registry's 'custom' routing.\n   */\n  options?: Omit<VerifyOptions<Env>, 'env'>\n}\n\nexport function createVerifierAdapter<Env>(opts: VerifierAdapterOpts<Env>): Analyst<Env> {\n  const id = opts.id ?? 'multi-layer-verifier'\n  const area = opts.area ?? 'verification'\n  return {\n    id,\n    description:\n      \"Runs a MultiLayerVerifier and lifts each layer's findings into the analyst envelope.\",\n    inputKind: 'custom',\n    cost: { kind: 'deterministic' },\n    version: `verifier-${ADAPTER_REV}`,\n    async analyze(env, ctx) {\n      const report = await opts.verifier.run({ env, ...opts.options })\n      const out: AnalystFinding[] = []\n      for (const layer of report.layers) {\n        for (const finding of layer.findings) {\n          out.push(liftLayerFinding(id, area, layer.layer, finding))\n        }\n        // Layer-level signal: a failed/error layer is itself a finding\n        // even if it didn't emit per-finding rows.\n        if (layer.status === 'fail' || layer.status === 'error' || layer.status === 'timeout') {\n          out.push(\n            makeFinding({\n              analyst_id: id,\n              area,\n              subject: layer.layer,\n              claim: `layer \"${layer.layer}\" ${layer.status}: ${layer.reason ?? 'no reason given'}`,\n              severity:\n                layer.status === 'error' ? 'high' : layer.status === 'timeout' ? 'medium' : 'high',\n              confidence: 1,\n              evidence_refs: [],\n              metadata: {\n                layer_status: layer.status,\n                duration_ms: layer.durationMs,\n                score: layer.score,\n                diagnostics: layer.diagnostics,\n              },\n            }),\n          )\n        }\n      }\n      ctx.log?.('verifier complete', {\n        layers: report.layers.length,\n        blended: report.blendedScore,\n        all_pass: report.allPass,\n      })\n      return out\n    },\n  }\n}\n\nfunction liftLayerFinding(\n  analyst_id: string,\n  area: string,\n  layer: string,\n  f: LayerFinding,\n): AnalystFinding {\n  return makeFinding({\n    analyst_id,\n    area,\n    subject: f.layer ?? layer,\n    claim: f.message,\n    severity: liftSeverity(f.severity),\n    confidence: 0.85,\n    evidence_refs: f.evidence\n      ? [{ kind: 'artifact', uri: 'inline:evidence', excerpt: f.evidence }]\n      : [],\n    metadata: f.detail,\n  })\n}\n\n// ── 3. RunCritic → Analyst ──────────────────────────────────────────\n\nexport interface RunCriticAdapterOpts {\n  id?: string\n  area?: string\n  critic?: RunCritic\n  /** Optional threshold below which a dimension is reported as a finding. Default 0.5. */\n  threshold?: number\n}\n\nexport function createRunCriticAdapter(opts: RunCriticAdapterOpts = {}): Analyst<RunTrace> {\n  const id = opts.id ?? 'run-critic'\n  const area = opts.area ?? 'run-quality'\n  const critic = opts.critic ?? new RunCritic()\n  const threshold = opts.threshold ?? 0.5\n  return {\n    id,\n    description:\n      'Scores a single run across success / grounding / drift / tool-quality and surfaces below-threshold dimensions.',\n    inputKind: 'custom',\n    cost: { kind: 'deterministic' },\n    version: `run-critic-${ADAPTER_REV}`,\n    async analyze(trace) {\n      const score = critic.scoreTrace(trace)\n      const out: AnalystFinding[] = []\n      const dims: Array<[keyof typeof score, AnalystSeverity, string]> = [\n        ['success', 'critical', 'run did not complete successfully'],\n        ['goalProgress', 'high', 'goal progress is low'],\n        ['repoGroundedness', 'high', 'output is poorly grounded in the repository'],\n        ['toolUseQuality', 'medium', 'tool use quality is low'],\n        ['patchQuality', 'medium', 'no real patch/edit evidence'],\n        ['testReality', 'high', 'no real test/build evidence'],\n        ['finalGate', 'critical', 'final gate is blocking'],\n      ]\n      for (const [dim, sev, msg] of dims) {\n        const value = score[dim] as number\n        if (typeof value === 'number' && value < threshold) {\n          out.push(\n            makeFinding({\n              analyst_id: id,\n              area,\n              subject: dim,\n              claim: msg,\n              rationale: `${dim}=${value.toFixed(2)} below threshold ${threshold}`,\n              severity: sev,\n              confidence: 1,\n              evidence_refs: [],\n              metadata: { dimension: dim, value, threshold, run_id: trace.run.runId },\n            }),\n          )\n        }\n      }\n      // Drift penalty is high → surface as a finding (inverse threshold).\n      if (score.driftPenalty > 1 - threshold) {\n        out.push(\n          makeFinding({\n            analyst_id: id,\n            area,\n            subject: 'drift',\n            claim: 'agent output drifted from repository signal',\n            rationale: `driftPenalty=${score.driftPenalty.toFixed(2)}`,\n            severity: 'medium',\n            confidence: 0.9,\n            evidence_refs: [],\n            metadata: { drift_penalty: score.driftPenalty, notes: score.notes },\n          }),\n        )\n      }\n      return out\n    },\n  }\n}\n\n// ── 4. JudgeFn → Analyst ────────────────────────────────────────────\n\nexport interface JudgeAdapterOpts {\n  id?: string\n  area?: string\n  judge: JudgeFn\n  /** TCloud handle the JudgeFn calls. */\n  tcloud: TCloud\n  /** Optional cost classification — most judges call an LLM. */\n  cost?: Analyst['cost']\n  /** Optional threshold below which a JudgeScore becomes a finding. Default 6 (on 0-10 scale). */\n  threshold?: number\n}\n\nexport function createJudgeAdapter(opts: JudgeAdapterOpts): Analyst<JudgeInput> {\n  const id = opts.id ?? 'judge'\n  const area = opts.area ?? 'judge'\n  const threshold = opts.threshold ?? 6\n  return {\n    id,\n    description:\n      'Wraps an agent-eval JudgeFn into an analyst; below-threshold dimensions surface as findings.',\n    inputKind: 'judge-input',\n    cost: opts.cost ?? { kind: 'llm' },\n    version: `judge-${ADAPTER_REV}`,\n    async analyze(input) {\n      const scores = await opts.judge(opts.tcloud, input)\n      return scores\n        .filter((s) => normalize10(s.score) < threshold)\n        .map((s) => liftJudgeScore(id, area, s))\n    },\n  }\n}\n\nfunction normalize10(s: number): number {\n  // JudgeScore convention is 0-10 but some judges emit 0-1. Coerce to 0-10.\n  return s <= 1 ? s * 10 : s\n}\n\nfunction liftJudgeScore(analyst_id: string, area: string, s: JudgeScore): AnalystFinding {\n  const score10 = normalize10(s.score)\n  const severity: AnalystSeverity =\n    score10 < 3 ? 'critical' : score10 < 5 ? 'high' : score10 < 7 ? 'medium' : 'low'\n  return makeFinding({\n    analyst_id,\n    area,\n    subject: s.dimension,\n    claim: `${s.judgeName}/${s.dimension} scored ${score10.toFixed(1)}/10`,\n    rationale: s.reasoning,\n    severity,\n    confidence: 0.8,\n    evidence_refs: s.evidence\n      ? [{ kind: 'artifact', uri: 'inline:evidence', excerpt: s.evidence }]\n      : [],\n    metadata: { judge_name: s.judgeName, dimension: s.dimension, score_10: score10 },\n  })\n}\n\n// ── 5. SemanticConceptJudge → Analyst ──────────────────────────────\n\nexport interface SemanticConceptJudgeAdapterOpts {\n  id?: string\n  area?: string\n  options?: SemanticConceptJudgeOptions\n}\n\nexport function createSemanticConceptJudgeAdapter(\n  opts: SemanticConceptJudgeAdapterOpts = {},\n): Analyst<SemanticConceptJudgeInput> {\n  const id = opts.id ?? 'semantic-concept-judge'\n  const area = opts.area ?? 'concept-coverage'\n  return {\n    id,\n    description:\n      'Runs the semantic-concept judge and surfaces missing / weak concepts as findings.',\n    inputKind: 'custom',\n    cost: { kind: 'llm', models: opts.options?.model ? [opts.options.model] : undefined },\n    version: `${SEMANTIC_CONCEPT_JUDGE_VERSION}-adapter-${ADAPTER_REV}`,\n    async analyze(input) {\n      const result = await runSemanticConceptJudge(input, opts.options)\n      if (!result.available) {\n        return [\n          makeFinding({\n            analyst_id: id,\n            area,\n            claim: 'semantic-concept judge unavailable',\n            rationale: result.error,\n            severity: 'info',\n            confidence: 1,\n            evidence_refs: [],\n            metadata: { reason: result.error },\n          }),\n        ]\n      }\n      const out: AnalystFinding[] = []\n      for (const f of result.findings) {\n        // Only surface gaps: missing concepts or low scores. Concepts at\n        // 7+/10 with present=true are not findings — they're successes.\n        if (f.present && f.score >= 7) continue\n        out.push(\n          makeFinding({\n            analyst_id: id,\n            area,\n            subject: f.concept,\n            claim: f.present\n              ? `concept \"${f.concept}\" is weak (${f.score}/10)`\n              : `concept \"${f.concept}\" is missing`,\n            rationale: f.evidence,\n            severity: liftSeverity(f.severity),\n            confidence: 0.85,\n            evidence_refs: [{ kind: 'artifact', uri: 'inline:evidence', excerpt: f.evidence }],\n            metadata: {\n              concept: f.concept,\n              present: f.present,\n              score_10: f.score,\n              cost_usd: result.costUsd ?? undefined,\n            },\n          }),\n        )\n      }\n      return out\n    },\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAuCA,IAAM,cAAc;AAIb,SAAS,aAAa,GAAmC;AAC9D,UAAQ,GAAG;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,IACT,KAAK;AACH,aAAO;AAAA,EACX;AACF;AAwBO,SAAS,0BACd,MAC6B;AAC7B,QAAM,KAAK,KAAK,MAAM;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,SAAO;AAAA,IACL;AAAA,IACA,aACE;AAAA,IACF,WAAW;AAAA,IACX,MAAM,EAAE,MAAM,OAAO,QAAQ,KAAK,QAAQ,CAAC,KAAK,KAAK,IAAI,OAAU;AAAA,IACnE,SAAS,iBAAiB,WAAW;AAAA,IACrC,MAAM,QAAQ,OAAO,KAAK;AACxB,YAAM,MAAwB,CAAC;AAC/B,iBAAW,YAAY,KAAK,WAAW;AACrC,YAAI,IAAI,QAAQ,QAAS;AACzB,cAAM,SAAS,MAAM;AAAA,UACnB,EAAE,SAAS;AAAA,UACX,EAAE,QAAQ,OAAO,IAAI,KAAK,IAAI,OAAO,KAAK,OAAO,GAAG,KAAK,MAAM;AAAA,QACjE;AACA,cAAM,UAAU,IAAI,MAAM,WAAW,SAAS,MAAM,GAAG,EAAE;AAIzD,YAAI,OAAO,SAAS,WAAW,GAAG;AAChC,cAAI;AAAA,YACF,YAAY;AAAA,cACV,YAAY;AAAA,cACZ;AAAA,cACA;AAAA,cACA,OAAO,OAAO,OAAO,MAAM,GAAG,GAAG;AAAA,cACjC,WAAW,OAAO;AAAA,cAClB,UAAU;AAAA,cACV,YAAY;AAAA,cACZ,eAAe,CAAC;AAAA,cAChB,UAAU;AAAA,gBACR,sBAAsB,OAAO;AAAA,gBAC7B,OAAO,OAAO;AAAA,cAChB;AAAA,YACF,CAAC;AAAA,UACH;AACA;AAAA,QACF;AACA,eAAO,SAAS,QAAQ,CAAC,OAAO,MAAM;AACpC,cAAI;AAAA,YACF,YAAY;AAAA,cACV,YAAY;AAAA,cACZ;AAAA,cACA;AAAA,cACA;AAAA,cACA,WAAW,MAAM,IAAI,OAAO,SAAS;AAAA,cACrC,UAAU;AAAA,cACV,YAAY;AAAA,cACZ,eAAe,CAAC;AAAA,cAChB,UAAU,EAAE,UAAU,OAAO,OAAO,WAAW,eAAe,EAAE;AAAA,YAClE,CAAC;AAAA,UACH;AAAA,QACF,CAAC;AAAA,MACH;AACA,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAeO,SAAS,sBAA2B,MAA8C;AACvF,QAAM,KAAK,KAAK,MAAM;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,SAAO;AAAA,IACL;AAAA,IACA,aACE;AAAA,IACF,WAAW;AAAA,IACX,MAAM,EAAE,MAAM,gBAAgB;AAAA,IAC9B,SAAS,YAAY,WAAW;AAAA,IAChC,MAAM,QAAQ,KAAK,KAAK;AACtB,YAAM,SAAS,MAAM,KAAK,SAAS,IAAI,EAAE,KAAK,GAAG,KAAK,QAAQ,CAAC;AAC/D,YAAM,MAAwB,CAAC;AAC/B,iBAAW,SAAS,OAAO,QAAQ;AACjC,mBAAW,WAAW,MAAM,UAAU;AACpC,cAAI,KAAK,iBAAiB,IAAI,MAAM,MAAM,OAAO,OAAO,CAAC;AAAA,QAC3D;AAGA,YAAI,MAAM,WAAW,UAAU,MAAM,WAAW,WAAW,MAAM,WAAW,WAAW;AACrF,cAAI;AAAA,YACF,YAAY;AAAA,cACV,YAAY;AAAA,cACZ;AAAA,cACA,SAAS,MAAM;AAAA,cACf,OAAO,UAAU,MAAM,KAAK,KAAK,MAAM,MAAM,KAAK,MAAM,UAAU,iBAAiB;AAAA,cACnF,UACE,MAAM,WAAW,UAAU,SAAS,MAAM,WAAW,YAAY,WAAW;AAAA,cAC9E,YAAY;AAAA,cACZ,eAAe,CAAC;AAAA,cAChB,UAAU;AAAA,gBACR,cAAc,MAAM;AAAA,gBACpB,aAAa,MAAM;AAAA,gBACnB,OAAO,MAAM;AAAA,gBACb,aAAa,MAAM;AAAA,cACrB;AAAA,YACF,CAAC;AAAA,UACH;AAAA,QACF;AAAA,MACF;AACA,UAAI,MAAM,qBAAqB;AAAA,QAC7B,QAAQ,OAAO,OAAO;AAAA,QACtB,SAAS,OAAO;AAAA,QAChB,UAAU,OAAO;AAAA,MACnB,CAAC;AACD,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAEA,SAAS,iBACP,YACA,MACA,OACA,GACgB;AAChB,SAAO,YAAY;AAAA,IACjB;AAAA,IACA;AAAA,IACA,SAAS,EAAE,SAAS;AAAA,IACpB,OAAO,EAAE;AAAA,IACT,UAAU,aAAa,EAAE,QAAQ;AAAA,IACjC,YAAY;AAAA,IACZ,eAAe,EAAE,WACb,CAAC,EAAE,MAAM,YAAY,KAAK,mBAAmB,SAAS,EAAE,SAAS,CAAC,IAClE,CAAC;AAAA,IACL,UAAU,EAAE;AAAA,EACd,CAAC;AACH;AAYO,SAAS,uBAAuB,OAA6B,CAAC,GAAsB;AACzF,QAAM,KAAK,KAAK,MAAM;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,QAAM,SAAS,KAAK,UAAU,IAAI,UAAU;AAC5C,QAAM,YAAY,KAAK,aAAa;AACpC,SAAO;AAAA,IACL;AAAA,IACA,aACE;AAAA,IACF,WAAW;AAAA,IACX,MAAM,EAAE,MAAM,gBAAgB;AAAA,IAC9B,SAAS,cAAc,WAAW;AAAA,IAClC,MAAM,QAAQ,OAAO;AACnB,YAAM,QAAQ,OAAO,WAAW,KAAK;AACrC,YAAM,MAAwB,CAAC;AAC/B,YAAM,OAA6D;AAAA,QACjE,CAAC,WAAW,YAAY,mCAAmC;AAAA,QAC3D,CAAC,gBAAgB,QAAQ,sBAAsB;AAAA,QAC/C,CAAC,oBAAoB,QAAQ,6CAA6C;AAAA,QAC1E,CAAC,kBAAkB,UAAU,yBAAyB;AAAA,QACtD,CAAC,gBAAgB,UAAU,6BAA6B;AAAA,QACxD,CAAC,eAAe,QAAQ,6BAA6B;AAAA,QACrD,CAAC,aAAa,YAAY,wBAAwB;AAAA,MACpD;AACA,iBAAW,CAAC,KAAK,KAAK,GAAG,KAAK,MAAM;AAClC,cAAM,QAAQ,MAAM,GAAG;AACvB,YAAI,OAAO,UAAU,YAAY,QAAQ,WAAW;AAClD,cAAI;AAAA,YACF,YAAY;AAAA,cACV,YAAY;AAAA,cACZ;AAAA,cACA,SAAS;AAAA,cACT,OAAO;AAAA,cACP,WAAW,GAAG,GAAG,IAAI,MAAM,QAAQ,CAAC,CAAC,oBAAoB,SAAS;AAAA,cAClE,UAAU;AAAA,cACV,YAAY;AAAA,cACZ,eAAe,CAAC;AAAA,cAChB,UAAU,EAAE,WAAW,KAAK,OAAO,WAAW,QAAQ,MAAM,IAAI,MAAM;AAAA,YACxE,CAAC;AAAA,UACH;AAAA,QACF;AAAA,MACF;AAEA,UAAI,MAAM,eAAe,IAAI,WAAW;AACtC,YAAI;AAAA,UACF,YAAY;AAAA,YACV,YAAY;AAAA,YACZ;AAAA,YACA,SAAS;AAAA,YACT,OAAO;AAAA,YACP,WAAW,gBAAgB,MAAM,aAAa,QAAQ,CAAC,CAAC;AAAA,YACxD,UAAU;AAAA,YACV,YAAY;AAAA,YACZ,eAAe,CAAC;AAAA,YAChB,UAAU,EAAE,eAAe,MAAM,cAAc,OAAO,MAAM,MAAM;AAAA,UACpE,CAAC;AAAA,QACH;AAAA,MACF;AACA,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAgBO,SAAS,mBAAmB,MAA6C;AAC9E,QAAM,KAAK,KAAK,MAAM;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,QAAM,YAAY,KAAK,aAAa;AACpC,SAAO;AAAA,IACL;AAAA,IACA,aACE;AAAA,IACF,WAAW;AAAA,IACX,MAAM,KAAK,QAAQ,EAAE,MAAM,MAAM;AAAA,IACjC,SAAS,SAAS,WAAW;AAAA,IAC7B,MAAM,QAAQ,OAAO;AACnB,YAAM,SAAS,MAAM,KAAK,MAAM,KAAK,QAAQ,KAAK;AAClD,aAAO,OACJ,OAAO,CAAC,MAAM,YAAY,EAAE,KAAK,IAAI,SAAS,EAC9C,IAAI,CAAC,MAAM,eAAe,IAAI,MAAM,CAAC,CAAC;AAAA,IAC3C;AAAA,EACF;AACF;AAEA,SAAS,YAAY,GAAmB;AAEtC,SAAO,KAAK,IAAI,IAAI,KAAK;AAC3B;AAEA,SAAS,eAAe,YAAoB,MAAc,GAA+B;AACvF,QAAM,UAAU,YAAY,EAAE,KAAK;AACnC,QAAM,WACJ,UAAU,IAAI,aAAa,UAAU,IAAI,SAAS,UAAU,IAAI,WAAW;AAC7E,SAAO,YAAY;AAAA,IACjB;AAAA,IACA;AAAA,IACA,SAAS,EAAE;AAAA,IACX,OAAO,GAAG,EAAE,SAAS,IAAI,EAAE,SAAS,WAAW,QAAQ,QAAQ,CAAC,CAAC;AAAA,IACjE,WAAW,EAAE;AAAA,IACb;AAAA,IACA,YAAY;AAAA,IACZ,eAAe,EAAE,WACb,CAAC,EAAE,MAAM,YAAY,KAAK,mBAAmB,SAAS,EAAE,SAAS,CAAC,IAClE,CAAC;AAAA,IACL,UAAU,EAAE,YAAY,EAAE,WAAW,WAAW,EAAE,WAAW,UAAU,QAAQ;AAAA,EACjF,CAAC;AACH;AAUO,SAAS,kCACd,OAAwC,CAAC,GACL;AACpC,QAAM,KAAK,KAAK,MAAM;AACtB,QAAM,OAAO,KAAK,QAAQ;AAC1B,SAAO;AAAA,IACL;AAAA,IACA,aACE;AAAA,IACF,WAAW;AAAA,IACX,MAAM,EAAE,MAAM,OAAO,QAAQ,KAAK,SAAS,QAAQ,CAAC,KAAK,QAAQ,KAAK,IAAI,OAAU;AAAA,IACpF,SAAS,GAAG,8BAA8B,YAAY,WAAW;AAAA,IACjE,MAAM,QAAQ,OAAO;AACnB,YAAM,SAAS,MAAM,wBAAwB,OAAO,KAAK,OAAO;AAChE,UAAI,CAAC,OAAO,WAAW;AACrB,eAAO;AAAA,UACL,YAAY;AAAA,YACV,YAAY;AAAA,YACZ;AAAA,YACA,OAAO;AAAA,YACP,WAAW,OAAO;AAAA,YAClB,UAAU;AAAA,YACV,YAAY;AAAA,YACZ,eAAe,CAAC;AAAA,YAChB,UAAU,EAAE,QAAQ,OAAO,MAAM;AAAA,UACnC,CAAC;AAAA,QACH;AAAA,MACF;AACA,YAAM,MAAwB,CAAC;AAC/B,iBAAW,KAAK,OAAO,UAAU;AAG/B,YAAI,EAAE,WAAW,EAAE,SAAS,EAAG;AAC/B,YAAI;AAAA,UACF,YAAY;AAAA,YACV,YAAY;AAAA,YACZ;AAAA,YACA,SAAS,EAAE;AAAA,YACX,OAAO,EAAE,UACL,YAAY,EAAE,OAAO,cAAc,EAAE,KAAK,SAC1C,YAAY,EAAE,OAAO;AAAA,YACzB,WAAW,EAAE;AAAA,YACb,UAAU,aAAa,EAAE,QAAQ;AAAA,YACjC,YAAY;AAAA,YACZ,eAAe,CAAC,EAAE,MAAM,YAAY,KAAK,mBAAmB,SAAS,EAAE,SAAS,CAAC;AAAA,YACjF,UAAU;AAAA,cACR,SAAS,EAAE;AAAA,cACX,SAAS,EAAE;AAAA,cACX,UAAU,EAAE;AAAA,cACZ,UAAU,OAAO,WAAW;AAAA,YAC9B;AAAA,UACF,CAAC;AAAA,QACH;AAAA,MACF;AACA,aAAO;AAAA,IACT;AAAA,EACF;AACF;","names":[]}

package/dist/analyst-t7zZS3TV.d.ts

@@ -0,0 +1,88 @@
+import { AxAIService } from '@ax-llm/ax';
+import { T as TraceAnalysisStore } from './store-GmBE2pZZ.js';
+
+interface AnalyzeTracesInput {
+    /** The user-facing question. Domain framing belongs here, not in the
+     *  actor description. */
+    question: string;
+}
+interface AnalyzeTracesResult {
+    /** The responder's prose answer. */
+    answer: string;
+    /** Bulleted findings extracted from the responder's structured output. */
+    findings: string[];
+    /** Per-actor-turn snapshots captured via `actorTurnCallback`. */
+    turns: AnalyzeTracesTurnSnapshot[];
+    /** Total turns the actor took. */
+    turnCount: number;
+    /** Token usage by role. */
+    usage: TraceAnalystUsage;
+    /** Full system + assistant + tool message log by role. */
+    chatLog: TraceAnalystChatLog;
+    /** Prompt version that produced this run. */
+    actorPromptVersion: string;
+}
+interface TraceAnalystUsage {
+    actor: TraceAnalystUsageEntry[];
+    responder: TraceAnalystUsageEntry[];
+}
+interface TraceAnalystUsageEntry {
+    [key: string]: unknown;
+}
+interface TraceAnalystChatLog {
+    actor: TraceAnalystChatMessage[];
+    responder: TraceAnalystChatMessage[];
+}
+interface TraceAnalystChatMessage {
+    [key: string]: unknown;
+}
+interface AnalyzeTracesTurnSnapshot {
+    turn: number;
+    isError: boolean;
+    /** The JS code the actor produced for this turn. */
+    code: string;
+    /** The formatted action-log entry the actor sees on the next turn. */
+    output: string;
+    /** Provider thought (when `actorOptions.showThoughts` is true and the
+     *  provider returns it). */
+    thought?: string;
+}
+interface AnalyzeTracesOptions {
+    /** Trace data source. Pass either an OTLP-JSONL path or a custom store. */
+    source: string | TraceAnalysisStore;
+    /** Caller-provided AxAIService. */
+    ai: AxAIService;
+    /** Model id forwarded to actor + responder. */
+    model?: string;
+    /** Recursion depth. 0 = no sub-agent dispatch. Default 1. */
+    maxDepth?: number;
+    /** Maximum actor turns. Default 12. */
+    maxTurns?: number;
+    /** Maximum parallel sub-agent calls in batched llmQuery. Default 2. */
+    maxParallelSubagents?: number;
+    /** Override the actor description. */
+    actorDescription?: string;
+    /** Override the subagent description. */
+    subagentDescription?: string;
+    /** Per-turn observability hook. */
+    onTurn?: (turn: AnalyzeTracesTurnSnapshot) => void | Promise<void>;
+    /** Override max runtime characters per turn. Default 6000. */
+    maxRuntimeChars?: number;
+    /** When set, every turn's snapshot is appended to this JSONL file
+     *  immediately. If the analyst crashes mid-loop (provider 503,
+     *  network error, validator reject) the partial reasoning is still
+     *  on disk. Replay the file with the responder afterward to recover
+     *  evidence. */
+    progressLogPath?: string;
+}
+/**
+ * Run the trace analyst.
+ *
+ * Throws:
+ *   - `TraceFileMissingError` if `source` is a path and doesn't exist.
+ *   - `AxAgentClarificationError` if the analyst asks for clarification.
+ *   - Provider errors (auth, rate limits) propagate from the AI service.
+ */
+declare function analyzeTraces(input: AnalyzeTracesInput, options: AnalyzeTracesOptions): Promise<AnalyzeTracesResult>;
+
+export { type AnalyzeTracesOptions as A, type AnalyzeTracesInput as a, type AnalyzeTracesResult as b, type AnalyzeTracesTurnSnapshot as c, analyzeTraces as d };