{
  "reviewedAt": "2026-10-02",
  "sourceRevision": "720c2a48c3d7599ec2d987cdff0283e5fea1d83b",
  "referenceDate": "2026-06-04",
  "timezone": "UTC",
  "models": {
    "originalChat": "us.anthropic.claude-sonnet-4-20250514-v1:0",
    "studioChat": "gpt-5.4",
    "reasoningEffort": "low",
    "publishedEvidenceReview": true,
    "nativeCompilation": "gpt-4.1-mini",
    "nativeEmbedding": {
      "model": "text-embedding-3-small",
      "dimensions": 512
    }
  },
  "sourceBundleSha256": "48fde0346954e7f2906370b02d0ee2655baf52fa1ba7cfc8d607d1788e5175d4",
  "mainRun": {
    "cases": 32,
    "turns": 52,
    "evidenceReview": false,
    "completed": 52,
    "timings": {
      "cciq": {
        "medianMs": 11599,
        "meanMs": 15195,
        "observedP95Ms": 35597
      },
      "studio": {
        "medianMs": 11665,
        "meanMs": 15116,
        "observedP95Ms": 39199
      }
    }
  },
  "reviewRun": {
    "cases": 5,
    "turns": 13,
    "evidenceReview": true,
    "completed": 13,
    "timings": {
      "cciq": {
        "medianMs": 17384,
        "meanMs": 20724,
        "observedP95Ms": 47442
      },
      "studio": {
        "medianMs": 19893,
        "meanMs": 24070,
        "observedP95Ms": 53765
      }
    },
    "casesReviewed": [
      "recorded-policy-follow-ups",
      "taxonomy-and-goals",
      "peer-anomalies",
      "internal-change-diagnostic",
      "survey-metrics"
    ]
  },
  "otherProfiles": {
    "turns": 4,
    "completed": 4,
    "profiles": [
      "deloitte",
      "airline-flightstar"
    ],
    "freshOriginalCalls": true
  },
  "publishedAgentSmoke": {
    "agents": 6,
    "countAndReportAnswers": 12,
    "enabledBriefingsReady": 5,
    "fnolBriefingEnabled": false
  },
  "sourceFactChecks": {
    "passed": 31,
    "failed": 0,
    "checks": [
      {
        "label": "All captured turns completed without fallback",
        "passed": true
      },
      {
        "label": "Policy confusion uses full-population multi-label counts",
        "passed": true
      },
      {
        "label": "Unqualified driver ranking uses the full snapshot",
        "passed": true
      },
      {
        "label": "recorded-policy-follow-ups: AHT breakdown uses the exact week and all six hubs",
        "passed": true
      },
      {
        "label": "recorded-policy-follow-ups: AHT headline rounding",
        "passed": true
      },
      {
        "label": "aht-comparison: AHT breakdown uses the exact week and all six hubs",
        "passed": true
      },
      {
        "label": "aht-comparison: AHT headline rounding",
        "passed": true
      },
      {
        "label": "Unqualified escalation breakdown uses full-snapshot observed flags",
        "passed": true
      },
      {
        "label": "Disposition chart covers all current contacts",
        "passed": true
      },
      {
        "label": "Atlanta baggage search: exact contacts, full hub name and weekly bounds",
        "passed": true
      },
      {
        "label": "Follow-up opens a contact from the preceding filtered result",
        "passed": true
      },
      {
        "label": "Baggage comparison includes both complete weekly hub populations",
        "passed": true
      },
      {
        "label": "Baggage headline volume, containment and AHT",
        "passed": true
      },
      {
        "label": "Explicit four-day comparison uses inclusive requested dates",
        "passed": true
      },
      {
        "label": "Survey csat: all hub values at respondent grain",
        "passed": true
      },
      {
        "label": "Survey nps: all hub values at respondent grain",
        "passed": true
      },
      {
        "label": "Survey top: all hub values at respondent grain",
        "passed": true
      },
      {
        "label": "Survey denominators cover every hub",
        "passed": true
      },
      {
        "label": "Reservation answer reads actual rebooked/flew outcomes",
        "passed": true
      },
      {
        "label": "Plural voucher examples include five topically relevant source calls",
        "passed": true
      },
      {
        "label": "Singular example contains exactly one distinct contact",
        "passed": true
      },
      {
        "label": "FNOL observed population is two accident claims",
        "passed": true
      },
      {
        "label": "FNOL follow-up retains the selected contact",
        "passed": true
      },
      {
        "label": "internal-change-diagnostic: consulted structured operational context",
        "passed": true
      },
      {
        "label": "external-disruption-diagnostic: consulted structured operational context",
        "passed": true
      },
      {
        "label": "Compiled weekly report keeps the original version",
        "passed": true
      },
      {
        "label": "Compiled weekly report has compact framing",
        "passed": true
      },
      {
        "label": "Compiled daily report keeps the original version",
        "passed": true
      },
      {
        "label": "Compiled daily report has compact framing",
        "passed": true
      },
      {
        "label": "Compiled per_interval report keeps the original version",
        "passed": true
      },
      {
        "label": "Compiled per_interval report has compact framing",
        "passed": true
      }
    ],
    "selection": "Main 52-turn capture, replacing the 13 explicitly reviewed turns with their rerun results; timings are reported separately for each run."
  },
  "browser": {
    "tests": 3,
    "passed": 3,
    "viewport": [
      1296,
      806
    ],
    "fixtures": [
      "chart matrix",
      "failed briefing retry"
    ],
    "live": [
      "compiled report",
      "citation picker",
      "transcript drawer",
      "private history"
    ]
  },
  "limitations": [
    "One observation per turn; not a load or stable p95 benchmark.",
    "Original responses reused for repair runs only after source and full conversation-history identity matched.",
    "Fact checks do not prove universal generated-prose equivalence.",
    "Delta seed and Netomi original operational tables were empty; native compilation expands original coverage."
  ],
  "routeAudit": {
    "analyticsFindings": 0,
    "existingOtherHandlerFindings": 4,
    "files": [
      "allianz-demo/reviews/route.ts",
      "embed/takeda/interactions/[conversationId]/route.ts"
    ]
  }
}