Chaque appel a un coût. Trouvez ceux à revoir.

{
  "schema_version": "1.0.0",
  "product": "ALPNAI Trace Audit",
  "payment_required": false,
  "persisted": false,
  "summary": {
    "event_count": 6,
    "task_count": 3,
    "total_cost_usd": 0.4,
    "tasks_by_outcome": {
      "success": 1,
      "failed": 1,
      "unknown": 1
    },
    "cost_by_outcome_usd": {
      "success": 0.16,
      "failed": 0.2,
      "unknown": 0.04
    },
    "cost_per_confirmed_success_usd": 0.4
  },
  "coverage": {
    "events_with_fingerprint": 4,
    "events_with_latency": 6,
    "tasks_with_known_outcome": 2,
    "cost_completeness": "caller_responsibility"
  },
  "repeated_call_candidates": [
    {
      "task_id": "research-01",
      "tool": "search",
      "occurrences": 2,
      "additional_event_count": 1,
      "additional_recorded_cost_usd": 0.04,
      "event_ids": [
        "e1",
        "e2"
      ],
      "omitted_event_ids": 0
    }
  ],
  "actions": [
    {
      "code": "investigate_failed_tasks",
      "priority": 1,
      "recorded_cost_usd": 0.2,
      "task_ids": [
        "research-01"
      ],
      "next_step": "Inspect the highest-cost failed tasks. Test a bounded retry or fallback policy on the same tasks."
    },
    {
      "code": "review_repeated_calls",
      "priority": 2,
      "recorded_cost_usd": 0.04,
      "task_ids": [
        "research-01"
      ],
      "next_step": "Check whether repeated same-tool, same-fingerprint calls are necessary. Evaluate deduplication or caching only for safe, equivalent calls."
    },
    {
      "code": "label_task_outcomes",
      "priority": 3,
      "recorded_cost_usd": 0.04,
      "task_ids": [
        "research-03"
      ],
      "next_step": "Record the business outcome of the entire task. A successful HTTP request is not proof of a successful task."
    }
  ],
  "task_evidence": [
    {
      "task_id": "research-01",
      "status": "failed",
      "cost_usd": 0.2,
      "event_ids": [
        "e1",
        "e2",
        "e3"
      ],
      "omitted_event_ids": 0
    },
    {
      "task_id": "research-02",
      "status": "success",
      "cost_usd": 0.16,
      "event_ids": [
        "e4",
        "e5"
      ],
      "omitted_event_ids": 0
    },
    {
      "task_id": "research-03",
      "status": "unknown",
      "cost_usd": 0.04,
      "event_ids": [
        "e6"
      ],
      "omitted_event_ids": 0
    }
  ],
  "method": {
    "outcome_basis": "Explicit caller-supplied whole-task verdicts; absent verdicts remain unknown.",
    "repetition_basis": "Same task, tool and caller-supplied input fingerprint; candidates for review, not proven waste.",
    "cost_basis": "All supplied event costs, including failed and unknown tasks. Missing costs are rejected; unreported events cannot be detected.",
    "overlapping_costs": "Repeated-call costs can overlap failed-task costs. Do not add them together.",
    "realized_savings_usd": null,
    "automatic_changes": false
  }
}