{
  "schema": "anvil-serving.manual-probe-summary/v1",
  "model": "gpt-oss-120b",
  "endpoint": "http://127.0.0.1:30002/v1",
  "prompt_id": "low-overhead-dashboard-architecture",
  "capture_method": "anvil_serving.benchmark.post_chat with the deterministic suite prompt and text-check engine",
  "original_cap": {
    "max_tokens": 384,
    "finish_reason": "length",
    "content_is_null": true,
    "reasoning_present": true,
    "completion_tokens": 384
  },
  "raised_cap": {
    "max_tokens": 2048,
    "finish_reason": "length",
    "content_length_chars": 5613,
    "reasoning_length_chars": 2266,
    "completion_tokens": 2048,
    "text_checks": {
      "python-stdlib": false,
      "bounded-retention": true,
      "external-evidence": false,
      "degraded-capability": false
    }
  },
  "limitations": [
    "This is a manually materialized summary of direct live responses, not the full raw response body.",
    "Character counts are not token counts.",
    "The benchmark evidence schema does not currently retain finish_reason or reasoning-channel metadata."
  ]
}