{
  "schema": "anvil-serving.recipe-feasibility-result/v1",
  "campaign": "2026-09-19-qwen38-huihui-ninfer",
  "equations": {
    "required_tokens": "sum(requirements.tokens)",
    "policy_available_vram": "physical_vram - sum(all vram_reserves)",
    "demand": "resident + concurrency*per_sequence + required_tokens*kv_token_multiplier*per_token",
    "relative_quality_loss": "1 - quality/reference_quality",
    "warm_e2e_gain": "1 - warm_e2e/no_spec_warm_e2e",
    "tasks_per_hour_ratio": "tasks_per_hour/reference_tasks_per_hour"
  },
  "required_tokens": {
    "min": 8192,
    "max": 8192
  },
  "physical_available_bytes": {
    "min": 33015463936,
    "max": 33015463936
  },
  "policy_available_bytes": {
    "min": 29794238464,
    "max": 29794238464
  },
  "candidates": [
    {
      "id": "huihui-8k-c1-scout",
      "classification": "benchmark-survivor",
      "reasons": [
        "resource bounds pass; behavioral evidence remains incomplete"
      ],
      "axes": {
        "context_limit": "pass",
        "measured_context": "not-measured",
        "measured_context_evidence_status": "not-supplied",
        "memory": "pass",
        "correctness": {
          "status": "unknown",
          "missing_variables": [
            "deterministic_pass_rate"
          ]
        },
        "quality": {
          "status": "unknown",
          "missing_variables": [
            "quality_score",
            "reference_quality_score"
          ]
        },
        "warm_e2e_speed": {
          "status": "unknown",
          "missing_variables": [
            "warm_e2e_seconds",
            "no_spec_warm_e2e_seconds"
          ]
        },
        "tasks_per_hour": {
          "status": "unknown",
          "missing_variables": [
            "successful_tasks_per_hour",
            "reference_tasks_per_hour"
          ]
        }
      },
      "required_tokens": {
        "min": 8192,
        "max": 8192
      },
      "demand_bytes": {
        "min": 20468203520,
        "max": 28051505152
      },
      "physical_available_bytes": {
        "min": 33015463936,
        "max": 33015463936
      },
      "policy_available_bytes": {
        "min": 29794238464,
        "max": 29794238464
      },
      "physical_margin_bytes": {
        "min": 4963958784,
        "max": 12547260416
      },
      "policy_margin_bytes": {
        "min": 1742733312,
        "max": 9326034944
      },
      "estimated_max_tokens_physical": {
        "min": 311168,
        "max": 1539840
      },
      "estimated_max_tokens_policy": {
        "min": 114560,
        "max": 1146624
      },
      "load_bearing_unknowns": [],
      "planning_bound_inputs": [
        "requirements.tokens.prompt",
        "requirements.tokens.output_reserve",
        "requirements.tokens.protocol",
        "requirements.concurrency",
        "requirements.vram_reserves.headroom.variable",
        "candidates[0].runtime_context_limit_tokens",
        "candidates[0].resident_components.target",
        "candidates[0].resident_components.runtime",
        "candidates[0].per_token_components.kv",
        "candidates[0].kv_token_multiplier"
      ],
      "missing_evidence": [
        "deterministic_pass_rate",
        "measured_max_stable_context_tokens",
        "no_spec_warm_e2e_seconds",
        "quality_score",
        "reference_quality_score",
        "reference_tasks_per_hour",
        "successful_tasks_per_hour",
        "warm_e2e_seconds"
      ]
    }
  ],
  "variables": [
    {
      "path": "requirements.tokens.prompt",
      "status": "assumed",
      "unit": "tokens",
      "min": 5888,
      "max": 5888,
      "source": "scout requirement",
      "observed_at": "2026-09-19",
      "notes": ""
    },
    {
      "path": "requirements.tokens.output_reserve",
      "status": "assumed",
      "unit": "tokens",
      "min": 2048,
      "max": 2048,
      "source": "scout requirement",
      "observed_at": "2026-09-19",
      "notes": ""
    },
    {
      "path": "requirements.tokens.protocol",
      "status": "assumed",
      "unit": "tokens",
      "min": 256,
      "max": 256,
      "source": "scout requirement",
      "observed_at": "2026-09-19",
      "notes": ""
    },
    {
      "path": "requirements.concurrency",
      "status": "assumed",
      "unit": "requests",
      "min": 1,
      "max": 1,
      "source": "scout requirement",
      "observed_at": "2026-09-19",
      "notes": ""
    },
    {
      "path": "requirements.physical_vram_bytes",
      "status": "measured",
      "unit": "bytes",
      "min": 34190917632,
      "max": 34190917632,
      "source": "local nvidia-smi",
      "observed_at": "2026-09-19",
      "notes": ""
    },
    {
      "path": "requirements.vram_reserves.desktop.variable",
      "status": "measured",
      "unit": "bytes",
      "min": 1175453696,
      "max": 1175453696,
      "source": "total minus free VRAM before stop",
      "observed_at": "2026-09-19",
      "notes": ""
    },
    {
      "path": "requirements.vram_reserves.headroom.variable",
      "status": "assumed",
      "unit": "bytes",
      "min": 3221225472,
      "max": 3221225472,
      "source": "campaign default reserve",
      "observed_at": "2026-09-19",
      "notes": ""
    },
    {
      "path": "candidates[0].runtime_context_limit_tokens",
      "status": "assumed",
      "unit": "tokens",
      "min": 8192,
      "max": 8192,
      "source": "scout recipe",
      "observed_at": "2026-09-19",
      "notes": ""
    },
    {
      "path": "candidates[0].resident_components.target",
      "status": "estimated",
      "unit": "bytes",
      "min": 19327352832,
      "max": 22548578304,
      "source": "producer GPU weights 18.98 GiB text / 19.25 GiB vision; conservative interval",
      "observed_at": "2026-09-19",
      "notes": null
    },
    {
      "path": "candidates[0].resident_components.runtime",
      "status": "estimated",
      "unit": "bytes",
      "min": 1073741824,
      "max": 5368709120,
      "source": "prior same-GPU NInfer total at much larger context; planning bound requires local confirmation",
      "observed_at": "2026-09-19",
      "notes": null
    },
    {
      "path": "candidates[0].per_token_components.kv",
      "status": "estimated",
      "unit": "bytes/token",
      "min": 8192,
      "max": 16384,
      "source": "prior NInfer hybrid attention planning interval",
      "observed_at": "2026-09-19",
      "notes": null
    },
    {
      "path": "candidates[0].kv_token_multiplier",
      "status": "assumed",
      "unit": "copies",
      "min": 1,
      "max": 1,
      "source": "C1",
      "observed_at": "2026-09-19",
      "notes": ""
    },
    {
      "path": "requirements.thresholds.min_deterministic_pass_rate",
      "status": "assumed",
      "unit": "ratio",
      "min": 1,
      "max": 1,
      "source": "hard gate",
      "observed_at": "2026-09-19",
      "notes": ""
    }
  ],
  "unresolved_variables": [],
  "promotion_authority": false
}
