{
  "schemaVersion": "kingy-task-pack/1",
  "id": "coding",
  "packageVersion": "20261031.4",
  "title": "Repair integer billing rounding",
  "scope": "One public Python fixture; one fixed change; original seven tests. No general coding ability claim.",
  "evidence": {
    "taskQuality": "bounded-public-fixture-failed",
    "completion": "observed-task-failed",
    "performance": "dated-failed-observation-only",
    "memory": "insufficient-information",
    "compatibility": "historically-exercised-exact-configuration",
    "fix": "unverified"
  },
  "model": {
    "repository": "Qwen/Qwen3-4B-GGUF",
    "revision": "bc640142c66e1fdd12af0bd68f40445458f3869b",
    "file": "Qwen3-4B-Q4_K_M.gguf",
    "sha256": "7485fe6f11af29433bc51cab58009521f205840f5b4ae3a32fa7f92e8534fdf5",
    "bytes": 2497280256,
    "quantization": "Q4_K_M"
  },
  "runtime": {
    "name": "llama.cpp",
    "build": 11429,
    "commit": "d81235049",
    "binarySha256": "51afbf4ea6485b9ada19e99af97181f12baca42ef704b1e291c26a9ac0644eae"
  },
  "hardware": {
    "cpu": "Apple M4 Pro",
    "memoryGiB": 24,
    "pool": "unified",
    "os": "macOS",
    "osVersion": "26.3",
    "osBuild": "25D125"
  },
  "settings": {
    "contextTokens": 8192,
    "concurrentSequences": 1,
    "gpuLayers": 0,
    "threads": 4,
    "batchThreads": 4,
    "cacheK": "f16",
    "cacheV": "f16",
    "temperature": 0,
    "seed": 42,
    "maxTokens": 512,
    "reasoning": "off"
  },
  "dependencies": [
    {
      "name": "Python",
      "version": "3.14.0",
      "purpose": "standard-library-only runner/verifier; no pip installation"
    }
  ],
  "downloads": {
    "weightsBytes": 2497280256,
    "runtimeIncluded": false,
    "modelIncluded": false,
    "automaticDownloads": false,
    "minimumFreeDiskBytes": 16106127360
  },
  "acceptanceTestsSha256": "e106c8a64f92e2f84c73cc6a2eeb2d900a0f4365f1e044f1babe8a7a7b20e601",
  "originalSourceSha256": "948edd83e427fe08769411722579b2b8933e9db3f74d170979880a75e5153bc2",
  "taskSha256": "bd7276e4d58bfbb55e3a4b944bd3dd37268a508976df6d816a611655c39237d3",
  "passCriteria": "All seven original unittest methods pass; no edits to tests; strict one-field JSON; bounded AST source accepted; complete response; failure retained.",
  "privacy": "Synthetic public fixture. No private repositories, generated commands or arbitrary report commands are executed.",
  "repairLinks": [],
  "limitations": [
    "All three submitted attempts failed the unchanged public fixture.",
    "The .4 prompt supplied implementation guidance; no independent coding competence claim.",
    "Request budget exhausted; no further automatic model tests.",
    "Authored control is not model output or a model repair."
  ],
  "qualificationStatus": "failed-unqualified",
  "runnerChange": "Same guarded .3 runner with version strings updated; original source and all seven tests unchanged.",
  "baseline": {
    "kind": "observed-local-model-run",
    "packageVersion": "20261029.2",
    "observedAt": "2026-10-30T02:36:18.243173+00:00",
    "passed": false,
    "attempts": 1,
    "modelRequestAttempts": 1,
    "modelResponseReceived": true,
    "inferencePerformed": true,
    "testMethodsRun": 7,
    "failedMethods": [
      "test_invalid_types"
    ],
    "failedSubcases": [
      "True",
      "False"
    ],
    "answerSeconds": 6.166544499999873,
    "coldStartSeconds": 2.58899058300085,
    "peakSampledProcessRssBytes": 5899993088,
    "promptSha256": "b2551654038a7090f4e83287dc47761375a63dc274dbbddcc04298cb38eba37c",
    "acceptanceTestsSha256": "e106c8a64f92e2f84c73cc6a2eeb2d900a0f4365f1e044f1babe8a7a7b20e601",
    "generatedSourceSha256": "900d987d2306f0096338c8f1513fa44ef56ede6f0098e699b1ccc7e7d53120e6",
    "modelSha256": "7485fe6f11af29433bc51cab58009521f205840f5b4ae3a32fa7f92e8534fdf5",
    "binarySha256": "51afbf4ea6485b9ada19e99af97181f12baca42ef704b1e291c26a9ac0644eae",
    "runtime": "version: 0.6.0 (build 11429, commit d81235049)\nbuilt with AppleClang 21.0.0.21000334 for Darwin arm64",
    "hardware": {
      "chip": "Apple M4 Pro",
      "memoryBytes": 25769803776,
      "os": "26.3",
      "build": "25D125"
    },
    "serverFlags": [
      "--alias",
      "kingy-qwen3-4b",
      "--host",
      "127.0.0.1",
      "--port",
      "18482",
      "-c",
      "8192",
      "-np",
      "1",
      "-t",
      "4",
      "-tb",
      "4",
      "-ngl",
      "0",
      "--no-op-offload",
      "--cache-ram",
      "0",
      "--no-cache-prompt",
      "--jinja",
      "--reasoning",
      "off",
      "--reasoning-budget",
      "0",
      "--no-webui",
      "--offline",
      "--no-context-shift",
      "--cache-type-k",
      "f16",
      "--cache-type-v",
      "f16",
      "--temp",
      "0"
    ],
    "requestSettings": {
      "temperature": 0,
      "seed": 42,
      "max_tokens": 512,
      "stream": false,
      "cache_prompt": false,
      "chat_template_kwargs": {
        "enable_thinking": false
      },
      "response_format": {
        "type": "json_schema",
        "json_schema": {
          "name": "billing_patch",
          "strict": true,
          "schema": {
            "type": "object",
            "properties": {
              "billing_py": {
                "type": "string"
              }
            },
            "required": [
              "billing_py"
            ],
            "additionalProperties": false
          }
        }
      }
    },
    "rawResponse": "raw-response.json",
    "failedReplay": "failed-replay.json",
    "originalTestsUnchanged": true,
    "ownedServerStopped": true,
    "ownedKeyRemoved": true,
    "limitations": [
      "One exposed synthetic baseline sample; no unseen/general coding quality claim.",
      "Observed latency and process RSS were measured during a failed task on a shared machine; not corrected-task performance.",
      "No corrected model output has been generated."
    ],
    "sourceProvenance": {
      "originalRunReceiptSha256": "e0f426e727d928b5517b941738a010bb731ed2a3cf8bffd5ad5f4edd7be0fe3b",
      "originalReplaySha256": "129c0d8c2defa548f58c5a60f7c501f838f9f6e07c977effa030e2d278d6c8bc"
    }
  },
  "correction": {
    "kind": "supplied-implementation-guidance",
    "status": "observed-failed",
    "reason": "Two observed attempts failed; the second accepted bool and overbilled exact minutes.",
    "promptSha256": "bd7276e4d58bfbb55e3a4b944bd3dd37268a508976df6d816a611655c39237d3",
    "previousPromptSha256": "12da42cd36a3c5462d672d6a69f933b0924d11545dee32e1bb511ea799887c47",
    "modelRequestAttempts": 1,
    "inferencePerformed": true,
    "testsUnchanged": true,
    "claimBoundary": "Supplied implementation guidance; not independent coding competence.",
    "priorAttemptReceipt": "../../evidence/coding-run-20261031-corrective1.json",
    "priorAttemptReceiptSha256": "e51bdb3203e74993fc4efc5b6dec05cd5965b300db75d62f394436f439b2c7be",
    "priorAttemptPassed": false
  },
  "attemptAccounting": {
    "totalSubmittedModelAttempts": 3,
    "maxTotalModelAttempts": 3,
    "maximumPerCycle": 1,
    "correctedPromptSubmittedAttempts": 2,
    "remainingModelAttempts": 0
  },
  "authoredControl": {
    "evidenceType": "coordinator-authored-minimal-code-correction-with-original-test-replay",
    "checkedAt": "2026-10-30T06:07:00.713325+00:00",
    "scope": "Retained baseline model source plus one explicit boolean rejection; not another model run or a corrected-prompt result.",
    "baselineSourceSha256": "900d987d2306f0096338c8f1513fa44ef56ede6f0098e699b1ccc7e7d53120e6",
    "authoredControlSha256": "a2d32ac36952f3b2bf5d19288b7a82ea02ded709476628f106f1605605513b52",
    "acceptanceTestsSha256": "e106c8a64f92e2f84c73cc6a2eeb2d900a0f4365f1e044f1babe8a7a7b20e601",
    "baselinePassed": false,
    "authoredControlPassed": true,
    "unchangedTestMethods": 7,
    "minimalCorrection": "Add isinstance(seconds, bool) or to the original input-type rejection condition.",
    "controlSource": "authored-control.py",
    "newModelRequests": 0,
    "correctedPromptModelQualified": false,
    "sameTaskModelReproduction": "unverified",
    "sourceProvenance": {
      "coordinatorProvidedSourceSha256": "a2d32ac36952f3b2bf5d19288b7a82ea02ded709476628f106f1605605513b52",
      "originalRawResponseSha256": "8163ca4f4d6d2316661191ed12b5990a2b752e15e5d6cc5252050ede79dff03d"
    },
    "limitations": [
      "Author-supplied correction passes this fixture only; it is not model output.",
      "Original seven acceptance methods remain unchanged.",
      "Replaying retained code is distinct from a model task rerun."
    ],
    "receipt": "history/20261030-cycle1/authored-control-verification.json"
  },
  "observedResult": {
    "status": "task-failed",
    "observedAt": "2026-10-31T09:04:19.903877+00:00",
    "passed": false,
    "modelRequestAttempts": 1,
    "inferencePerformed": true,
    "promptSha256": "bd7276e4d58bfbb55e3a4b944bd3dd37268a508976df6d816a611655c39237d3",
    "sourceSha256": "a6565a43acdd6f78250f10785fddeac566759a9c1fb7d6e326764dc0f7b3b005",
    "acceptanceTestsSha256": "e106c8a64f92e2f84c73cc6a2eeb2d900a0f4365f1e044f1babe8a7a7b20e601",
    "independentReplayPassed": false,
    "failedSubcases": [
      "True",
      "False"
    ],
    "answerSeconds": 9.349002082999505,
    "receipt": "history/20261031-corrective2/result.json"
  }
}