{
  "schemaVersion": 1,
  "id": "codebase-memory",
  "benchmarkId": "codebase-memory",
  "label": "Codebase Memory retrieval cost",
  "status": "complete",
  "publicationEligible": true,
  "coverage": {
    "cases": 3,
    "runs": 18,
    "trialsPerArmPerCase": 3,
    "caseIds": [
      "cbm-extension-architecture",
      "cbm-extension-config-scope",
      "cbm-project-instructions"
    ]
  },
  "arms": {
    "baseline": "felan-cbm-off",
    "candidate": "felan-cbm-on"
  },
  "trialReducer": "median",
  "caseEstimator": "ratioOfReducedSums",
  "comparisons": {
    "cost.total": {
      "baselineReducedCaseSum": 5.727142,
      "candidateReducedCaseSum": 4.135741,
      "ratioOfReducedSumsGainPercent": 27.78700091598915,
      "macroMeanCaseGainPercent": 11.269148006055824,
      "caseGainsPercent": [
        -1.031951759483988,
        42.405964690752015,
        -7.566568913100557
      ],
      "rawAttemptTotalGainPercent": 25.052814651585376
    },
    "duration.stepsMs": {
      "baselineReducedCaseSum": 1167282,
      "candidateReducedCaseSum": 1076719,
      "ratioOfReducedSumsGainPercent": 7.758450828505879,
      "macroMeanCaseGainPercent": 7.722228120587097,
      "caseGainsPercent": [
        -4.424578138032808,
        8.454515222057623,
        19.136747277736475
      ],
      "rawAttemptTotalGainPercent": 7.991582891817235
    },
    "usage.promptTokens": {
      "baselineReducedCaseSum": 7116614,
      "candidateReducedCaseSum": 4211297,
      "ratioOfReducedSumsGainPercent": 40.82442858359327,
      "macroMeanCaseGainPercent": 12.229176806088622,
      "caseGainsPercent": [
        6.630136175439376,
        51.406820352517045,
        -21.349426109690555
      ],
      "rawAttemptTotalGainPercent": 37.78097163969734
    },
    "usage.outputTokens": {
      "baselineReducedCaseSum": 28887,
      "candidateReducedCaseSum": 30014,
      "ratioOfReducedSumsGainPercent": -3.901408938276733,
      "macroMeanCaseGainPercent": -4.0493830621669975,
      "caseGainsPercent": [
        -12.590516234524642,
        -0.9349079150850556,
        1.3772749631087065
      ],
      "rawAttemptTotalGainPercent": -2.8099209881427267
    },
    "usage.totalTokens": {
      "baselineReducedCaseSum": 7146562,
      "candidateReducedCaseSum": 4240437,
      "ratioOfReducedSumsGainPercent": 40.664658055159954,
      "macroMeanCaseGainPercent": 12.155455036876923,
      "caseGainsPercent": [
        6.393150779721577,
        51.32073563138205,
        -21.24752130047286
      ],
      "rawAttemptTotalGainPercent": 37.61287572109261
    }
  },
  "cases": [
    {
      "caseId": "cbm-extension-architecture",
      "sourceCommits": [
        "e5867637569bd1c7ad08420b79ec4031a5733f57"
      ],
      "baseline": {
        "reducedByMedian": {
          "cost.total": 1.2001530000000002,
          "duration.stepsMs": 233785,
          "usage.promptTokens": 946059,
          "usage.outputTokens": 8562,
          "usage.totalTokens": 953802
        },
        "passRate": 1.0,
        "meanVerifierReward": 0.8666666666666667
      },
      "candidate": {
        "reducedByMedian": {
          "cost.total": 1.2125380000000001,
          "duration.stepsMs": 244129,
          "usage.promptTokens": 883334,
          "usage.outputTokens": 9640,
          "usage.totalTokens": 892824
        },
        "passRate": 1.0,
        "meanVerifierReward": 1.0
      }
    },
    {
      "caseId": "cbm-extension-config-scope",
      "sourceCommits": [
        "51a18d8f0c853a06867ddbd48046ad4a84307058"
      ],
      "baseline": {
        "reducedByMedian": {
          "cost.total": 3.8947869999999996,
          "duration.stepsMs": 727694,
          "usage.promptTokens": 5717681,
          "usage.outputTokens": 14226,
          "usage.totalTokens": 5733850
        },
        "passRate": 1.0,
        "meanVerifierReward": 1.0
      },
      "candidate": {
        "reducedByMedian": {
          "cost.total": 2.243165,
          "duration.stepsMs": 666171,
          "usage.promptTokens": 2778403,
          "usage.outputTokens": 14359,
          "usage.totalTokens": 2791196
        },
        "passRate": 1.0,
        "meanVerifierReward": 1.0
      }
    },
    {
      "caseId": "cbm-project-instructions",
      "sourceCommits": [
        "104faa5559029c8be9e8a1eb504d87974a5864e9"
      ],
      "baseline": {
        "reducedByMedian": {
          "cost.total": 0.632202,
          "duration.stepsMs": 205803,
          "usage.promptTokens": 452874,
          "usage.outputTokens": 6099,
          "usage.totalTokens": 458910
        },
        "passRate": 1.0,
        "meanVerifierReward": 1.0
      },
      "candidate": {
        "reducedByMedian": {
          "cost.total": 0.680038,
          "duration.stepsMs": 166419,
          "usage.promptTokens": 549560,
          "usage.outputTokens": 6015,
          "usage.totalTokens": 556417
        },
        "passRate": 1.0,
        "meanVerifierReward": 1.0
      }
    }
  ],
  "qualityDisclosure": {
    "frameworkPassRate": "All 18 published attempts reported status passed and pass=true.",
    "architectureBaselineVerifierRewards": [
      0.8,
      0.8,
      1.0
    ],
    "architectureCandidateVerifierRewards": [
      1.0,
      1.0,
      1.0
    ],
    "architectureVerifierNote": "Two Codebase Memory-off attempts received reward 0.8 (coverage 4/5); all three Codebase Memory-on attempts scored 1.0.",
    "configScopeVerifierRewards": {
      "felan-cbm-off": [
        1.0,
        1.0,
        1.0
      ],
      "felan-cbm-on": [
        1.0,
        1.0,
        1.0
      ]
    },
    "projectInstructionsVerifierRewards": {
      "felan-cbm-off": [
        1.0,
        1.0,
        1.0
      ],
      "felan-cbm-on": [
        1.0,
        1.0,
        1.0
      ]
    },
    "nonRequiredAssertionNote": "4 of 18 attempts recorded a non-required clean-tool-execution failure, primarily grep ENOENT (cbm-extension-architecture: 4); these failures did not affect framework pass status and occur in both arms."
  },
  "evidence": "../../sources/provenance/codebase-memory.json",
  "sourceBatchCoverage": {
    "declaredCases": 3,
    "declaredRuns": 18,
    "retainedCases": 3,
    "retainedRuns": 18,
    "complete": true
  },
  "runtime": {
    "packageVersion": "0.23.2",
    "extension": "@felan-ai/ext-codebase-memory 0.3.1",
    "candidateNote": "felan-cbm-on is the enabled arm, run as the harness agent felan-cbm-curated: builtinExtensions.codebaseMemory=true with extensionConfig.codebaseMemory.mode=curated, the extension's default model-facing surface."
  }
}
