NodeRoom/docs/eval/proofloop-prod-proxy-longrun-plan.json at main · HomenShum/NodeRoom · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
{
  "schema": "proofloop-prod-proxy-longrun-v1",
  "generatedAt": "2026-07-05T08:11:53.015Z",
  "runId": "prod-proxy-longrun-2026-07-05T08-11-53-015Z",
  "baseUrl": "https://noderoom.live",
  "root": "D:\\VSCode Projects\\cafecorner_nodebench\\nodebench_ai4\\noderoom",
  "matrixDigest": "07831f7607550f00795461f19b54efc663eb0a5535ccb859dd819b51d1b499a6",
  "budget": {
    "capUsd": 100,
    "unknownAttemptCostUsd": 0.02,
    "historicalMeasuredSpendUsd": 0.502,
    "queuedEstimatedNewSpendUsd": 99.960136,
    "fullMatrixEstimatedUsd": 246.812536,
    "fullCurrentModelMatrixFitsBudget": false,
    "runnableQueueFitsBudget": true
  },
  "controlPlane": {
    "goalContract": "Run every tracked prod proxy benchmark task through the live NodeRoom browser UI with memory mode off, score each model-task attempt, and refuse an all-task winner until the full matrix is proven.",
    "evaluator": "Detached verifier receipts and scorecards decide pass/fail; model output or agent narration is not accepted as proof.",
    "verifiers": [
      "real prod URL must be https://noderoom.live",
      "memoryModeAllowed=false for every attempt",
      "browser evidence must include trace/cost/verifier receipts before a pass is promoted",
      "official score claims stay false unless upstream scorer/judge artifacts are imported"
    ],
    "outerLoop": "state is persisted under .proofloop/prod-proxy-longrun/<runId>; running attempts carry a lease and can be resumed after process/network failure; budget checks run before every live attempt",
    "orchestration": [
      "cheap models are sorted first inside each runnable family",
      "expensive or unknown-cost attempts are deferred when they would exceed the budget cap",
      "blocked adapter families remain visible instead of being dropped from the denominator"
    ],
    "observability": [
      "state.json",
      "queue.jsonl",
      "events.jsonl",
      "dashboard.json",
      "budget-ledger.json",
      "adapter-gaps.json"
    ],
    "memory": [
      "session memory is local-only under .proofloop/prod-proxy-longrun",
      "durable lessons must be promoted separately; generated memory stores are not committed"
    ]
  },
  "summary": {
    "totalAttempts": 5416,
    "passedExistingAttempts": 10,
    "queuedAttempts": 3516,
    "runningAttempts": 0,
    "passedAttempts": 0,
    "failedAttempts": 0,
    "blockedAdapterAttempts": 0,
    "blockedBudgetAttempts": 1890,
    "uniqueTaskTargets": 1354,
    "modelCount": 4,
    "families": 10,
    "familiesWithQueuedAttempts": 9,
    "familiesBlockedByAdapter": 0,
    "currentAllTaskWinner": null,
    "currentAdapterSmokeWinner": "poolside/laguna-xs-2.1"
  },
  "modelCosts": [
    {
      "modelId": "z-ai/glm-5.2",
      "prodAdapterSmokePassed": 3,
      "prodAdapterSmokeTotal": 3,
      "costBasis": "measured_and_estimated_smoke",
      "estimatedCostPerAttemptUsd": 0.081806,
      "fullMatrixEstimatedUsd": 110.765324,
      "runnableQueueEstimatedUsd": 0,
      "historicalMeasuredSpendUsd": 0.245
    },
    {
      "modelId": "deepseek/deepseek-v4-flash",
      "prodAdapterSmokePassed": 1,
      "prodAdapterSmokeTotal": 3,
      "costBasis": "measured_and_estimated_smoke",
      "estimatedCostPerAttemptUsd": 0.023,
      "fullMatrixEstimatedUsd": 31.142,
      "runnableQueueEstimatedUsd": 31.119,
      "historicalMeasuredSpendUsd": 0.025
    },
    {
      "modelId": "poolside/laguna-xs-2.1",
      "prodAdapterSmokePassed": 3,
      "prodAdapterSmokeTotal": 3,
      "costBasis": "measured_and_estimated_smoke",
      "estimatedCostPerAttemptUsd": 0.011,
      "fullMatrixEstimatedUsd": 14.894,
      "runnableQueueEstimatedUsd": 14.861,
      "historicalMeasuredSpendUsd": 0.033
    },
    {
      "modelId": "qwen/qwen3.7-plus",
      "prodAdapterSmokePassed": 3,
      "prodAdapterSmokeTotal": 3,
      "costBasis": "measured_and_estimated_smoke",
      "estimatedCostPerAttemptUsd": 0.066478,
      "fullMatrixEstimatedUsd": 90.011212,
      "runnableQueueEstimatedUsd": 53.980136,
      "historicalMeasuredSpendUsd": 0.199
    }
  ],
  "adapterGaps": [],
  "attemptsOmitted": {
    "reason": "Tracked docs/eval JSON omits the full attempt queue; durable resumable state is written under .proofloop/prod-proxy-longrun/.",
    "localStatePath": ".proofloop/prod-proxy-longrun/prod-proxy-longrun-2026-07-05T08-11-53-015Z/state.json",
    "queuePath": ".proofloop/prod-proxy-longrun/prod-proxy-longrun-2026-07-05T08-11-53-015Z/queue.jsonl",
    "totalAttempts": 5416,
    "sample": [
      {
        "schema": "proofloop-prod-proxy-longrun-attempt-v1",
        "attemptId": "spreadsheetbench-v1-full-912--102-20--deepseek_deepseek-v4-flash--b48daaac0e",
        "familyId": "spreadsheetbench-v1-full-912",
        "taskId": "102-20",
        "modelId": "deepseek/deepseek-v4-flash",
        "title": "I need guidance with reformatting my Excel raw data using VBA. Specifically, I want to combine and sum up data from a co",
        "status": "queued",
        "adapterStatus": "prod_browser_available",
        "realUserProdUiRequired": true,
        "memoryModeAllowed": false,
        "officialScoreClaim": false,
        "estimatedCostUsd": 0.023,
        "command": {
          "shell": "npm run proofloop:live:spreadsheetbench-v1",
          "env": {
            "BENCH_BASE_URL": "https://noderoom.live",
            "PLAYWRIGHT_BASE_URL": "https://noderoom.live",
            "PLAYWRIGHT_REUSE_SERVER": "1",
            "PLAYWRIGHT_RETRIES": "0",
            "PLAYWRIGHT_OUTPUT_DIR": ".proofloop/prod-proxy-longrun/prod-proxy-longrun-2026-07-05T08-11-53-015Z/receipts/playwright/spreadsheetbench-v1-full-912--102-20--deepseek_deepseek-v4-flash--b48daaac0e",
            "BENCH_AGENT_MODEL_MODE": "specific",
            "BENCH_AGENT_MODEL_POLICY": "deepseek/deepseek-v4-flash",
            "PROOFLOOP_REAL_USER_MODE": "1",
            "PROOFLOOP_FOCUS_MODE": "0",
            "PROOFLOOP_NODEAGENT_RUNTIME_PROFILE": "",
            "PROOFLOOP_RUN_ID": "prod-proxy-longrun-2026-07-05T08-11-53-015Z-spreadsheetbench-v1-full-912--102-20--deepseek_deepseek-v4-flash--b48daaac0e",
            "PROOFLOOP_CASE_ID": "spreadsheetbench-v1-full-912--102-20--deepseek_deepseek-v4-flash--b48daaac0e",
            "PROOFLOOP_FRESH_ROOM_ROOT": ".proofloop/prod-proxy-longrun/prod-proxy-longrun-2026-07-05T08-11-53-015Z/receipts/fresh-room",
            "PROOFLOOP_SUITE_PROOF_PATH": ".proofloop/prod-proxy-longrun/prod-proxy-longrun-2026-07-05T08-11-53-015Z/receipts/spreadsheetbench-v1-full-912--102-20--deepseek_deepseek-v4-flash--b48daaac0e.json",
            "SPREADSHEETBENCH_TRACK": "spreadsheetbench-v1",
            "SPREADSHEETBENCH_STAGE_ROOT": ".tmp/official-benchmarks/staged-v1-912",
            "SPREADSHEETBENCH_TASK_ID": "102-20",
            "SPREADSHEETBENCH_LIVE_PROOF_PATH": ".proofloop/prod-proxy-longrun/prod-proxy-longrun-2026-07-05T08-11-53-015Z/receipts/spreadsheetbench-v1-full-912--102-20--deepseek_deepseek-v4-flash--b48daaac0e.json"
          }
        },
        "evidence": [
          "docs/eval/official-benchmark-task-coverage.json",
          "docs/eval/spreadsheetbench-v1-912-stage.json",
          "docs/eval/spreadsheetbench-v1-912-copy-input-baseline.json",
          "docs/eval/official-benchmark-readiness.json",
          ".tmp/official-benchmarks/staged-v1-912/tasks/102-20/agent/task.json",
          ".tmp/official-benchmarks/staged-v1-912/tasks/102-20/evaluator/evaluator.json",
          "docs/eval/spreadsheetbench-live-room-proof.json"
        ],
        "blockers": [
          "Existing SpreadsheetBench receipt is local live-browser only; rerun against https://noderoom.live."
        ],
        "attempts": 0
      },
      {
        "schema": "proofloop-prod-proxy-longrun-attempt-v1",
        "attemptId": "spreadsheetbench-v1-full-912--102-20--poolside_laguna-xs-2.1--33b65c6844",
        "familyId": "spreadsheetbench-v1-full-912",
        "taskId": "102-20",
        "modelId": "poolside/laguna-xs-2.1",
        "title": "I need guidance with reformatting my Excel raw data using VBA. Specifically, I want to combine and sum up data from a co",
        "status": "queued",
        "adapterStatus": "prod_browser_available",
        "realUserProdUiRequired": true,
        "memoryModeAllowed": false,
        "officialScoreClaim": false,
        "estimatedCostUsd": 0.011,
        "command": {
          "shell": "npm run proofloop:live:spreadsheetbench-v1",
          "env": {
            "BENCH_BASE_URL": "https://noderoom.live",
            "PLAYWRIGHT_BASE_URL": "https://noderoom.live",
            "PLAYWRIGHT_REUSE_SERVER": "1",
            "PLAYWRIGHT_RETRIES": "0",
            "PLAYWRIGHT_OUTPUT_DIR": ".proofloop/prod-proxy-longrun/prod-proxy-longrun-2026-07-05T08-11-53-015Z/receipts/playwright/spreadsheetbench-v1-full-912--102-20--poolside_laguna-xs-2.1--33b65c6844",
            "BENCH_AGENT_MODEL_MODE": "specific",
            "BENCH_AGENT_MODEL_POLICY": "poolside/laguna-xs-2.1",
            "PROOFLOOP_REAL_USER_MODE": "1",
            "PROOFLOOP_FOCUS_MODE": "0",
            "PROOFLOOP_NODEAGENT_RUNTIME_PROFILE": "",
            "PROOFLOOP_RUN_ID": "prod-proxy-longrun-2026-07-05T08-11-53-015Z-spreadsheetbench-v1-full-912--102-20--poolside_laguna-xs-2.1--33b65c6844",
            "PROOFLOOP_CASE_ID": "spreadsheetbench-v1-full-912--102-20--poolside_laguna-xs-2.1--33b65c6844",
            "PROOFLOOP_FRESH_ROOM_ROOT": ".proofloop/prod-proxy-longrun/prod-proxy-longrun-2026-07-05T08-11-53-015Z/receipts/fresh-room",
            "PROOFLOOP_SUITE_PROOF_PATH": ".proofloop/prod-proxy-longrun/prod-proxy-longrun-2026-07-05T08-11-53-015Z/receipts/spreadsheetbench-v1-full-912--102-20--poolside_laguna-xs-2.1--33b65c6844.json",
            "SPREADSHEETBENCH_TRACK": "spreadsheetbench-v1",
            "SPREADSHEETBENCH_STAGE_ROOT": ".tmp/official-benchmarks/staged-v1-912",
            "SPREADSHEETBENCH_TASK_ID": "102-20",
            "SPREADSHEETBENCH_LIVE_PROOF_PATH": ".proofloop/prod-proxy-longrun/prod-proxy-longrun-2026-07-05T08-11-53-015Z/receipts/spreadsheetbench-v1-full-912--102-20--poolside_laguna-xs-2.1--33b65c6844.json"
          }
        },
        "evidence": [
          "docs/eval/official-benchmark-task-coverage.json",
          "docs/eval/spreadsheetbench-v1-912-stage.json",
          "docs/eval/spreadsheetbench-v1-912-copy-input-baseline.json",
          "docs/eval/official-benchmark-readiness.json",
          ".tmp/official-benchmarks/staged-v1-912/tasks/102-20/agent/task.json",
          ".tmp/official-benchmarks/staged-v1-912/tasks/102-20/evaluator/evaluator.json",
          "docs/eval/spreadsheetbench-live-room-proof.json"
        ],
        "blockers": [
          "Existing SpreadsheetBench receipt is local live-browser only; rerun against https://noderoom.live."
        ],
        "attempts": 0
      },
      {
        "schema": "proofloop-prod-proxy-longrun-attempt-v1",
        "attemptId": "finch-prod-proxy-task--finch-local-financial-evidence-qa--z-ai_glm-5.2--163633ef8d",
        "familyId": "finch-prod-proxy-task",
        "taskId": "finch-local-financial-evidence-qa",
        "modelId": "z-ai/glm-5.2",
        "title": "Financial evidence QA through NodeRoom story workflow",
        "status": "passed_existing",
        "adapterStatus": "prod_browser_available",
        "realUserProdUiRequired": true,
        "memoryModeAllowed": false,
        "officialScoreClaim": false,
        "estimatedCostUsd": 0.081806,
        "measuredCostUsd": 0.021,
        "evidence": [
          "proofloop/benchmarks/finch/adapter.json",
          "docs/eval/proofloop-external-adapter-live-room-runs/finch.json",
          "docs/eval/proofloop-proxy-model-sweep.json",
          "docs/eval/proofloop-proxy-model-sweep-runs/proxy-model-sweep-2026-07-04-live/z-ai_glm-5.2/finch.json",
          "https://noderoom.live/?room=NR1RHOT8YL9&name=Proof+Loop"
        ],
        "blockers": [],
        "attempts": 0
      },
      {
        "schema": "proofloop-prod-proxy-longrun-attempt-v1",
        "attemptId": "finch-prod-proxy-task--finch-local-financial-evidence-qa--poolside_laguna-xs-2.1--7b96cd3899",
        "familyId": "finch-prod-proxy-task",
        "taskId": "finch-local-financial-evidence-qa",
        "modelId": "poolside/laguna-xs-2.1",
        "title": "Financial evidence QA through NodeRoom story workflow",
        "status": "passed_existing",
        "adapterStatus": "prod_browser_available",
        "realUserProdUiRequired": true,
        "memoryModeAllowed": false,
        "officialScoreClaim": false,
        "estimatedCostUsd": 0.011,
        "measuredCostUsd": 0.012,
        "evidence": [
          "proofloop/benchmarks/finch/adapter.json",
          "docs/eval/proofloop-external-adapter-live-room-runs/finch.json",
          "docs/eval/proofloop-proxy-model-sweep.json",
          "docs/eval/proofloop-proxy-model-sweep-runs/proxy-model-sweep-2026-07-04-live/poolside_laguna-xs-2.1/finch.json",
          "https://noderoom.live/?room=NRTSTABCXSW&name=Proof+Loop"
        ],
        "blockers": [],
        "attempts": 0
      }
    ]
  }
}