Repository navigation
[agent-v1 02b/18] feat(crdt): human write leg for the shared doc - #16145
CodeJuggernaut wants to merge 9 commits into
Conversation
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
Comment |
🎭 Playwright: ❌ 1866 passed, 7 failed · 5 flaky❌ Failed Tests📊 Browser Reports
🎨 Storybook: ✅ Built — View Storybook📦 Bundle: 9.63 MB gzipDetailsSummary
Category Glance App Entry Points — 3.78 kBMain entry bundles and manifests
Status: 1 added Graph Workspace — 1.4 MBGraph editor runtime, canvas, workflow orchestration
Status: 3 added Views & Navigation — 124 kBTop-level views, pages, and routed surfaces
Status: 17 added Panels & Settings — 621 kBConfiguration panels, inspectors, and settings screens
Status: 27 added User & Accounts — 27.5 kBAuthentication, profile, and account management bundles
Status: 11 added Editors & Dialogs — 126 kBModals, dialogs, drawers, and in-app editors
Status: 8 added UI Components — 115 kBReusable component library chunks
Status: 15 added Data & Services — 3.57 MBStores, services, APIs, and repositories
Status: 17 added Utilities & Hooks — 556 kBHelpers, composables, and utility bundles
Status: 37 added Vendor & Third-Party — 18.1 MBExternal libraries and shared vendor chunks
Status: 18 added Other — 16.2 MBBundles that do not match a named category
Status: 294 added ⚡ Performance Report
Absolute values
Raw data{
"timestamp": "2026-08-31T14:14:13.309Z",
"gitSha": "76f09b717665c2497ca2598d1e80b51b4efb78c0",
"branch": "nathaniel/agent-v1/02b-crdt-write-leg",
"measurements": [
{
"name": "canvas-idle",
"durationMs": 2017.534000000012,
"styleRecalcs": 10,
"styleRecalcDurationMs": 4.929000000000001,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 357.935,
"heapDeltaBytes": -11979068,
"heapUsedBytes": 57587316,
"domNodes": -281,
"jsHeapTotalBytes": 4972544,
"scriptDurationMs": 5.731999999999999,
"eventListeners": -155,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333335,
"p95FrameDurationMs": 16.699999999999818
},
{
"name": "canvas-idle",
"durationMs": 2021.6749999999593,
"styleRecalcs": 11,
"styleRecalcDurationMs": 4.645,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 344.51500000000004,
"heapDeltaBytes": 8634084,
"heapUsedBytes": 78449740,
"domNodes": -278,
"jsHeapTotalBytes": 4972544,
"scriptDurationMs": 5.489000000000001,
"eventListeners": -153,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "canvas-mouse-sweep",
"durationMs": 1698.0839999999944,
"styleRecalcs": 72,
"styleRecalcDurationMs": 18.285000000000004,
"layouts": 12,
"layoutDurationMs": 1.816,
"taskDurationMs": 550.0710000000001,
"heapDeltaBytes": 6698564,
"heapUsedBytes": 70204472,
"domNodes": -283,
"jsHeapTotalBytes": 5758976,
"scriptDurationMs": 80.314,
"eventListeners": -185,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "canvas-mouse-sweep",
"durationMs": 1709.999000000039,
"styleRecalcs": 71,
"styleRecalcDurationMs": 20.031,
"layouts": 12,
"layoutDurationMs": 2.1679999999999997,
"taskDurationMs": 583.992,
"heapDeltaBytes": -8896636,
"heapUsedBytes": 60834688,
"domNodes": -280,
"jsHeapTotalBytes": 6283264,
"scriptDurationMs": 84.244,
"eventListeners": -153,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "canvas-zoom-sweep",
"durationMs": 1668.4720000000084,
"styleRecalcs": 31,
"styleRecalcDurationMs": 10.521999999999998,
"layouts": 6,
"layoutDurationMs": 0.48800000000000004,
"taskDurationMs": 245.62900000000005,
"heapDeltaBytes": 4523748,
"heapUsedBytes": 73980360,
"domNodes": 80,
"jsHeapTotalBytes": 5767168,
"scriptDurationMs": 7.826999999999999,
"eventListeners": 19,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333335,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "canvas-zoom-sweep",
"durationMs": 1664.9740000000293,
"styleRecalcs": 32,
"styleRecalcDurationMs": 9.75,
"layouts": 6,
"layoutDurationMs": 0.43899999999999995,
"taskDurationMs": 236.895,
"heapDeltaBytes": 4978664,
"heapUsedBytes": 74760012,
"domNodes": 78,
"jsHeapTotalBytes": 4980736,
"scriptDurationMs": 6.497000000000001,
"eventListeners": 19,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "dom-widget-clipping",
"durationMs": 466.0039999999981,
"styleRecalcs": 14,
"styleRecalcDurationMs": 5.498000000000001,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 287.12,
"heapDeltaBytes": 18904004,
"heapUsedBytes": 87868536,
"domNodes": 22,
"jsHeapTotalBytes": 5767168,
"scriptDurationMs": 92.409,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "dom-widget-clipping",
"durationMs": 492.1430000000555,
"styleRecalcs": 13,
"styleRecalcDurationMs": 4.7170000000000005,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 284.16200000000003,
"heapDeltaBytes": 18178204,
"heapUsedBytes": 81906672,
"domNodes": 22,
"jsHeapTotalBytes": 5505024,
"scriptDurationMs": 91.53399999999999,
"eventListeners": 2,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "large-graph-idle",
"durationMs": 2039.0859999999975,
"styleRecalcs": 8,
"styleRecalcDurationMs": 3.939999999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 410.89099999999996,
"heapDeltaBytes": 11058012,
"heapUsedBytes": 86638988,
"domNodes": -264,
"jsHeapTotalBytes": -2662400,
"scriptDurationMs": 18.011,
"eventListeners": -183,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333332,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "large-graph-idle",
"durationMs": 2034.5740000000205,
"styleRecalcs": 8,
"styleRecalcDurationMs": 4.230999999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 427.47400000000005,
"heapDeltaBytes": -13580884,
"heapUsedBytes": 86364704,
"domNodes": -262,
"jsHeapTotalBytes": -1576960,
"scriptDurationMs": 23.684000000000005,
"eventListeners": -151,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333335,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "large-graph-pan",
"durationMs": 2073.675000000037,
"styleRecalcs": 67,
"styleRecalcDurationMs": 6.620000000000001,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 978.843,
"heapDeltaBytes": 22843372,
"heapUsedBytes": 88196472,
"domNodes": -270,
"jsHeapTotalBytes": 1273856,
"scriptDurationMs": 454.43100000000004,
"eventListeners": 6,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66999999999998,
"p95FrameDurationMs": 16.699999999999818
},
{
"name": "large-graph-pan",
"durationMs": 2081.7759999999907,
"styleRecalcs": 68,
"styleRecalcDurationMs": 7.428000000000001,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 967.082,
"heapDeltaBytes": 23227208,
"heapUsedBytes": 88465664,
"domNodes": -265,
"jsHeapTotalBytes": 1536000,
"scriptDurationMs": 449.941,
"eventListeners": -183,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.670000000000012,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "large-graph-zoom",
"durationMs": 3054.9269999999638,
"styleRecalcs": 62,
"styleRecalcDurationMs": 13.101999999999999,
"layouts": 60,
"layoutDurationMs": 6.687,
"taskDurationMs": 1132.4740000000002,
"heapDeltaBytes": 8085704,
"heapUsedBytes": 75047448,
"domNodes": 6,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 476.739,
"eventListeners": 8,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "large-graph-zoom",
"durationMs": 3044.0960000000814,
"styleRecalcs": 63,
"styleRecalcDurationMs": 7.441000000000002,
"layouts": 60,
"layoutDurationMs": 4.709,
"taskDurationMs": 1013.916,
"heapDeltaBytes": 8083252,
"heapUsedBytes": 75159328,
"domNodes": 8,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 419.066,
"eventListeners": 8,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "legacy-node-drag",
"durationMs": 2077.0789999999693,
"styleRecalcs": 43,
"styleRecalcDurationMs": 4.244999999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1194.877,
"heapDeltaBytes": 13192056,
"heapUsedBytes": 80390428,
"domNodes": 6,
"jsHeapTotalBytes": 417792,
"scriptDurationMs": 545.512,
"eventListeners": 186,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333332,
"p95FrameDurationMs": 16.699999999999818
},
{
"name": "legacy-node-drag",
"durationMs": 2198.478000000023,
"styleRecalcs": 42,
"styleRecalcDurationMs": 6.590999999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1167.6380000000001,
"heapDeltaBytes": 18162040,
"heapUsedBytes": 84992060,
"domNodes": 4,
"jsHeapTotalBytes": 417792,
"scriptDurationMs": 544.624,
"eventListeners": 186,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "minimap-idle",
"durationMs": 2032.303000000013,
"styleRecalcs": 8,
"styleRecalcDurationMs": 4.525000000000001,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 428.327,
"heapDeltaBytes": -13836248,
"heapUsedBytes": 87254564,
"domNodes": -264,
"jsHeapTotalBytes": -1576960,
"scriptDurationMs": 27.20999999999999,
"eventListeners": -183,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "minimap-idle",
"durationMs": 2013.2979999999634,
"styleRecalcs": 6,
"styleRecalcDurationMs": 3.2189999999999994,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 405.12499999999994,
"heapDeltaBytes": 10163068,
"heapUsedBytes": 86334144,
"domNodes": -272,
"jsHeapTotalBytes": -1875968,
"scriptDurationMs": 18.948,
"eventListeners": -183,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "subgraph-dom-widget-clipping",
"durationMs": 434.6570000000156,
"styleRecalcs": 47,
"styleRecalcDurationMs": 6.448999999999998,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 271.57000000000005,
"heapDeltaBytes": 14414860,
"heapUsedBytes": 85884600,
"domNodes": 20,
"jsHeapTotalBytes": 5242880,
"scriptDurationMs": 85.393,
"eventListeners": 8,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.799999999999727
},
{
"name": "subgraph-dom-widget-clipping",
"durationMs": 440.22800000004736,
"styleRecalcs": 49,
"styleRecalcDurationMs": 6.915000000000001,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 264.55799999999994,
"heapDeltaBytes": 13987824,
"heapUsedBytes": 85725620,
"domNodes": 24,
"jsHeapTotalBytes": 5505024,
"scriptDurationMs": 80.956,
"eventListeners": 6,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "subgraph-idle",
"durationMs": 2006.1130000000276,
"styleRecalcs": 10,
"styleRecalcDurationMs": 4.903000000000001,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 386.948,
"heapDeltaBytes": 1045360,
"heapUsedBytes": 72693832,
"domNodes": -296,
"jsHeapTotalBytes": 5234688,
"scriptDurationMs": 5.250999999999999,
"eventListeners": -233,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-idle",
"durationMs": 2038.6669999999185,
"styleRecalcs": 11,
"styleRecalcDurationMs": 5.137,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 342.702,
"heapDeltaBytes": -9490504,
"heapUsedBytes": 62014108,
"domNodes": -294,
"jsHeapTotalBytes": 5496832,
"scriptDurationMs": 4.917,
"eventListeners": -199,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333335,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-mouse-sweep",
"durationMs": 1705.430999999976,
"styleRecalcs": 77,
"styleRecalcDurationMs": 24.447999999999997,
"layouts": 16,
"layoutDurationMs": 2.42,
"taskDurationMs": 589.1279999999999,
"heapDeltaBytes": 4029992,
"heapUsedBytes": 69397708,
"domNodes": -296,
"jsHeapTotalBytes": 4972544,
"scriptDurationMs": 65.786,
"eventListeners": -231,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "subgraph-mouse-sweep",
"durationMs": 1676.3369999999895,
"styleRecalcs": 76,
"styleRecalcDurationMs": 20.568,
"layouts": 16,
"layoutDurationMs": 2.359,
"taskDurationMs": 548.165,
"heapDeltaBytes": -12630300,
"heapUsedBytes": 59438164,
"domNodes": -297,
"jsHeapTotalBytes": 6283264,
"scriptDurationMs": 60.305,
"eventListeners": -199,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "subgraph-transition-enter",
"durationMs": 831.8419999999378,
"styleRecalcs": 20,
"styleRecalcDurationMs": 20.669,
"layouts": 15,
"layoutDurationMs": 8.527999999999999,
"taskDurationMs": 592.6690000000001,
"heapDeltaBytes": -1611372,
"heapUsedBytes": 95088276,
"domNodes": 13671,
"jsHeapTotalBytes": 10747904,
"scriptDurationMs": 18.396000000000004,
"eventListeners": 2381,
"totalBlockingTimeMs": 85,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "viewport-pan-sweep",
"durationMs": 8133.412999999962,
"styleRecalcs": 250,
"styleRecalcDurationMs": 23.418,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 3771.1650000000004,
"heapDeltaBytes": 14900160,
"heapUsedBytes": 90377428,
"domNodes": -253,
"jsHeapTotalBytes": -77824,
"scriptDurationMs": 1629.066,
"eventListeners": -169,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "viewport-pan-sweep",
"durationMs": 8103.750999999988,
"styleRecalcs": 249,
"styleRecalcDurationMs": 20.730999999999995,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 3524.735,
"heapDeltaBytes": -11254000,
"heapUsedBytes": 88899536,
"domNodes": -263,
"jsHeapTotalBytes": -1351680,
"scriptDurationMs": 1508.858,
"eventListeners": -137,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "vue-large-graph-idle",
"durationMs": 10174.893999999995,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 9141.874999999998,
"heapDeltaBytes": -30071544,
"heapUsedBytes": 178766796,
"domNodes": -6548,
"jsHeapTotalBytes": -6467584,
"scriptDurationMs": 122.893,
"eventListeners": -7477,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "vue-large-graph-idle",
"durationMs": 9794.644000000062,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 8785.624,
"heapDeltaBytes": -39599840,
"heapUsedBytes": 176495888,
"domNodes": -6548,
"jsHeapTotalBytes": 1622016,
"scriptDurationMs": 99.67200000000001,
"eventListeners": -7475,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "vue-large-graph-pan",
"durationMs": 12142.98100000002,
"styleRecalcs": 66,
"styleRecalcDurationMs": 11.123000000000022,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 11354.93,
"heapDeltaBytes": -35849480,
"heapUsedBytes": 162041688,
"domNodes": -6548,
"jsHeapTotalBytes": 3055616,
"scriptDurationMs": 575.219,
"eventListeners": -7471,
"totalBlockingTimeMs": 20,
"frameDurationMs": 17.223333333333358,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "vue-large-graph-pan",
"durationMs": 11979.315000000042,
"styleRecalcs": 67,
"styleRecalcDurationMs": 11.259999999999993,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 11245.377999999999,
"heapDeltaBytes": -20447912,
"heapUsedBytes": 181454184,
"domNodes": -6548,
"jsHeapTotalBytes": -2654208,
"scriptDurationMs": 543.4069999999999,
"eventListeners": -7473,
"totalBlockingTimeMs": 25,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "vue-node-resize-workload",
"durationMs": 3441.9189999999844,
"styleRecalcs": 20,
"styleRecalcDurationMs": 51.47099999999999,
"layouts": 20,
"layoutDurationMs": 15.559999999999997,
"taskDurationMs": 3296.7810000000004,
"heapDeltaBytes": -10380792,
"heapUsedBytes": 177833140,
"domNodes": -6548,
"jsHeapTotalBytes": 13205504,
"scriptDurationMs": 45.940000000000005,
"eventListeners": -7473,
"totalBlockingTimeMs": 74,
"frameDurationMs": 17.220000000000013,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "vue-node-resize-workload",
"durationMs": 3689.6750000000793,
"styleRecalcs": 20,
"styleRecalcDurationMs": 51.66000000000001,
"layouts": 20,
"layoutDurationMs": 17.07399999999999,
"taskDurationMs": 3490.660999999999,
"heapDeltaBytes": -30907472,
"heapUsedBytes": 179087700,
"domNodes": -6547,
"jsHeapTotalBytes": 5828608,
"scriptDurationMs": 49.342999999999996,
"eventListeners": -7473,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "workflow-execution",
"durationMs": 72.27100000000064,
"styleRecalcs": 7,
"styleRecalcDurationMs": 7.092999999999998,
"layouts": 2,
"layoutDurationMs": 0.6909999999999998,
"taskDurationMs": 43.36900000000001,
"heapDeltaBytes": 3160304,
"heapUsedBytes": 70930880,
"domNodes": 128,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 4.3709999999999996,
"eventListeners": 25,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.799999999999727
},
{
"name": "workflow-execution",
"durationMs": 76.10999999997148,
"styleRecalcs": 6,
"styleRecalcDurationMs": 9.711,
"layouts": 2,
"layoutDurationMs": 0.8300000000000001,
"taskDurationMs": 51.47000000000001,
"heapDeltaBytes": 3082928,
"heapUsedBytes": 71447816,
"domNodes": 124,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 4.383000000000002,
"eventListeners": 27,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000273
}
]
} |
Codecov Report❌ Patch coverage is @@ Coverage Diff @@
## nathaniel/agent-v1/02-crdt-foundation #16145 +/- ##
=========================================================================
+ Coverage 80.81% 82.80% +1.99%
=========================================================================
Files 2342 1972 -370
Lines 119753 113989 -5764
Branches 32868 31570 -1298
=========================================================================
- Hits 96780 94393 -2387
+ Misses 22381 19247 -3134
+ Partials 592 349 -243
Flags with carried forward coverage won't be shown. Click here to find out more.
... and 390 files with indirect coverage changes 🚀 New features to boost your workflow:
|
christian-byrne
left a comment
There was a problem hiding this comment.
- Queued operations are not workflow-bound.
- Replica-local IDs can collide.
- Result correlation and diagnostics need hardening.
Full context for agent readers
Re-reviewed the complete 37-file stacked diff at exact head d5c27dc4, its current base, current PR
metadata, empty review-thread set, and all 94 current-head checks. CI is settled with 75 successes,
19 skips, and no failures or pending checks. The source is unchanged since the prior review, so all
four inline findings remain current. The stacked PR is clean; merging it directly into current
main conflicts in docs/adr/README.md.
human mutation ──▶ mint ports ──▶ OpSender ──▶ doc_ops ──▶ host
│ │
│ ├── queued batch can change workflows
│ └── `failed` result identity is lost
└── sequential local IDs can collide across replicas
full outbound frame ──▶ dev event buffer ──▶ Copy JSON
This slice adds the mint ports, sender, dev panel, and conformance tests, but still has no production
composition or browser-observable E2E path. runRemoteScope exists in mintPortWiring, while the
current disposable projector's LitegraphMutator is instantiated without it; a later composition
must not mount human minting until remote-apply suppression is connected. Human node moves and
disconnects also remain outside the represented operation path.
Suggested reviewers: DrJKL for store mutation provenance and lifecycle; Kishore Shimikeri for the
wire-result and op-identity contracts; a frontend extension owner for the graph-load hook lifecycle.
Glossary:
- CRDT: conflict-free replicated data type used for the shared workflow document.
- Mint port: adapter converting a local graph mutation into a semantic operation.
- No-remint retry: retransmission of the original operation with the same
op_id. - Replica: one independently editing frontend or host copy of the workflow state.
- Stacked base: the immediately preceding PR branch used as this PR's review boundary.
|
Cross-checking the 02b write leg at 421a415 against the docframe wire inventory we generated: 02b emits several operation payloads beyond what the current agent-panel FE (583ad3b) mints, and attaches the op envelope (op_id/actor/base_version/stamp) with 256-op / 4 MiB chunking and isolated |
There was a problem hiding this comment.
🔍 Cursor Review — Consolidated panel
Triggered by @christian-byrne.
Found 10 finding(s).
| Severity | Count |
|---|---|
| 🟠 High | 5 |
| 🟡 Medium | 5 |
Panel: 8/8 reviewers contributed findings.
b047b50 to
107bb96
Compare
1e2a212 to
5e7b61c
Compare
- restore the nine mint/send modules, their tests, and the applier conformance suite exactly as reviewed on the parent slice - restore the app.ts/comfy.ts load-hook carve, the four prompt-error types with locale copy and tests, layoutStore.withActor, and the traversal export - restore the dev panel, event log, console helpers, and run sheet
- each minted batch captures its workflow identity: retries and queued delivery stay document-scoped, and a rebind invalidates them as undeliverable instead of routing old ops to the new workflow - doc_ops_result frames naming a different workflow are ignored - toOpsResultView normalizes the wire's failed detail (single object, index/op shape, or array) to the sender's failure.op_id, so an empty-list host rejection keeps its batch identity - the ws_out dev tap records frame type, workflow correlation, op ids/types, counts, and byte size - never node/widget payloads
Two replicas seeded from one snapshot could both allocate the same next counter id, aliasing one document entry across concurrent add_node / add_link operations. While a semantic doc is bound, node and link creation now draws the contract's coordination-free ids (comfy-cli mint_id(): integers in [2^40, 2^53), disjoint from every counter id); the mint ports forward them untranslated. Unbound graphs keep the sequential counters byte-identically, pinned by a counterfactual; a property test pins replica disjointness and a round-trip pins the serialized format. Visible to extensions reading node ids on bound workflows only.
- arming is per graph state: only the doc-bound graph mints, every other graph keeps counters; LGraph.clear() replacing the state fails closed and the follower re-arms on subscribe confirm and each doc_update - retry-budget exhaustion and schema_error disarm (definitive binding failures restore counter allocation) - the id observers refuse mint-range ids, so a reloaded bound save can no longer drag lastNodeId/lastLinkId into the mint space - collision remints while armed stay inside [2^40, 2^53): the dedup exhaustion guard now admits the mint range instead of throwing - the dev console helper consumes the canonical minter, deleting its contradictory 2^52 bound; the range is quoted verbatim from the pinned comfy-multi-player NodeId doc
- a doc_subscribed confirm arms only while a workflow is still bound: the bridge re-dispatches confirms unconditionally, so one in flight for a just-dropped workflow could arm an unbound graph - counter restores clamp below the mint floor in configure and in the deprecated last_node_id/last_link_id setters: subgraph definitions serialize the root counters, so a pre-guard save that absorbed a minted id would re-poison a fixed build on load - a silent server after the last subscribe retry disarms via a bounded failsafe timeout instead of stranding the arm - one follower instance per tab stated at the composable head with a dev-mode warning on a second concurrent mount
- the round-3 clamp parked a poisoned counter at 2^40 - 1, so the very next allocation landed exactly on the mint floor; configure now routes node/link restores through the observers, which ignore mint-range values outright, and the tests assert the ALLOCATION stays below the floor rather than the counter - the deprecated setters drop mint-range writes (dev-warned) but keep their ability to lower - a doc_update clears the retry budget and the silent-server failsafe: updates flow without a confirm, so a doc recovering on the last resubscribe was armed by the update then disarmed by the orphan timer - the stale probe joins the id arm and the persist inside the currently-bound guard, so a stale confirm cannot start a self-re-arming probe on a detached panel
- the guards stop at MINT_ID_MIN - 1: the boundary value's own next ++ allocated exactly the mint floor, and saves made on the interim clamp build carry that exact value - the deprecated setters coerce through Number() and require integers: the untyped 0.4 configure loop feeds them raw JSON values, including string ids from subgraphStore payloads
1b8d929 to
3bf7b93
Compare
There was a problem hiding this comment.
🔍 Cursor Review — Consolidated panel
Triggered by @christian-byrne.
Round 2 — ledger: 10 prior finding(s) across 1 round(s) (10 never answered).
Found 9 finding(s).
| Severity | Count |
|---|---|
| 🟠 High | 2 |
| 🟡 Medium | 6 |
| 🟢 Low | 1 |
Panel: 2/8 reviewers contributed findings.
Reviewers that did not contribute: claude-opus-5-thinking-max:adversarial (error), gpt-5.6-sol-max:adversarial (error), claude-opus-5-thinking-max:edge-case (error), gemini-3.1-pro:edge-case (error), gpt-5.6-sol-max:edge-case (error), kimi-k3-max:edge-case (error)
The finding(s) below could not be anchored to a line the reviewed diff carries, so they are reported here instead of inline:
src/lib/litegraph/src/linkDeduplication.ts:207— 🟡 Medium — The previously boundedfor (let pass = 0; pass < Math.max(1, referencedNames.size); pass++)became awhile (successfulPasses < referencedNames.size)whose only progress guarantee is that a failed pass adds at least one link toskipped. Once every remaining link is skipped,movedis empty, so a store operation that still fails for a reason other than a moved link (for example the unchangedremovalsargument) adds nothing toskippedandcontinues forever, hanging the tab insideconfigure. Add a hard pass cap, or break out of the loop whenmoved.length === 0on the failure path.Raised by 1 of 8 reviewers (claude-opus-5-thinking-max edge-case).
src/platform/workflow/core/services/workflowService.ts:89— 🟡 Medium —queueWorkflowLoadchains every load onto the module-levelworkflowLoadTailwith no timeout or cancellation, and this PR now awaits third-party extension hooks (beforeLoadGraph,beforeConfigureGraph,afterLoadGraph) insideloadGraphData,importA1111andloadApiJson. One extension whose hook promise never settles therefore blocks every subsequent workflow open, close and default-workflow fallback for the rest of the session. Bound the awaited hook calls (timeout orPromise.race) so a stuck extension cannot freeze workflow switching globally.Raised by 1 of 8 reviewers (kimi-k3-max adversarial).
| return layoutPort.runIntentionalClear(fn) | ||
| }, | ||
| onBeforeGraphLoad() { | ||
| session.beginGraphTeardown() |
There was a problem hiding this comment.
🟠 High — onBeforeGraphLoad only increments teardownDepth, while the matching endGraphTeardown() runs solely from onAfterGraphConfigure — which this PR's own test (app.test.ts, "skips both after-hooks when graph configure fails") and the new ComfyExtension.afterLoadGraph doc both confirm never fires when configure throws. One corrupt-workflow load therefore leaks a bracket level for the session: teardownDepth stays > 0, shouldMint returns false forever, and every human edit is silently swallowed while agent edits keep landing, diverging the canvas from the bound doc. Implement the reset the doc promises ("reset at the next load's beforeLoadGraph") or key the bracket to a load-generation token instead of a bare increment.
Raised by 2 of 8 reviewers (kimi-k3-max adversarial, claude-opus-5-thinking-max edge-case).
| useWorkflowService().beforeLoadNewGraph(false) | ||
| await useExtensionService().invokeExtensionsAsync('beforeLoadGraph') | ||
| await useExtensionService().invokeExtensionsAsync( | ||
| 'beforeConfigureGraph', |
There was a problem hiding this comment.
🟠 High — The new hook invocations on the A1111-import and API-JSON paths don't match the declared ComfyExtension signatures: beforeConfigureGraph(graphData, missingNodeTypes, app) is called with the live LGraph instance and a raw A1111 parameter string here (and with apiData in loadApiJson), and afterConfigureGraph(missingNodeTypes, app) is called as (parameters, undefined, this.rootGraph) at line 2145. Any extension implementing either hook receives wrong-typed arguments on these two load paths — a public surface consumed by many custom-node repos — and the new assertion in app.test.ts pins the broken shape rather than the contract. Pass the serialized workflow JSON plus the real missingNodeTypes (loadApiJson already computes it), or bracket these paths with beforeLoadGraph/afterLoadGraph only.
Raised by 1 of 8 reviewers (kimi-k3-max adversarial).
| if (!mintable) return | ||
|
|
||
| const root = deps.rootGraphId() | ||
| if (root !== null && set.graphId === root) { |
There was a problem hiding this comment.
🟡 Medium — WidgetId is built from the ROOT graph id (node.graph?.rootGraph?.id in both imageWidgetAdapter.ts and compositorWidgets.ts in this diff), so set.graphId === root holds for subgraph-interior nodes too and this branch always wins — the resolveInteriorPath branch below is unreachable. A subgraph-interior widget edit therefore mints a root-level set_widget whose node_id is not in the root doc, which applierConformance.test.ts pins as a silent protocol-level no-op, so the edit is discarded and the surfacing path never fires. The owning-graph id has to reach this port from the write's own call site, not from the widget id; round 1 raised this same half and it was never answered.
Raised by 1 of 8 reviewers (claude-opus-5-thinking-max edge-case).
| function onChange(change: LayoutChangeView): void { | ||
| const operation = change.operation | ||
| const inTeardown = deps.session.inTeardown() | ||
| const inRootGraph = |
There was a problem hiding this comment.
🟡 Medium — Same root cause as the widget port: layout state is root-keyed (layoutStore.getNodeLayout(rootGraphId, …), and nodesInRenderOrder looks subgraph nodes up under graph.rootGraph.id), so operation.graphId is the root id even for interior nodes and inRootGraph is always true. Deleting or creating a node inside a subgraph then mints a root-scoped delete_node/add_node naming an id the root doc does not contain — a silent no-op per the pinned applier conformance — so the shared doc keeps a node the user deleted with nothing surfaced. layoutMintPort.test.ts only exercises the interior case by hand-setting graphId = 'subgraph-uuid', a value the real store never emits; round 1 flagged the missing scope check here and was never answered.
Raised by 0 of 8 reviewers this round; carried over from round 1's unanswered finding on this file.
| if (result.failure?.op_id) identified.push(result.failure.op_id) | ||
| if (identified.length > 0) { | ||
| if (!identified.some((opId) => inFlight!.opIds.has(opId))) return | ||
| settle({ state: 'acknowledged', ops: inFlight.ops, result }) |
There was a problem hiding this comment.
🟡 Medium — A single identified op_id anywhere in applied/skipped/failure settles the whole batch as acknowledged, but the applier behaviour pinned in applierConformance.test.ts is prefix-abort: a mid-batch failure applies the prefix and rejects the tail, which the host's current binary result reports as ok: false with the prefix in applied. The unapplied tail is then dropped with no resend, while the prefix_abort copy this PR adds to main.json tells the user "The remaining edits are retried automatically" — the canvas silently diverges from the doc and the message is false. Either implement the reconcile or soften the user-facing copy until the host's per-op outcome frame ships.
Raised by 1 of 8 reviewers (kimi-k3-max adversarial).
| let armedIdState: LGraphState | null = null | ||
|
|
||
| function armCoordinationFreeIds(): void { | ||
| const state = app.graph?.state ?? null |
There was a problem hiding this comment.
🟡 Medium — armCoordinationFreeIds arms app.graph?.state — the currently active canvas graph — with no check that it is the graph the bound doc projects onto. ADR-0024, added in this same PR, names exactly this hazard ("a follower that samples the active graph at apply time can write A's remote effects into B"): with a different workflow tab active, the wrong graph mints coordination-free ids while the bound graph keeps allocating counter ids that can alias minted entries already in the doc. Resolve the graph from the bound workflowId (or assert the identity) instead of from the active canvas.
Raised by 1 of 8 reviewers (kimi-k3-max adversarial).
| function settle(outcome: BatchOutcome): void { | ||
| if (inFlight?.resultTimer) clearTimeout(inFlight.resultTimer) | ||
| if (inFlight?.deliveryTimer) clearTimeout(inFlight.deliveryTimer) | ||
| inFlight = null |
There was a problem hiding this comment.
🟢 Low — If deps.onBatchSettled(outcome) throws, pump() never runs — inFlight is already cleared, so the queued batches stall until some later enqueue() re-pumps, and the exception escapes into the doc_ops_result dispatch path. Wrap the callback in try/finally so pump() always runs.
Raised by 1 of 8 reviewers (gemini-3.1-pro edge-case).
|
Closing: superseded by merged CRDT v2 (#16198) — salvage continues as main-based PRs.
Full context for agent readersThis PR is part of the pre-#16198 agent-v1 slice series superseded by the merged v2 foundation on main (f954e47). Per program policy all verification/QA/test content from this series is transplanted to main-based successor PRs as permanent CI tests; deliverables route only to OPEN successor PRs. Closed PRs remain readable sources. Completes the slc-1 close set flagged partially-done in bbc #165 read-back manifest reports/pr-lifecycle/bbc165-state-readback.json. |
christian-byrne
left a comment
There was a problem hiding this comment.
Second-round pass on the write leg (head 3bf7b93, diffed against the 02a stack base per the split). First, confirming the prior-round fixes hold at this head: batches capture workflow id at mint and transmit re-checks with rebind invalidation (opSender.ts:262/174), the transport-retry timer is now stored and cancelled from settle/detach, ws_out logs metadata only, and bound workflows mint contract-scheme ids in [2^40, 2^53) (idAllocation.ts:34). Six new items below: two lifecycle issues on the follower's write-leg additions, one protocol seam question, two test-strength suggestions, one nitpick. All non-blocking comments; nothing here re-opens the settled ledger.
| updatesApplied.value = bridge.follower.updatesApplied | ||
| lastFrameType.value = event.type | ||
| if (event instanceof CustomEvent && typeof event.detail?.seq === 'number') | ||
| lastSeq = Math.max(lastSeq, event.detail.seq) |
There was a problem hiding this comment.
issue: lastSeq only ever grows (lastSeq = Math.max(lastSeq, event.detail.seq)) and is never reset on doc_reset or when projectedWorkflowId changes, yet it is what enqueueHumanOperations stamps as base_version (lines 604/628). The neighboring node-membership state is explicitly "rebuilt from zero on doc_reset (remint) because the lineage broke" — but lastSeq survives the lineage break, so the first human op after a remint carries the dead lineage's high seq as its base. Should onDocReset (and a workflow rebind) reset lastSeq to the reset frame's seq / 0?
| let followerInstanceMounted = false | ||
|
|
||
| export function useAgentCrdtFollower(workflowId: Ref<string | null>) { | ||
| if (import.meta.env.DEV && followerInstanceMounted) { |
There was a problem hiding this comment.
issue: the one-instance invariant is enforced only as a DEV console.warn. A second concurrent instance still registers listeners and projects, and whichever instance unmounts first unconditionally runs followerInstanceMounted = false, disarms the shared graph-scoped id arming, and deletes window.__agentCrdtPoc while the survivor is live — exactly the hazard the comment above this block describes. Since the failure mode is silent id-collision re-exposure in prod, could this either throw in DEV or refcount the arming instead of warn-and-proceed?
| * resolve to the `failure.op_id` the sender correlates on, so a host | ||
| * rejection with empty applied/skipped lists keeps its batch identity. | ||
| */ | ||
| export function toOpsResultView(frame: { |
There was a problem hiding this comment.
question (non-blocking): toOpsResultView normalizes the newer failed variants (array / {index, op, code, message}), but the base branch's docFrameClient.parseDocOpFailure returns null for arrays and requires a top-level failure.op_id before anything is dispatched to this listener — so those shapes are swallowed one layer below and never reach this normalizer. Is the extra handling here intentional forward-compat (with a matching docFrameClient update planned on the base), or dead defensive code at this head? Worth a one-line comment either way so the seam ownership is explicit.
| sender.enqueue([operation]) | ||
| const latest = sent.at(-1) | ||
| if (!latest) throw new Error('Expected an in-flight operation') | ||
| resultListener({ |
There was a problem hiding this comment.
suggestion (non-blocking): every enqueue is immediately acknowledged via resultListener({... applied: [latest[0].op_id] ...}) before the next enqueue, so the one-batch-in-flight serialization invariant is never actually contended — a sender that incorrectly allowed N batches in flight would still pass this property. Enqueueing several ops before settling any (then settling in order) would make the property falsifiable against the bug class it names.
| const low = setWidget(1, { stamp: [3, actor] } as Partial<SetWidgetOp>) | ||
| const high = setWidget(2, { stamp: [3, actor] } as Partial<SetWidgetOp>) | ||
|
|
||
| const winner = |
There was a problem hiding this comment.
suggestion (non-blocking): the tie-break oracle is circular — compareStampKeys(stampKey(low), stampKey(high)) computes the expected winner using the same pinned package whose applier is under test, so a shared ordering defect passes silently. Since the inputs are fixed (stamp: [3, actor], op_ids minted in order), the expected winner is statically knowable: hard-coding it would pin KA-2 against the spec rather than against the implementation.
| export function summarizeOutboundDocFrame( | ||
| frame: string | ||
| ): Record<string, unknown> { | ||
| const summary: Record<string, unknown> = { bytes: frame.length } |
There was a problem hiding this comment.
nitpick (non-blocking): bytes: frame.length counts UTF-16 code units, not transmitted bytes — fine for order-of-magnitude dev logging, but the key name overpromises; chars: or a Blob(frame).size would be exact.
|
This slice is reference-only. Accepted work is routed to #16198 and #16353. Full context for agent readersPer Christian's main-target close ruling, #16145 (02b — human write leg) is no longer a landing vehicle. It remains readable implementation and review provenance. No accepted write-leg property residual is left without a named carrier. If a genuinely unaccounted delta is found, reopen this source or file a bounded main-based carrier rather than reviving the stacked merge path. |
Merge after: #16135 (02a) - stacked on its branch; retargets when it lands.
Decisions: Agent V1 Decision Log
Summary
The human write leg of the CRDT foundation, split out of #16135 so the follower (read leg) and the write leg get separate reviews. This branch restores the write-leg half of the reviewed integration state: the union of this branch and #16135 is byte-identical to the pre-split head (tree-hash verified).
Changes
src/workbench/extensions/agent/crdt/: op envelope minting, mint gate and session, layout/link/widget mint ports, port wiring, serialized op sender - human edits mint semantic ops into the shared doc.ComfyExtensionload hooks,layoutStore.withActor, a graph-traversal helper, four agent prompt-error registrations (resolver + en locale).Review Focus
Linear tickets