Repository navigation
fix(agent): gate persisted workflow tab bindings on document identity - #18077
Conversation
|
Important Review skippedBot user detected. To trigger a single review, invoke the ⚙️ Run configurationConfiguration used: Repository: Comfy-Org/ComfyUI_frontend/.coderabbit.yaml Review profile: ASSERTIVE Plan: Advanced Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
Comment |
🎭 Playwright: ✅ 2064 passed, 0 failed · 2 flaky📊 Browser Reports
🎨 Storybook: ✅ Built — View Storybook📦 Bundle: 9.95 MB gzip ⚪ 0 BDetailsSummary
Category Glance App Entry Points — 3.81 kB (baseline 3.81 kB) • ⚪ 0 BMain entry bundles and manifests Status: 1 unchanged Graph Workspace — 1.46 MB (baseline 1.46 MB) • ⚪ 0 BGraph editor runtime, canvas, workflow orchestration Status: 3 unchanged Views & Navigation — 132 kB (baseline 132 kB) • ⚪ 0 BTop-level views, pages, and routed surfaces Status: 18 unchanged Panels & Settings — 638 kB (baseline 638 kB) • ⚪ 0 BConfiguration panels, inspectors, and settings screens Status: 28 unchanged User & Accounts — 37.9 kB (baseline 37.9 kB) • ⚪ 0 BAuthentication, profile, and account management bundles Status: 11 unchanged Editors & Dialogs — 126 kB (baseline 126 kB) • ⚪ 0 BModals, dialogs, drawers, and in-app editors Status: 8 unchanged UI Components — 124 kB (baseline 124 kB) • ⚪ 0 BReusable component library chunks Status: 16 unchanged Data & Services — 3.87 MB (baseline 3.87 MB) • ⚪ 0 BStores, services, APIs, and repositories Status: 16 unchanged Utilities & Hooks — 566 kB (baseline 566 kB) • ⚪ 0 BHelpers, composables, and utility bundles Status: 38 unchanged Vendor & Third-Party — 17.4 MB (baseline 17.4 MB) • ⚪ 0 BExternal libraries and shared vendor chunks Status: 18 unchanged Other — 17.8 MB (baseline 17.8 MB) • ⚪ 0 BBundles that do not match a named category Status: 301 unchanged ⚡ Performance Report
Show regressions
All metrics
Historical variance (last 15 runs)
Trend (last 15 commits on main)
Raw data{
"timestamp": "2026-09-20T17:45:00.821Z",
"gitSha": "d42707160f8a1a32470eea3d7cd3d657840de436",
"branch": "test/pm-1255-stale-workflow-tab-binding",
"measurements": [
{
"name": "canvas-idle",
"durationMs": 2005.3379999999947,
"styleRecalcs": 10,
"styleRecalcDurationMs": 5.805999999999998,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 233.03799999999998,
"heapDeltaBytes": -1708072,
"heapUsedBytes": 61041652,
"domNodes": 20,
"jsHeapTotalBytes": 4456448,
"scriptDurationMs": 8.338999999999999,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.699999999999818
},
{
"name": "canvas-idle",
"durationMs": 2008.950999999911,
"styleRecalcs": 9,
"styleRecalcDurationMs": 5.417000000000002,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 346.997,
"heapDeltaBytes": -388808,
"heapUsedBytes": 61709324,
"domNodes": -262,
"jsHeapTotalBytes": 4186112,
"scriptDurationMs": 8.352,
"eventListeners": -152,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333335,
"p95FrameDurationMs": 16.699999999999818
},
{
"name": "canvas-mouse-sweep",
"durationMs": 1810.623000000021,
"styleRecalcs": 76,
"styleRecalcDurationMs": 23.27,
"layouts": 14,
"layoutDurationMs": 4.17,
"taskDurationMs": 714.7260000000001,
"heapDeltaBytes": 3354892,
"heapUsedBytes": 65504288,
"domNodes": -260,
"jsHeapTotalBytes": 5758976,
"scriptDurationMs": 120.80700000000002,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "canvas-mouse-sweep",
"durationMs": 1695.2559999999721,
"styleRecalcs": 73,
"styleRecalcDurationMs": 19.497,
"layouts": 14,
"layoutDurationMs": 3.581,
"taskDurationMs": 613.178,
"heapDeltaBytes": 10679412,
"heapUsedBytes": 72832496,
"domNodes": -260,
"jsHeapTotalBytes": 5234688,
"scriptDurationMs": 106.05799999999999,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "canvas-zoom-sweep",
"durationMs": 1720.882999999958,
"styleRecalcs": 32,
"styleRecalcDurationMs": 11.505,
"layouts": 6,
"layoutDurationMs": 1.278,
"taskDurationMs": 244.978,
"heapDeltaBytes": 2158600,
"heapUsedBytes": 64437708,
"domNodes": 78,
"jsHeapTotalBytes": 4980736,
"scriptDurationMs": 10.381,
"eventListeners": 15,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.670000000000012,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "canvas-zoom-sweep",
"durationMs": 1708.943999999974,
"styleRecalcs": 30,
"styleRecalcDurationMs": 11.601,
"layouts": 6,
"layoutDurationMs": 1.08,
"taskDurationMs": 249.847,
"heapDeltaBytes": 2839404,
"heapUsedBytes": 70127760,
"domNodes": 77,
"jsHeapTotalBytes": 5242880,
"scriptDurationMs": 10.078999999999999,
"eventListeners": 15,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.699999999999818
},
{
"name": "dom-widget-clipping",
"durationMs": 563.4860000000117,
"styleRecalcs": 9,
"styleRecalcDurationMs": 4.812999999999998,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 337.01399999999995,
"heapDeltaBytes": 17857524,
"heapUsedBytes": 80042044,
"domNodes": 14,
"jsHeapTotalBytes": 5505024,
"scriptDurationMs": 121.589,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "dom-widget-clipping",
"durationMs": 515.9780000000183,
"styleRecalcs": 11,
"styleRecalcDurationMs": 5.5840000000000005,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 325.584,
"heapDeltaBytes": 17759004,
"heapUsedBytes": 80054864,
"domNodes": 18,
"jsHeapTotalBytes": 5505024,
"scriptDurationMs": 116.57,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "large-graph-idle",
"durationMs": 2013.165000000015,
"styleRecalcs": 7,
"styleRecalcDurationMs": 4.3770000000000024,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 345.549,
"heapDeltaBytes": 1677488,
"heapUsedBytes": 83042220,
"domNodes": -243,
"jsHeapTotalBytes": -1544192,
"scriptDurationMs": 31.792999999999996,
"eventListeners": -152,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "large-graph-idle",
"durationMs": 2020.4160000000684,
"styleRecalcs": 8,
"styleRecalcDurationMs": 5.051,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 420.269,
"heapDeltaBytes": 1941536,
"heapUsedBytes": 83432248,
"domNodes": -242,
"jsHeapTotalBytes": -233472,
"scriptDurationMs": 28.816999999999997,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "large-graph-pan",
"durationMs": 2091.7720000000486,
"styleRecalcs": 67,
"styleRecalcDurationMs": 12.584999999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1220.2450000000003,
"heapDeltaBytes": 16162660,
"heapUsedBytes": 80100572,
"domNodes": -246,
"jsHeapTotalBytes": 1798144,
"scriptDurationMs": 594.866,
"eventListeners": -152,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "large-graph-pan",
"durationMs": 2154.5149999999467,
"styleRecalcs": 70,
"styleRecalcDurationMs": 12.507,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1192.5249999999999,
"heapDeltaBytes": -673192,
"heapUsedBytes": 79919348,
"domNodes": -216,
"jsHeapTotalBytes": 1265664,
"scriptDurationMs": 577.6629999999999,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "large-graph-zoom",
"durationMs": 3157.9189999999926,
"styleRecalcs": 64,
"styleRecalcDurationMs": 14.595000000000004,
"layouts": 60,
"layoutDurationMs": 14.94,
"taskDurationMs": 1319.3139999999999,
"heapDeltaBytes": 10074920,
"heapUsedBytes": 75079524,
"domNodes": -248,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 570.468,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "large-graph-zoom",
"durationMs": 3163.4950000000117,
"styleRecalcs": 64,
"styleRecalcDurationMs": 14.63,
"layouts": 60,
"layoutDurationMs": 14.395999999999999,
"taskDurationMs": 1339.931,
"heapDeltaBytes": -7018484,
"heapUsedBytes": 73883420,
"domNodes": -250,
"jsHeapTotalBytes": -1056768,
"scriptDurationMs": 569.409,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333335,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "legacy-node-drag",
"durationMs": 2174.254000000019,
"styleRecalcs": 44,
"styleRecalcDurationMs": 8.27,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1354.1129999999998,
"heapDeltaBytes": 18414448,
"heapUsedBytes": 84149380,
"domNodes": -248,
"jsHeapTotalBytes": 942080,
"scriptDurationMs": 712.138,
"eventListeners": 180,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "legacy-node-drag",
"durationMs": 2319.5899999999483,
"styleRecalcs": 41,
"styleRecalcDurationMs": 6.075,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 1394.1970000000001,
"heapDeltaBytes": 8158332,
"heapUsedBytes": 73825388,
"domNodes": 2,
"jsHeapTotalBytes": 225280,
"scriptDurationMs": 686.6560000000001,
"eventListeners": 180,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "minimap-idle",
"durationMs": 2037.4269999999797,
"styleRecalcs": 8,
"styleRecalcDurationMs": 4.915000000000001,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 445.75399999999996,
"heapDeltaBytes": -9917600,
"heapUsedBytes": 79108436,
"domNodes": -254,
"jsHeapTotalBytes": -757760,
"scriptDurationMs": 29.470999999999997,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "minimap-idle",
"durationMs": 2002.8479999999718,
"styleRecalcs": 8,
"styleRecalcDurationMs": 4.545000000000003,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 356.797,
"heapDeltaBytes": -11203540,
"heapUsedBytes": 69511412,
"domNodes": -261,
"jsHeapTotalBytes": 253952,
"scriptDurationMs": 27.092,
"eventListeners": -152,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333335,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "subgraph-dom-widget-clipping",
"durationMs": 425.1770000000761,
"styleRecalcs": 49,
"styleRecalcDurationMs": 8.4,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 282.58700000000005,
"heapDeltaBytes": 14115196,
"heapUsedBytes": 83689768,
"domNodes": 23,
"jsHeapTotalBytes": 5505024,
"scriptDurationMs": 92.497,
"eventListeners": 6,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.663333333333338,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "subgraph-dom-widget-clipping",
"durationMs": 463.1520000000364,
"styleRecalcs": 46,
"styleRecalcDurationMs": 7.4,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 301.285,
"heapDeltaBytes": 14391832,
"heapUsedBytes": 77913240,
"domNodes": 18,
"jsHeapTotalBytes": 5505024,
"scriptDurationMs": 102.063,
"eventListeners": 6,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66333333333332,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-idle",
"durationMs": 2004.1960000000358,
"styleRecalcs": 11,
"styleRecalcDurationMs": 6.705000000000001,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 350.28900000000004,
"heapDeltaBytes": 15920296,
"heapUsedBytes": 79560956,
"domNodes": -273,
"jsHeapTotalBytes": 4710400,
"scriptDurationMs": 7.254999999999997,
"eventListeners": -198,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "subgraph-idle",
"durationMs": 1990.4779999999391,
"styleRecalcs": 10,
"styleRecalcDurationMs": 5.942000000000001,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 333.189,
"heapDeltaBytes": -8648760,
"heapUsedBytes": 54647980,
"domNodes": -276,
"jsHeapTotalBytes": 4972544,
"scriptDurationMs": 6.850999999999999,
"eventListeners": -198,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "subgraph-mouse-sweep",
"durationMs": 1675.803999999971,
"styleRecalcs": 72,
"styleRecalcDurationMs": 19.86,
"layouts": 18,
"layoutDurationMs": 4.913,
"taskDurationMs": 625.1089999999999,
"heapDeltaBytes": -3084072,
"heapUsedBytes": 60741996,
"domNodes": -273,
"jsHeapTotalBytes": 5234688,
"scriptDurationMs": 82.76599999999999,
"eventListeners": -220,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-mouse-sweep",
"durationMs": 1694.5839999999635,
"styleRecalcs": 77,
"styleRecalcDurationMs": 20.317,
"layouts": 18,
"layoutDurationMs": 4.710000000000001,
"taskDurationMs": 605.2059999999999,
"heapDeltaBytes": 1047984,
"heapUsedBytes": 64743612,
"domNodes": -278,
"jsHeapTotalBytes": 6021120,
"scriptDurationMs": 77.042,
"eventListeners": -230,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-transition-enter",
"durationMs": 798.7060000000383,
"styleRecalcs": 19,
"styleRecalcDurationMs": 20.116999999999997,
"layouts": 5,
"layoutDurationMs": 7.861,
"taskDurationMs": 509.09400000000005,
"heapDeltaBytes": 19956496,
"heapUsedBytes": 103210592,
"domNodes": 13587,
"jsHeapTotalBytes": 11272192,
"scriptDurationMs": 19.309,
"eventListeners": 2383,
"totalBlockingTimeMs": 91,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "viewport-pan-sweep",
"durationMs": 8123.4980000000405,
"styleRecalcs": 249,
"styleRecalcDurationMs": 35.126999999999995,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 3770.595,
"heapDeltaBytes": -12084036,
"heapUsedBytes": 69469552,
"domNodes": -241,
"jsHeapTotalBytes": 4485120,
"scriptDurationMs": 1912.8,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.670000000000012,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "viewport-pan-sweep",
"durationMs": 8126.735000000053,
"styleRecalcs": 249,
"styleRecalcDurationMs": 39.342999999999996,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 4215.363,
"heapDeltaBytes": -5573696,
"heapUsedBytes": 75938628,
"domNodes": -243,
"jsHeapTotalBytes": 516096,
"scriptDurationMs": 1944.671,
"eventListeners": -174,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "vue-large-graph-idle",
"durationMs": 12025.41000000008,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 11262.101,
"heapDeltaBytes": -52841984,
"heapUsedBytes": 178162932,
"domNodes": -8062,
"jsHeapTotalBytes": -10805248,
"scriptDurationMs": 188.482,
"eventListeners": -14928,
"totalBlockingTimeMs": 0,
"frameDurationMs": 17.223333333333358,
"p95FrameDurationMs": 16.80000000000291
},
{
"name": "vue-large-graph-idle",
"durationMs": 11981.411000000095,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 11167.377000000002,
"heapDeltaBytes": -35415112,
"heapUsedBytes": 175341792,
"domNodes": -8062,
"jsHeapTotalBytes": -12521472,
"scriptDurationMs": 157.64800000000002,
"eventListeners": -14928,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "vue-large-graph-pan",
"durationMs": 14719.792999999981,
"styleRecalcs": 72,
"styleRecalcDurationMs": 16.011000000000024,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 14253.855000000001,
"heapDeltaBytes": -19638884,
"heapUsedBytes": 190153744,
"domNodes": -8062,
"jsHeapTotalBytes": -11644928,
"scriptDurationMs": 699.901,
"eventListeners": -14928,
"totalBlockingTimeMs": 5,
"frameDurationMs": 17.223333333333358,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "vue-large-graph-pan",
"durationMs": 14003.239000000009,
"styleRecalcs": 67,
"styleRecalcDurationMs": 14.172999999999991,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 13540.799,
"heapDeltaBytes": -42057316,
"heapUsedBytes": 189106300,
"domNodes": -8062,
"jsHeapTotalBytes": -17428480,
"scriptDurationMs": 710.4929999999999,
"eventListeners": -14928,
"totalBlockingTimeMs": 48,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "vue-node-resize-workload",
"durationMs": 4400.36699999996,
"styleRecalcs": 20,
"styleRecalcDurationMs": 55.22100000000002,
"layouts": 20,
"layoutDurationMs": 23.81699999999999,
"taskDurationMs": 4292.615,
"heapDeltaBytes": -43862512,
"heapUsedBytes": 192738288,
"domNodes": -8061,
"jsHeapTotalBytes": 11993088,
"scriptDurationMs": 75.43799999999999,
"eventListeners": -14928,
"totalBlockingTimeMs": 17,
"frameDurationMs": 17.223333333333358,
"p95FrameDurationMs": 16.80000000000109
},
{
"name": "vue-node-resize-workload",
"durationMs": 4341.4000000000215,
"styleRecalcs": 20,
"styleRecalcDurationMs": 55.26500000000001,
"layouts": 20,
"layoutDurationMs": 23.18200000000001,
"taskDurationMs": 4245.44,
"heapDeltaBytes": -13785008,
"heapUsedBytes": 202637516,
"domNodes": -8066,
"jsHeapTotalBytes": 13828096,
"scriptDurationMs": 71.954,
"eventListeners": -14928,
"totalBlockingTimeMs": 22,
"frameDurationMs": 17.223333333333297,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "vue-zoom-out-idle",
"durationMs": 6073.141999999962,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 5633.187999999999,
"heapDeltaBytes": -21659836,
"heapUsedBytes": 168543112,
"domNodes": 0,
"jsHeapTotalBytes": -20971520,
"scriptDurationMs": 83.69399999999999,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "vue-zoom-out-idle",
"durationMs": 4770.727999999963,
"styleRecalcs": 1,
"styleRecalcDurationMs": 0.13699999999999823,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 4463.328,
"heapDeltaBytes": -715972,
"heapUsedBytes": 181562204,
"domNodes": 0,
"jsHeapTotalBytes": -15204352,
"scriptDurationMs": 72.961,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "workflow-execution",
"durationMs": 431.0450000000401,
"styleRecalcs": 14,
"styleRecalcDurationMs": 15.799000000000001,
"layouts": 3,
"layoutDurationMs": 0.9129999999999999,
"taskDurationMs": 95.04899999999999,
"heapDeltaBytes": 4977628,
"heapUsedBytes": 66321316,
"domNodes": 123,
"jsHeapTotalBytes": 524288,
"scriptDurationMs": 7.9700000000000015,
"eventListeners": 81,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.670000000000012,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "workflow-execution",
"durationMs": 443.80799999999,
"styleRecalcs": 12,
"styleRecalcDurationMs": 15.077,
"layouts": 3,
"layoutDurationMs": 1.007,
"taskDurationMs": 93.71999999999998,
"heapDeltaBytes": 4871224,
"heapUsedBytes": 66596648,
"domNodes": 121,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 8.793,
"eventListeners": 81,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666666666666668,
"p95FrameDurationMs": 16.799999999999727
}
]
}🔬 E2E Coverage
Top 10 files by uncovered lines
🌾 Fallow: ✅ PassedFallow audits the diff for duplication, complexity and dead code. It gates on new findings only, so when the audit completes, a red is something this PR introduced or moved — never inherited backlog. A red can also mean the audit itself did not finish; the details below say which happened. No new findings in the changed files. Reproduce locally# the base the gate compared against
npx fallow audit --root . --quiet --format json \
--changed-since cab72b46b739dfa009812ecac7cfd7b5032e5eaa
Full docs: ADR-DEVEX-LINT-0015 🧩 Custom Nodes Ecosystem Matrix: ✅ PassedExecutes the frontend JS of every registry pack that ships any, against this change, in both renderers. The eight
On a pin-bump PR, a red gate means the ecosystem moved — not that your diff broke something. That is what the bump PR is for. Outside a pin bump or a cold rebuild of the unpinned tail, a red means the diff. Note the corpus size and the executed count are different numbers, and the verdict is against the executed count: roughly 3,100 of ~5,100 packs ship no extension-shaped JS and are skipped. Reproduce locally# build one shard's fixture, then run it
python3 scripts/registry-census/build_matrix.py --shard 1/4
MATRIX_RENDERER=vue MATRIX_VUE=1 \
npx vitest run --config vitest.matrix.config.mts
Full docs: scripts/registry-census/README.md |
d04a997 to
c7bf240
Compare
…255/PM-986) Every new/unsaved workflow tab defaults to the same path, and agentWorkflowTabBindingStore caches workflowId -> tabPath in plain, unscoped localStorage with no TTL and no per-page-load nonce. A tab closed without the SPA's own unbind() running leaves that binding behind, so a brand-new tab reopening the default path silently inherits it and the CRDT follower pulls the old workflow's real content into the new tab. Adds three failing/red repros pinning the exact gap (a fresh store/module instantiation racing a pre-seeded stale localStorage entry, not the in-app close path the existing tests already cover): - agentWorkflowTabBindingStore.test.ts: store-level, asserts matchesWorkflow/tabPathFor reject the stale binding. - AgentPanelRoot.test.ts: component-level, asserts the real CRDT follower does not send a doc_subscribe for the abandoned workflow. - browser_tests/tests/agent/agentStaleWorkflowTabBinding.spec.ts: Playwright E2E through a routed /ws and a seeded HostDoc, asserting the canvas stays empty and attaching a screenshot as visual proof. All three use test.fails()/test.fail() so CI records them as known, tracked failures. No fix included; this is tests only.
c7bf240 to
a829da6
Compare
|
Fixed a test-authoring bug CI's Moved |
Playwright videos for new testsDownload the recorded videos and HTML report Recorded projects: cloud Recorded test titles:
|
Codecov Report❌ Patch coverage is
@@ Coverage Diff @@
## main #18077 +/- ##
==========================================
+ Coverage 82.78% 83.95% +1.16%
==========================================
Files 2939 2260 -679
Lines 144683 128703 -15980
Branches 37233 35080 -2153
==========================================
- Hits 119773 108048 -11725
+ Misses 23636 20365 -3271
+ Partials 1274 290 -984
Flags with carried forward coverage won't be shown. Click here to find out more.
... and 833 files with indirect coverage changes 🚀 New features to boost your workflow:
|
Triviality verdict: not trivial. Plan below, no fix pushed.I read Why the FEC-5 pattern does not transplant cleanlyFEC-5's nonce is minted once per module evaluation (
A per-page-load nonce would refuse both of these on every genuine reload, since the nonce is, by construction, different after every reload. That regresses a shipped feature (the sidebar/chat staying attached to its workflow across a refresh or browser restart), not a hypothetical one. The actual disambiguation problemThe bug and the legitimate case are structurally identical from the store's point of view: both are "a workflow appears at a previously-bound path, with no known in-memory instance." The only thing separating "same draft, legitimately restored" from "unrelated new tab, coincidental collision" is that every anonymous new tab deterministically gets the exact same default path once the old one is gone ( A primitive that might actually fit, already in this codebase
Recommended approach
Open questions that need a human call before implementing
Alternatives considered
No source files were changed; the three |
…point A browser tab closed without the SPA's cleanup leaves its workflow binding at the shared default path. On the next boot the surviving thread hydrates, names that workflow, and the panel resolves it to the brand-new tab that merely reuses the path: the next turn posts as that workflow and the CRDT follower pulls its content onto the empty canvas. Store tests pin the fix's contract (document identity as proof, a separate v2 storage key migrated once from the legacy map, non-destructive refusal, a confirmedAt TTL). The component and E2E repros now exercise the real chain (thread pointer -> hydrate -> onWorkflowRestored -> bind) instead of an agent_active_tab push. All new tests are red against the current store by design; the fix follows. The ADR AGENT-BINDING-0035 records the decision, the rollback story and the upgrade blast radius.
A binding record is still keyed by tab path, but a tab may only claim it when its workflow JSON id matches the id captured at bind time. Every fresh draft mints a new id and every restore preserves the persisted one, so a same-tab reload and a browser-restart restore keep the sidebar bound while a brand-new tab that merely reuses the default path is refused. Refusal is non-destructive: the record is neither adopted nor deleted while the unproven occupant is open, and only the proven owner releases it on close. Records move to Comfy.Agent.WorkflowTabBindings.v2 with a one-shot migration from the legacy map, which is left in place so a rollback build keeps reading the shape it wrote. Migrated records carry no id and can only be claimed by saved workflows. A confirmedAt stamp, refreshed by bind() and by a proven adoption, prunes records older than 30 days at store creation. A draft refused for a record cannot become its owner by being saved in place at the record's path, and saved tabs compare ids too when both are known. The watcher reads records ungated; only tabPathFor and workflowIdFor apply the occupant gate, since the navigation consumers resolve them to a tab without consulting matchesWorkflow. The ADR AGENT-BINDING-0035 records the decision.
…rkflow-tab-binding
|
Fix landed on top of the repros, red then green. Design. Bindings stay keyed by tab path, but the workflow JSON What the design review changed before implementation. (1) New storage key instead of widening the old one: under the old key a rollback build would have returned the new record objects as paths and broken ack adoption. (2) The component and E2E repros were rewritten around the real entry point. The earlier note above that the leak fires on the boot WS Commits. Not run locally. The Playwright spec needs the app server, which this environment does not have; it is type-checked and linted, and the cloud project in CI runs it. Everything else (agent vitest suite, Generated by Claude Code |
|
Source fix merged. Both release backports tested and opened. Original-session acceptance remains open. Full context for agent readersBinding PR18077 merged at 18:31:34 UTC from the reviewed, tested Claude head as the source commit. This supersedes the earlier queued status; no competing refactor was pushed. Opened Core 1.54 backport at tested Core head and Cloud 1.54 backport at tested Cloud head. Both preserve the production fix byte-for-byte, adapt release test isolation/imports and use an equivalent stale-document seed instead of an unavailable main-only loader. Each passes 239 focused tests, source/browser types, lint, formatting, knip and the black-box stale-binding browser case. Screenshots were inspected. Restoring the old store on the Cloud carrier fails the outgoing workflow-ID assertion; restoring the fix passes. Earlier missing-button attempts used the wrong serving process, not a proven feature-gate regression. Normal auto-merge is enabled on both release PRs; hosted checks and non-author review remain required. The completion lane holds their PR reservations and retains original-session acceptance, both release merges and the proposed retention decision. No incident closure, architecture ratification or deployment. Program receipt. Glossary: binding associates an Agent workflow with an editor tab; backport applies a merged fix to a release branch. |
The open-tab watcher released closed tabs and adopted open ones in a single callback whose branching tripped the new-code complexity gate. Each half now lives in its own function with the same behavior. Also covers the two parse fallbacks: a legacy key that does not hold JSON yields no bindings, and a saved tab whose stored content is not JSON can still claim its record.
christian-byrne
left a comment
There was a problem hiding this comment.
The document-identity check prevents a fresh draft at a reused path from inheriting an abandoned Agent workflow. Restored drafts still reconnect. Original-session and release acceptance remain tracked separately.
Full context for agent readers
Reviewed the exact published head, including persisted-record validation, lookup gating, verified-owner cleanup, saving a refused draft in place, reload ordering, and the caller paths that restore or adopt a target. No blocking correctness finding in this change.
The unchanged head has 239 passing store/component tests and a passing Cloud browser regression. The controlled pre-fix-store run sent the abandoned workflow ID; the fixed store sends the fresh target and preserves the starting canvas. The local exact-base Fallow audit has no introduced findings. Hosted checks are terminal: 85 successful check runs, 16 skipped, and the successful CodeRabbit status. No unresolved review threads or human changes-requested reviews were present.
Main has advanced through disjoint website, asset-query and opt-in conversation-host fixture changes. None overlaps this patch or the panel mock used by its browser case. Normal merge-queue integration checks remain required; no administrative bypass or redundant branch refresh.
This review does not ratify the proposed binding architecture decision or settle the thirty-day retention product choice. The original-session check, retention decision and core/1.54 plus cloud/1.54 qualification stay with the stale-binding incident and completion lane. It does not establish full incident closure.
Glossary: binding associates an Agent workflow with a canvas tab; exact head is the reviewed commit; Fallow is the static-analysis gate.
…#18077) <!-- ccr-slack-attribution --> _Requested by **Christian Byrne** · [Slack thread](https://comfy-organization.slack.com/archives/C0BGH348Z0C/p1789592076376929?thread_ts=1789592076.376929&cid=C0BGH348Z0C)_ **Before:** Every new unsaved tab opens at the same default path, and the agent panel remembers which tab belongs to which agent workflow by that path. When a bound tab was closed without the app's own cleanup (the browser tab was simply closed), its record stayed behind. On the next visit the still-open chat thread named that workflow, the panel matched it to the brand-new tab by path, the next message was sent as that workflow, and the agent's document was pulled onto the fresh canvas over the default graph. **After:** A remembered binding only attaches to a tab that carries the same workflow document id it was bound with. A fresh tab has a new id, so it is refused and keeps its default canvas and its own identity; a reload or a browser-restart restore of the same draft keeps its id, so the sidebar stays attached as before. This change makes the agent workflow tab binding store prove ownership by document identity instead of by tab path alone. **How:** Records move to a new storage key (`Comfy.Agent.WorkflowTabBindings.v2`) as `{ tabPath, graphId, confirmedAt }`, migrated once from the legacy map, which is left in place so a rollback keeps reading the shape it wrote. The store's watcher adopts a tab only when the ids match; the exported lookups hide a record while an unproven tab occupies its path, without deleting it, so a live binding in another browser tab is never damaged. A 30-day `confirmedAt` TTL prunes dead records at store creation. A draft refused for a record cannot become its owner by being saved in place at that path. Alternatives (page-load nonce, tab identity, re-keying on the graph id) and the consequences, including the one-time upgrade cost for open unsaved drafts, are in ADR `docs/adr/AGENT-BINDING-0035-document-identity-gates-persisted-workflow-tab-bindings.md`. - Red then green in two commits: `d44406a` (tests and ADR only; the new store, component and E2E tests fail against the old store) then `f7b06a2` (the fix, all green). - `eccfc83` repairs the E2E spec: a fresh tab boots with the default graph, not an empty canvas, so the spec now captures the boot node titles and asserts they are unchanged after the turn, that the outgoing turn carries the newly saved workflow id, and that the stale ack is rejected visibly. - `77915be` splits the store watcher into two named helpers (same behavior) to clear the fallow complexity gate, and adds tests for the two parse fallbacks (non-JSON legacy key, non-JSON `originalContent`) that Codecov flagged. - `pnpm vitest run src/workbench/extensions/agent/AgentPanelRoot.test.ts src/workbench/extensions/agent/stores/agent/agentWorkflowTabBindingStore.test.ts`: 239 passed; store file at 100% line coverage. - `pnpm typecheck`, `pnpm typecheck:browser`, `pnpm knip`, `pnpm adr:check`, eslint and `oxfmt --check` on every touched file, and `pnpm fallow audit --changed-since origin/main` (verdict pass, no introduced findings): clean on the pushed head. - Not run locally: `browser_tests/tests/agent/agentStaleWorkflowTabBinding.spec.ts` needs the app server, which this environment does not have. It is type-checked and linted; CI's cloud Playwright project (and the new-tests video job) is the only place it runs. --------- Co-authored-by: bymyself <cbyrne@comfy.org> (cherry picked from commit be3d753)
## Summary Backport the merged [persisted-tab binding fix](#18077) so a fresh canvas cannot inherit an abandoned chat's workflow identity. ## Changes - Production store is identical to the merged source; release-specific test setup and browser imports are preserved. - Browser coverage uses an equivalent stale MarkdownNote document because the main-only recorded-conversation loader is absent on the release branch. ## Review Focus Document identity, saved-in-place refusal, restoration ordering and cross-tab ownership. The thirty-day retention ADR remains Proposed. [Original incident](https://linear.app/comfyorg/issue/PM-1255) stays open through original-session and release acceptance. <details><summary>Full context for agent readers</summary> Source: [merged commit](be3d753). This is a manual backport because the test setup and ADR index do not apply unchanged; no unrelated main commits are imported. Target-native validation: 239 store/component tests, source and browser typechecks, lint, formatting and knip pass. The black-box browser regression passes on a verified Cloud-distribution server against the qualified local backend. It checks the newly saved outgoing workflow identity, rejected stale acknowledgement, unchanged fresh canvas and active tab. Screenshots were inspected. Sensitivity control on the Cloud release: restoring only the old production store makes the outgoing workflow identity assertion fail with the abandoned workflow ID. The two release carriers use the same production fix and equivalent browser assertions. Earlier browser attempts used an occupied/wrong-checkout port and are not evidence of a product regression. No tests were skipped or weakened. This PR does not deploy the release or close the broader incident. Glossary: backport applies a merged fix to a release branch; binding associates an Agent workflow with an editor tab. </details>
## Summary Backport the merged [persisted-tab binding fix](#18077) so a fresh canvas cannot inherit an abandoned chat's workflow identity. ## Changes - Production store is identical to the merged source; release-specific test setup and browser imports are preserved. - Browser coverage uses an equivalent stale MarkdownNote document because the main-only recorded-conversation loader is absent on the release branch. ## Review Focus Document identity, saved-in-place refusal, restoration ordering and cross-tab ownership. The thirty-day retention ADR remains Proposed. [Original incident](https://linear.app/comfyorg/issue/PM-1255) stays open through original-session and release acceptance. <details><summary>Full context for agent readers</summary> Source: [merged commit](be3d753). This is a manual backport because the test setup and ADR index do not apply unchanged; no unrelated main commits are imported. Target-native validation: 239 store/component tests, source and browser typechecks, lint, formatting and knip pass. The black-box browser regression passes on a verified Cloud-distribution server against the qualified local backend. It checks the newly saved outgoing workflow identity, rejected stale acknowledgement, unchanged fresh canvas and active tab. Screenshots were inspected. Sensitivity control on the Cloud release: restoring only the old production store makes the outgoing workflow identity assertion fail with the abandoned workflow ID. The two release carriers use the same production fix and equivalent browser assertions. Earlier browser attempts used an occupied/wrong-checkout port and are not evidence of a product regression. No tests were skipped or weakened. This PR does not deploy the release or close the broader incident. Glossary: backport applies a merged fix to a release branch; binding associates an Agent workflow with an editor tab. </details> Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
…option, op coalescing) + contract e2e (Comfy-Org#18064) <!-- ccr-slack-attribution --> _Requested by **Christian Byrne** · [Slack thread](https://comfy-organization.slack.com/archives/C0BGH348Z0C/p1789670879302689?thread_ts=1789670879.302689&cid=C0BGH348Z0C)_ **Before:** Clearing a saved workflow's canvas, deleting the chat and starting a new one left the agent reasoning about the removed nodes: an N-node delete went out as N serialized single-op batches and most of them were dropped when the chat unbound. The e2e for this flow was an expected-failure repro asserting the wrong contract (a fresh `workflow_id`). The second door to the same symptom, a blank scratch tab opened after a reload adopting a persisted binding for the shared default path, landed separately in Comfy-Org#18077 (document identity gates persisted tab bindings); this PR is rebased onto it and no longer carries a gate of its own for that door. **After:** Operations minted in one tick are coalesced into one batch, so a clear made while bound is one round trip; a `doc_reset` settles pending human batches `undeliverable` at once. The e2e is a plain `@cloud` contract test: the first turn of the new chat posts to a new thread with the same `workflow_id` and an empty draft, every cleared node id reaches the host document as a `delete_node`, and after the ack and re-subscribe the canvas stays empty. Keeps a new agent chat from inheriting a stale canvas through the clear-while-bound door and pins the client side of the contract the backend change relies on. **How.** `docs/adr/CRDT-AUTHORITY-0035-human-canvas-authority-and-draft-reconciliation.md` records the decisions: the human-authored `draft` is the reconciliation primitive (conditional on the backend change that makes a posted human canvas authoritative, merged to the agent release branch and awaiting promotion to main), a new chat keeps the same `workflow_id`, ops are best-effort with same-tick coalescing (`crdt/opCoalescer.ts`, wired in `useAgentCrdtFollower`), `opSender.abortAll()` on `doc_reset`, and the blank-tab door is closed by the document-identity gate of ADR AGENT-BINDING-0035 (Comfy-Org#18077) rather than by a second gate here. The fake host's apply mode (`humanOpsHost: 'apply'`, from Comfy-Org#18078) applies human ops through the real library and answers `doc_ops_result` plus the `doc_update` echo; the contract spec opts into it, and this PR adds only the relay's `ok:false` answer for a `doc_ops` batch aimed at a workflow the host does not serve. Coalescing only narrows the clear-while-bound ordering; edits made while unbound reach the document only through the next turn's draft. **Commits, red to green.** `test:` reworked spec + failing unit tests (red: one op reaches the never-answering host; `abortAll` missing; eight batches instead of one; no settle on `doc_reset`) → `fix:` coalescer, `abortAll` + reset abort (units green; spec still red at the host-document outcome because all eight ops arrive but the fake host does not apply them) → `docs:` ADR → `test:` spec opts into the fake host's apply mode, host answers foreign-workflow `doc_ops` (spec green; `agentConversationReplay`, `agentTabSwitchCatchUp`, `load3dAgentUpdate` stay green) → `test:` flush the coalescer before asserting in the R-73 cross-workflow tests → `refactor:` drop the unused coalescer `pending()`. **Verified locally** on the rebased head: eslint, oxfmt check, `tsc --noEmit`, `vue-tsc` for `browser_tests`, knip, `fallow audit --changed-since origin/main` pass (0 introduced), `adr:check`, Vitest for `src/workbench/extensions/agent` and `src/platform/workflow/persistence` (128 files, 1976 passed, 5 expected fail). Playwright `cloud` project on an earlier head of this branch: the new spec, `agentConversationReplay` (all recordings), `agentTabSwitchCatchUp`, `load3dAgentUpdate`, 19 passed; not re-run after the rebase. The one local failure there was the pre-existing `@screenshot` replay case, identical 25279-pixel diff on the base commit (local Chromium revision drift, Linux CI owns the baseline). **Still open.** Backend promotion to main (until then the saved-target door still fails in production). Follow-ups: withhold `draft` while a bound follower is disconnected or in a gap; surface `undeliverable` human ops to the user (`detach()` drops the queue with no event today); the target-keyed follower of GRAPH-DOCUMENT-0024; a server-side thread delete so the client tombstone can go; two browser tabs on one document are out of scope. 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_01LZ1S2w986aWaX4XTw2ynmo --------- Co-authored-by: Claude <noreply@anthropic.com> Co-authored-by: Christian Byrne <cbyrne@comfy.org>
…18280) <!-- ccr-slack-attribution --> _Requested by **Christian Byrne** · [Slack thread](https://comfy-organization.slack.com/archives/C0BGH348Z0C/p1789946234541539?thread_ts=1789946234.541539&cid=C0BGH348Z0C)_ ## Summary **Before:** once the agent service refuses a turn's `workflow_id` with `403 {"error":"workflow not found or access denied"}`, the tab that produced that id is broken permanently. Every later message from that tab posts the same dead id and fails the same way, and reloading the page does not help. The only escape is opening a brand new workflow tab. This is what a user hit after completing the subscribe flow: credits landed, the prompt never ran, and re-sending in the same tab kept failing with "Message failed to send: workflow not found or access denied". **After:** that refusal releases the tab's workflow binding and the session's bound workflow instead of keeping them. The failed send still reports the failure, and the next send from the same tab resolves the tab again rather than re-posting an id the server has already rejected. No new tab needed. ## Changes - **What**: `useAgentSession`'s send-failure path now recognises the workflow-scoped 403 and calls `bindingStore.unbind()` for the tab the turn was posted from, clearing `boundWorkflowId` and the page-lifetime remembered workflow when they name the same id. A `zDisownedWorkflowError` schema sits next to `zAgentAdmissionError` and matches that one body. ## Review Focus The mechanism, and the limits of what is proven here. The tab → workflow binding lives in `localStorage`, and the thread carries its own server-side workflow pointer, so the bad state is doubly persistent: `onWorkflowRestored()` re-affirms the binding from the thread's pointer on every boot, which is why a refresh does not clear it. Nothing in the client ever treated a server refusal as evidence that the binding is dead. That is the divergence this fixes, and it is the reason a recoverable one-off becomes a permanently broken tab. It is the same family as Comfy-Org#18077 (fresh tab inherits a stale binding) and Comfy-Org#18196 (new chat inherits a stale target), one layer further out. What this does **not** prove is *why* the workflow becomes unservable around the subscribe round trip. The backend collapses two cases into one message (the row is gone, or it belongs to another workspace), deliberately, so the response cannot tell them apart. Confirming which one it is needs a live repro with a network capture plus the server-side log line for the rejected `workflow_id`; until then this is a fix for the persistence of the failure, not for its trigger. Worth noting the thread itself passes its own workspace check in the same handler, so whatever happens, it separates the thread from its workflow rather than moving both. Discrimination is by response body, because a 403 refusing the thread or the message has the same shape with a different subject; the regression test pins both directions so a thread-scoped refusal cannot start eating workflow bindings. ## Testing - `pnpm vitest run src/workbench/extensions/agent/composables/agent/useAgentSession.test.ts` — the new `(k2)` table-driven case was committed red first (binding survived the workflow-scoped 403) and goes green with the fix; the thread-scoped row passed in both states. - `pnpm vitest run src/workbench/extensions/agent` — 121 files, 1868 passed. - `pnpm typecheck` clean, `pnpm lint` 0 errors (warnings all pre-existing), `pnpm knip` clean, `pnpm format` applied. Not exercised against a live checkout round trip; no browser test added, since the behaviour under test is a client-side reaction to one HTTP response and the composable is the lowest level that proves it. 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_01GNFfTjWvYLP9CZ1udGpjdA --- _Generated by [Claude Code](https://claude.ai/code/session_01GNFfTjWvYLP9CZ1udGpjdA)_ --------- Co-authored-by: Claude <noreply@anthropic.com>
…g#18221) <!-- CURSOR_AGENT_PR_BODY_BEGIN --> **Summary** Comfy-Org#18077's document-identity restore finishes after history Edit (`onSelectHistory` resets the composer state immediately, then `restoreTarget` → `setWorkflowTarget` → `setNodeScope`). `setNodeScope` used to bump `promptEpoch` even when there were no node chips, resetting the editor's ProseMirror state; Select All + type then replaced the prompt with ` again`. Scope changes without node references no longer reset editor history, and ArrowLeft/ArrowRight collapse a non-empty selection so chips survive Send → history → Edit. In-session tab binds from Comfy-Org#18077 stay gated the same way. **Verification** ```bash DISTRIBUTION=cloud pnpm exec playwright test --project=cloud \ browser_tests/tests/agent/agentInlinePrompt.spec.ts --grep "preserves inline workflow references" ``` Expected: chips survive Send, history, and Edit (not wiped to ` again`). This Playwright scenario predates this PR and already covers the full Send → history → Edit integration path. This PR extends the component regression coverage for both ArrowLeft and ArrowRight rather than duplicating the E2E test. ```bash pnpm exec vitest run \ src/workbench/extensions/agent/stores/agent/agentComposerStore.test.ts \ src/workbench/extensions/agent/stores/agent/agentWorkflowTabBindingStore.test.ts \ src/workbench/extensions/agent/components/agent/Composer.test.ts ``` <!-- CURSOR_AGENT_PR_BODY_END --> <div><a href="https://cursor.com/agents/bc-081de356-67e6-5393-bc9d-c5ffe92edcaa?cursor_ref=pr_footer&cursor_cta=open_in_web"><picture><source media="(prefers-color-scheme: dark)" srcset="https://cursor.com/assets/images/open-in-web-dark.png"><source media="(prefers-color-scheme: light)" srcset="https://cursor.com/assets/images/open-in-web-light.png"><img alt="Open in Web" width="114" height="28" src="https://cursor.com/assets/images/open-in-web-dark.png"></picture></a> <a href="https://cursor.com/background-agent?bcId=bc-081de356-67e6-5393-bc9d-c5ffe92edcaa&cursor_ref=pr_footer&cursor_cta=open_in_cursor"><picture><source media="(prefers-color-scheme: dark)" srcset="https://cursor.com/assets/images/open-in-cursor-dark.png"><source media="(prefers-color-scheme: light)" srcset="https://cursor.com/assets/images/open-in-cursor-light.png"><img alt="Open in Cursor" width="131" height="28" src="https://cursor.com/assets/images/open-in-cursor-dark.png"></picture></a> </div> --------- Co-authored-by: Cursor Agent <cursoragent@cursor.com> Co-authored-by: Amp <amp@ampcode.com>
Requested by Christian Byrne · Slack thread
Before: Every new unsaved tab opens at the same default path, and the agent panel remembers which tab belongs to which agent workflow by that path. When a bound tab was closed without the app's own cleanup (the browser tab was simply closed), its record stayed behind. On the next visit the still-open chat thread named that workflow, the panel matched it to the brand-new tab by path, the next message was sent as that workflow, and the agent's document was pulled onto the fresh canvas over the default graph.
After: A remembered binding only attaches to a tab that carries the same workflow document id it was bound with. A fresh tab has a new id, so it is refused and keeps its default canvas and its own identity; a reload or a browser-restart restore of the same draft keeps its id, so the sidebar stays attached as before.
This change makes the agent workflow tab binding store prove ownership by document identity instead of by tab path alone.
How: Records move to a new storage key (
Comfy.Agent.WorkflowTabBindings.v2) as{ tabPath, graphId, confirmedAt }, migrated once from the legacy map, which is left in place so a rollback keeps reading the shape it wrote. The store's watcher adopts a tab only when the ids match; the exported lookups hide a record while an unproven tab occupies its path, without deleting it, so a live binding in another browser tab is never damaged. A 30-dayconfirmedAtTTL prunes dead records at store creation. A draft refused for a record cannot become its owner by being saved in place at that path. Alternatives (page-load nonce, tab identity, re-keying on the graph id) and the consequences, including the one-time upgrade cost for open unsaved drafts, are in ADRdocs/adr/AGENT-BINDING-0035-document-identity-gates-persisted-workflow-tab-bindings.md.Testing
d44406a(tests and ADR only; the new store, component and E2E tests fail against the old store) thenf7b06a2(the fix, all green).eccfc83repairs the E2E spec: a fresh tab boots with the default graph, not an empty canvas, so the spec now captures the boot node titles and asserts they are unchanged after the turn, that the outgoing turn carries the newly saved workflow id, and that the stale ack is rejected visibly.77915besplits the store watcher into two named helpers (same behavior) to clear the fallow complexity gate, and adds tests for the two parse fallbacks (non-JSON legacy key, non-JSONoriginalContent) that Codecov flagged.pnpm vitest run src/workbench/extensions/agent/AgentPanelRoot.test.ts src/workbench/extensions/agent/stores/agent/agentWorkflowTabBindingStore.test.ts: 239 passed; store file at 100% line coverage.pnpm typecheck,pnpm typecheck:browser,pnpm knip,pnpm adr:check, eslint andoxfmt --checkon every touched file, andpnpm fallow audit --changed-since origin/main(verdict pass, no introduced findings): clean on the pushed head.browser_tests/tests/agent/agentStaleWorkflowTabBinding.spec.tsneeds the app server, which this environment does not have. It is type-checked and linted; CI's cloud Playwright project (and the new-tests video job) is the only place it runs.