Repository navigation
feat(agent): delete approval card for the agent's node deletes - #20092
Conversation
The agent's generic ask_user prompt reached the panel as an agent_ask frame
but was dropped as 'unrendered-kind', so a turn parked on it showed nothing.
Render it as a card with every option: radios for a required single
choice, checkboxes (capped at max_selections) otherwise, and an Other
free-text field when allow_other is set. Free text counts as one selection,
mirroring the server's validateSelection.
The answer goes through the same path run_approval uses (answerAsk ->
POST /agent/threads/:id/asks/:ask_id/answer), now carrying
{ selected, other_text } bound to the card's own ask_id, with the existing
retry, 409/403/404 handling, watchdog and superseded-answer reporting.
On agent_ask_resolved (or an out-of-band retirement) a run-approval card is
still removed, while an ask_user card stays read-only showing the settled
options and, only when the settled ids match what this client sent, its
own free text. A redelivered agent_ask for the same ask_id no longer adds a
second card. Pending asks hydrate from transcript history, including a
warning notice for a kind the panel cannot render, and keep the turn live.
…note resolved cards are session-only
…ach turn
A delete_approval ask reuses the question card: the host's delete/keep
options plus a compact list of the nodes it would remove, read from
context.nodes (title, else type, with the id). A malformed node list still
shows the prompt and options. The answer goes through the shared answerAsk
path as { selected: ["delete" | "keep"] }.
The turn request now sends ask_kinds, the kinds this panel renders. It comes
from RENDERED_ASK_KINDS, the keys of the same builder map toAskPart
dispatches through, so the advertised list and the renderer cannot drift.
Resolves the turn POST body conflict: main's TurnPostBody now also carries ask_kinds.
A delete_approval past the 50-node display cap, or with unreadable entries, now shows "N more nodes not listed" so a Delete answer never covers nodes the user was not told about. The node and ask-kind tests are table cases.
…ds as a literal tuple context.nodes entries are read through zAgentAskNodeRef in agentApiSchema, the same per-entry safeParse the transcript uses for tool calls, instead of an unknown-to-record cast. RENDERED_ASK_KINDS is a literal tuple and the builder map is a Record over it, so a kind without a builder (or a builder without a kind) fails typecheck, with no Object.keys cast.
- ask_kinds advertises run_approval and ask_user only. RENDERED_ASK_KINDS and RenderedAskKind live in agentApiSchema, so the REST client no longer imports the UI parts module. delete_approval moves to its own PR. - The answer body is the generated AgentAnswerRequest. - agentMessageParts no longer imports i18n: an unrenderable ask is an askUnavailable part, and AgentNoticeCard renders its text. One askIdOf helper replaces the four hand-written "belongs to this ask" checks. - Run-approval telemetry fires only for an ask whose run-approval card was shown, so an ask_user offering run/cancel never reports as one. - A resolved ask_user with nothing to read back says it was answered somewhere else instead of "Answered" over an empty list. - AskUserCheckboxOptions mirrors AskUserRadioOptions, with shared option id helpers; the radios pass disabled to their labels; the dead askId watch and the duplicate submit string are gone (g.submit). - Tests: no loops in bodies, an options object for renderCard, test.step phases in the e2e with i18n-built labels, unit edge cases left to unit tests.
A delete_approval ask renders as its own card, built like the run
approval card: the host's prompt, the nodes it would remove (title, else
type, plus the id; unique by id, capped at 50, with a count of anything not
listed), and two buttons, Keep and Delete (destructive). The answer goes
through the shared answerAsk path as { selected: ["delete" | "keep"] }, and
the resolved card stays read-only with the decision. An ask that does not
offer exactly delete/keep as one required choice falls back to the warning
notice. ask_kinds now advertises delete_approval.
📦 Bundle: 9.49 MB gzip 🔴 +57 BDetailsSummary
Category Glance App Entry Points — 6.62 kB (baseline 6.62 kB) • ⚪ 0 BMain entry bundles and manifests
Status: 1 added / 1 removed Graph Workspace — 3.46 MB (baseline 3.46 MB) • ⚪ 0 BGraph editor runtime, canvas, workflow orchestration
Status: 3 added / 3 removed / 1 unchanged Views & Navigation — 147 kB (baseline 147 kB) • ⚪ 0 BTop-level views, pages, and routed surfaces
Status: 12 added / 12 removed / 6 unchanged Panels & Settings — 670 kB (baseline 670 kB) • ⚪ 0 BConfiguration panels, inspectors, and settings screens
Status: 11 added / 11 removed / 18 unchanged User & Accounts — 151 kB (baseline 151 kB) • ⚪ 0 BAuthentication, profile, and account management bundles
Status: 9 added / 9 removed / 8 unchanged Editors & Dialogs — 359 kB (baseline 359 kB) • ⚪ 0 BModals, dialogs, drawers, and in-app editors
Status: 8 added / 8 removed / 4 unchanged UI Components — 212 kB (baseline 212 kB) • ⚪ 0 BReusable component library chunks
Status: 7 added / 7 removed / 17 unchanged Data & Services — 1.11 MB (baseline 1.11 MB) • ⚪ 0 BStores, services, APIs, and repositories
Status: 31 added / 31 removed / 6 unchanged Utilities & Hooks — 998 kB (baseline 998 kB) • ⚪ 0 BHelpers, composables, and utility bundles
Status: 37 added / 37 removed / 35 unchanged Vendor & Third-Party — 13.9 MB (baseline 13.9 MB) • ⚪ 0 BExternal libraries and shared vendor chunks Status: 17 unchanged Other — 19.7 MB (baseline 19.7 MB) • 🔴 +570 BBundles that do not match a named category
Status: 96 added / 96 removed / 270 unchanged ⚡ Performance Report
Show regressions
All metrics
Historical variance (last 15 runs)
Trend (last 15 commits on main)
Raw data{
"timestamp": "2026-10-16T01:20:03.351Z",
"gitSha": "7ecb043d8f3db71dff878e88f92135ae5ee66156",
"branch": "kishore/agent-delete-approval-card",
"measurements": [
{
"name": "canvas-idle",
"durationMs": 2024.4619999999998,
"styleRecalcs": 9,
"styleRecalcDurationMs": 2.222,
"layouts": 9,
"layoutDurationMs": 2.5590000000000006,
"taskDurationMs": 546.9340000000001,
"heapDeltaBytes": -20301156,
"heapUsedBytes": 58326096,
"domNodes": -162,
"jsHeapTotalBytes": 7077888,
"scriptDurationMs": 8.745000000000003,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665573770491804,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "canvas-idle",
"durationMs": 2008.3550000000514,
"styleRecalcs": 10,
"styleRecalcDurationMs": 2.788,
"layouts": 10,
"layoutDurationMs": 3.004,
"taskDurationMs": 553.918,
"heapDeltaBytes": -23025096,
"heapUsedBytes": 55702252,
"domNodes": -162,
"jsHeapTotalBytes": 7077888,
"scriptDurationMs": 8.711999999999998,
"eventListeners": -67,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66611570247934,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "canvas-mouse-sweep",
"durationMs": 1792.7530000000047,
"styleRecalcs": 76,
"styleRecalcDurationMs": 12.988,
"layouts": 9,
"layoutDurationMs": 2.5399999999999996,
"taskDurationMs": 911.222,
"heapDeltaBytes": 897440,
"heapUsedBytes": 79627272,
"domNodes": -162,
"jsHeapTotalBytes": 7077888,
"scriptDurationMs": 153.609,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66635514018692,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "canvas-mouse-sweep",
"durationMs": 1959.5760000000269,
"styleRecalcs": 82,
"styleRecalcDurationMs": 13.963,
"layouts": 10,
"layoutDurationMs": 2.7239999999999998,
"taskDurationMs": 933.5749999999999,
"heapDeltaBytes": 639748,
"heapUsedBytes": 79415292,
"domNodes": -162,
"jsHeapTotalBytes": 8126464,
"scriptDurationMs": 156.234,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66610169491526,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "canvas-zoom-sweep",
"durationMs": 1739.0320000000088,
"styleRecalcs": 19,
"styleRecalcDurationMs": 4.736000000000001,
"layouts": 11,
"layoutDurationMs": 2.7070000000000003,
"taskDurationMs": 391.227,
"heapDeltaBytes": -205052,
"heapUsedBytes": 78428768,
"domNodes": 47,
"jsHeapTotalBytes": 7602176,
"scriptDurationMs": 10.595,
"eventListeners": 37,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665384615384614,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "canvas-zoom-sweep",
"durationMs": 1757.0959999999332,
"styleRecalcs": 18,
"styleRecalcDurationMs": 4.476999999999999,
"layouts": 12,
"layoutDurationMs": 2.767,
"taskDurationMs": 294.323,
"heapDeltaBytes": -188588,
"heapUsedBytes": 78526144,
"domNodes": 47,
"jsHeapTotalBytes": 7864320,
"scriptDurationMs": 10.123,
"eventListeners": 37,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66698113207547,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "dom-widget-clipping",
"durationMs": 594.0739999999778,
"styleRecalcs": 15,
"styleRecalcDurationMs": 2.778,
"layouts": 10,
"layoutDurationMs": 2.571,
"taskDurationMs": 432.805,
"heapDeltaBytes": 6243512,
"heapUsedBytes": 85467352,
"domNodes": 0,
"jsHeapTotalBytes": 7864320,
"scriptDurationMs": 174.54,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66388888888889,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "dom-widget-clipping",
"durationMs": 649.4720000000029,
"styleRecalcs": 14,
"styleRecalcDurationMs": 2.5149999999999997,
"layouts": 9,
"layoutDurationMs": 2.5380000000000003,
"taskDurationMs": 478.80999999999995,
"heapDeltaBytes": 6671912,
"heapUsedBytes": 85379816,
"domNodes": 0,
"jsHeapTotalBytes": 7340032,
"scriptDurationMs": 165.34300000000002,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.664102564102578,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "dom-widget-order-dense-n200-w200-i120",
"durationMs": 1156.8490000000224,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 832.2579999999999,
"heapDeltaBytes": 14293772,
"heapUsedBytes": 94947704,
"domNodes": 0,
"jsHeapTotalBytes": 18251776,
"scriptDurationMs": 65.63799999999999,
"eventListeners": 0,
"totalBlockingTimeMs": 518,
"frameDurationMs": 33.82058823529411,
"p95FrameDurationMs": 16.699999999999818
},
{
"name": "dom-widget-order-dense-n200-w200-i120",
"durationMs": 1129.2050000000131,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 818.3659999999999,
"heapDeltaBytes": 18741276,
"heapUsedBytes": 99050332,
"domNodes": 0,
"jsHeapTotalBytes": 19038208,
"scriptDurationMs": 63.207,
"eventListeners": 0,
"totalBlockingTimeMs": 507,
"frameDurationMs": 33.332352941176474,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "dom-widget-order-sparse-n1000-w20-i120",
"durationMs": 839.6729999999764,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 433.05499999999995,
"heapDeltaBytes": -4597452,
"heapUsedBytes": 74423772,
"domNodes": 0,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 153.619,
"eventListeners": 0,
"totalBlockingTimeMs": 211,
"frameDurationMs": 24.9970588235294,
"p95FrameDurationMs": 33.400000000000546
},
{
"name": "dom-widget-order-sparse-n1000-w20-i120",
"durationMs": 843.7009999999532,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 481.0120000000002,
"heapDeltaBytes": -4215484,
"heapUsedBytes": 73753420,
"domNodes": 0,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 156.221,
"eventListeners": 0,
"totalBlockingTimeMs": 213,
"frameDurationMs": 24.99705882352943,
"p95FrameDurationMs": 16.80000000000109
},
{
"name": "extension-clean-combined-draw",
"durationMs": 9811.225000000008,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 8964.053,
"heapDeltaBytes": -15189640,
"heapUsedBytes": 68787616,
"domNodes": 0,
"jsHeapTotalBytes": 4194304,
"scriptDurationMs": 7250.598,
"eventListeners": 0,
"totalBlockingTimeMs": 709,
"frameDurationMs": 31.665483870967744,
"p95FrameDurationMs": 50.100000000000364
},
{
"name": "extension-clean-combined-draw",
"durationMs": 10478.673000000015,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 9730.926,
"heapDeltaBytes": 13942380,
"heapUsedBytes": 97962884,
"domNodes": 0,
"jsHeapTotalBytes": 7077888,
"scriptDurationMs": 7953.347999999999,
"eventListeners": 0,
"totalBlockingTimeMs": 1056,
"frameDurationMs": 31.430030030030025,
"p95FrameDurationMs": 50.100000000000364
},
{
"name": "extension-rgthree-prototype-draw",
"durationMs": 10309.123999999996,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 9505.358999999999,
"heapDeltaBytes": 12249540,
"heapUsedBytes": 95297540,
"domNodes": 0,
"jsHeapTotalBytes": 3670016,
"scriptDurationMs": 7750.038,
"eventListeners": 0,
"totalBlockingTimeMs": 960,
"frameDurationMs": 31.21090909090909,
"p95FrameDurationMs": 50.100000000000364
},
{
"name": "extension-rgthree-prototype-draw",
"durationMs": 9710.381999999981,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 8877.325,
"heapDeltaBytes": -2712588,
"heapUsedBytes": 80914456,
"domNodes": 0,
"jsHeapTotalBytes": 2883584,
"scriptDurationMs": 7183.316,
"eventListeners": 0,
"totalBlockingTimeMs": 730,
"frameDurationMs": 31.042172523961664,
"p95FrameDurationMs": 50.100000000000364
},
{
"name": "large-combo-open",
"durationMs": 200.81099999998742,
"styleRecalcs": 22,
"styleRecalcDurationMs": 12.096,
"layouts": 9,
"layoutDurationMs": 2.347,
"taskDurationMs": 152.555,
"heapDeltaBytes": 13306288,
"heapUsedBytes": 77524892,
"domNodes": 917,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 8.393,
"eventListeners": 196,
"totalBlockingTimeMs": 0,
"frameDurationMs": 21.680000000000017,
"p95FrameDurationMs": 66.69999999999982
},
{
"name": "large-combo-open",
"durationMs": 211.07200000000148,
"styleRecalcs": 22,
"styleRecalcDurationMs": 11.945999999999998,
"layouts": 8,
"layoutDurationMs": 2.3879999999999995,
"taskDurationMs": 159.19,
"heapDeltaBytes": 13311476,
"heapUsedBytes": 78048400,
"domNodes": 917,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 8.331999999999999,
"eventListeners": 196,
"totalBlockingTimeMs": 0,
"frameDurationMs": 21.66999999999998,
"p95FrameDurationMs": 66.59999999999945
},
{
"name": "large-graph-idle",
"durationMs": 2034.1000000000236,
"styleRecalcs": 8,
"styleRecalcDurationMs": 1.8419999999999999,
"layouts": 8,
"layoutDurationMs": 1.8060000000000003,
"taskDurationMs": 601.009,
"heapDeltaBytes": -23897252,
"heapUsedBytes": 75006684,
"domNodes": -258,
"jsHeapTotalBytes": -225280,
"scriptDurationMs": 36.47,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66557377049181,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "large-graph-idle",
"durationMs": 2042.6860000000033,
"styleRecalcs": 7,
"styleRecalcDurationMs": 1.5370000000000001,
"layouts": 7,
"layoutDurationMs": 1.633,
"taskDurationMs": 579.685,
"heapDeltaBytes": -25496736,
"heapUsedBytes": 74076344,
"domNodes": -278,
"jsHeapTotalBytes": -225280,
"scriptDurationMs": 32.763000000000005,
"eventListeners": -68,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665040650406507,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "large-graph-pan",
"durationMs": 2648.116000000016,
"styleRecalcs": 66,
"styleRecalcDurationMs": 4.855999999999999,
"layouts": 5,
"layoutDurationMs": 1.1650000000000003,
"taskDurationMs": 1586.6280000000002,
"heapDeltaBytes": 14706964,
"heapUsedBytes": 80810144,
"domNodes": 0,
"jsHeapTotalBytes": 225280,
"scriptDurationMs": 786.554,
"eventListeners": 2,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.772611464968154,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "large-graph-pan",
"durationMs": 2154.8360000000457,
"styleRecalcs": 63,
"styleRecalcDurationMs": 4.537,
"layouts": 2,
"layoutDurationMs": 0.3769999999999997,
"taskDurationMs": 1383.722,
"heapDeltaBytes": 14835564,
"heapUsedBytes": 80493628,
"domNodes": 0,
"jsHeapTotalBytes": 225280,
"scriptDurationMs": 796.9920000000001,
"eventListeners": 2,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.796093749999997,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "large-graph-zoom",
"durationMs": 3620.22300000001,
"styleRecalcs": 68,
"styleRecalcDurationMs": 5.824,
"layouts": 61,
"layoutDurationMs": 9.241,
"taskDurationMs": 1663.8429999999998,
"heapDeltaBytes": 3815232,
"heapUsedBytes": 70838320,
"domNodes": 24,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 748.0110000000001,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.66589861751152,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "large-graph-zoom",
"durationMs": 3628.2899999999927,
"styleRecalcs": 69,
"styleRecalcDurationMs": 6.013,
"layouts": 62,
"layoutDurationMs": 9.376000000000001,
"taskDurationMs": 1688.94,
"heapDeltaBytes": 489024,
"heapUsedBytes": 66692616,
"domNodes": 24,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 750.923,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665740740740738,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "legacy-node-drag",
"durationMs": 2948.44599999999,
"styleRecalcs": 42,
"styleRecalcDurationMs": 3.7479999999999998,
"layouts": 2,
"layoutDurationMs": 0.5379999999999994,
"taskDurationMs": 1895.349,
"heapDeltaBytes": 8492072,
"heapUsedBytes": 76016756,
"domNodes": 0,
"jsHeapTotalBytes": 3932160,
"scriptDurationMs": 935.0099999999999,
"eventListeners": 182,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.953448275862065,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "legacy-node-drag",
"durationMs": 2681.4219999999978,
"styleRecalcs": 41,
"styleRecalcDurationMs": 3.6839999999999997,
"layouts": 2,
"layoutDurationMs": 0.392,
"taskDurationMs": 1692.377,
"heapDeltaBytes": 8215680,
"heapUsedBytes": 77245152,
"domNodes": 0,
"jsHeapTotalBytes": 3108864,
"scriptDurationMs": 946.394,
"eventListeners": 182,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.77106918238994,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "minimap-idle",
"durationMs": 2022.9590000000144,
"styleRecalcs": 4,
"styleRecalcDurationMs": 0.7039999999999997,
"layouts": 3,
"layoutDurationMs": 0.6439999999999996,
"taskDurationMs": 548.055,
"heapDeltaBytes": 10481296,
"heapUsedBytes": 76192860,
"domNodes": 0,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 30.635999999999996,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665289256198346,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "minimap-idle",
"durationMs": 2035.9260000000177,
"styleRecalcs": 4,
"styleRecalcDurationMs": 1.1199999999999994,
"layouts": 4,
"layoutDurationMs": 1.0190000000000003,
"taskDurationMs": 592.318,
"heapDeltaBytes": -26643084,
"heapUsedBytes": 75033712,
"domNodes": -258,
"jsHeapTotalBytes": -225280,
"scriptDurationMs": 30.54799999999999,
"eventListeners": -67,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665289256198346,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-dom-widget-clipping",
"durationMs": 609.0019999999754,
"styleRecalcs": 48,
"styleRecalcDurationMs": 4.705,
"layouts": 10,
"layoutDurationMs": 2.623,
"taskDurationMs": 429.24300000000005,
"heapDeltaBytes": 1297940,
"heapUsedBytes": 81473836,
"domNodes": 0,
"jsHeapTotalBytes": 8388608,
"scriptDurationMs": 149.80500000000004,
"eventListeners": 46,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.667567567567563,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "subgraph-dom-widget-clipping",
"durationMs": 571.4659999999867,
"styleRecalcs": 47,
"styleRecalcDurationMs": 4.5840000000000005,
"layouts": 9,
"layoutDurationMs": 2.5839999999999996,
"taskDurationMs": 424.68600000000004,
"heapDeltaBytes": 20786424,
"heapUsedBytes": 85845360,
"domNodes": 0,
"jsHeapTotalBytes": 3670016,
"scriptDurationMs": 143.637,
"eventListeners": 46,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.667647058823523,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "subgraph-idle",
"durationMs": 2022.278999999969,
"styleRecalcs": 11,
"styleRecalcDurationMs": 2.6670000000000003,
"layouts": 11,
"layoutDurationMs": 2.8199999999999994,
"taskDurationMs": 545.609,
"heapDeltaBytes": -17122216,
"heapUsedBytes": 63538584,
"domNodes": -180,
"jsHeapTotalBytes": 6815744,
"scriptDurationMs": 8.197000000000001,
"eventListeners": -114,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665289256198342,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-idle",
"durationMs": 2023.665000000051,
"styleRecalcs": 10,
"styleRecalcDurationMs": 2.404,
"layouts": 9,
"layoutDurationMs": 2.4999999999999996,
"taskDurationMs": 576.613,
"heapDeltaBytes": -5297884,
"heapUsedBytes": 60216668,
"domNodes": -180,
"jsHeapTotalBytes": 7340032,
"scriptDurationMs": 8.143,
"eventListeners": -114,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.666942148760327,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-mouse-sweep",
"durationMs": 1725.481000000002,
"styleRecalcs": 77,
"styleRecalcDurationMs": 12.268999999999998,
"layouts": 9,
"layoutDurationMs": 2.3289999999999993,
"taskDurationMs": 838.031,
"heapDeltaBytes": -2005460,
"heapUsedBytes": 78649208,
"domNodes": -180,
"jsHeapTotalBytes": 8650752,
"scriptDurationMs": 106.047,
"eventListeners": -114,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665048543689323,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "subgraph-mouse-sweep",
"durationMs": 1976.9469999999956,
"styleRecalcs": 86,
"styleRecalcDurationMs": 14.926000000000002,
"layouts": 11,
"layoutDurationMs": 3.025,
"taskDurationMs": 958.344,
"heapDeltaBytes": 938104,
"heapUsedBytes": 81698100,
"domNodes": -180,
"jsHeapTotalBytes": 8388608,
"scriptDurationMs": 109.48,
"eventListeners": -114,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665546218487393,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "subgraph-transition-enter",
"durationMs": 1435.116999999991,
"styleRecalcs": 19,
"styleRecalcDurationMs": 25.653000000000002,
"layouts": 5,
"layoutDurationMs": 15.954999999999997,
"taskDurationMs": 1006.6590000000002,
"heapDeltaBytes": 2505708,
"heapUsedBytes": 100405816,
"domNodes": 15950,
"jsHeapTotalBytes": 16392192,
"scriptDurationMs": 28.185999999999996,
"eventListeners": 3244,
"totalBlockingTimeMs": 155,
"frameDurationMs": 19.331999999999994,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "viewport-pan-sweep",
"durationMs": 8272.981000000016,
"styleRecalcs": 247,
"styleRecalcDurationMs": 16.578,
"layouts": 5,
"layoutDurationMs": 1.218,
"taskDurationMs": 5130.563999999999,
"heapDeltaBytes": 36788596,
"heapUsedBytes": 101389684,
"domNodes": -258,
"jsHeapTotalBytes": 487424,
"scriptDurationMs": 2764.592,
"eventListeners": -66,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.69979797979798,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "viewport-pan-sweep",
"durationMs": 8965.667000000052,
"styleRecalcs": 248,
"styleRecalcDurationMs": 16.602999999999998,
"layouts": 5,
"layoutDurationMs": 1.2429999999999997,
"taskDurationMs": 5705.598,
"heapDeltaBytes": 2464116,
"heapUsedBytes": 66650040,
"domNodes": -258,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 2629.73,
"eventListeners": -66,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.697020484171322,
"p95FrameDurationMs": 16.799999999999272
},
{
"name": "vue-large-graph-idle",
"durationMs": 16365.845000000036,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 15370.502000000002,
"heapDeltaBytes": 4118012,
"heapUsedBytes": 283503532,
"domNodes": -3330,
"jsHeapTotalBytes": 12570624,
"scriptDurationMs": 248.325,
"eventListeners": -653,
"totalBlockingTimeMs": 0,
"frameDurationMs": 28.217241379310344,
"p95FrameDurationMs": 50.100000000000364
},
{
"name": "vue-large-graph-idle",
"durationMs": 22716.990999999893,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 21799.473,
"heapDeltaBytes": -15105292,
"heapUsedBytes": 283187480,
"domNodes": -3330,
"jsHeapTotalBytes": 12046336,
"scriptDurationMs": 308.002,
"eventListeners": -653,
"totalBlockingTimeMs": 0,
"frameDurationMs": 39.03058419243986,
"p95FrameDurationMs": 83.29999999999927
},
{
"name": "vue-large-graph-pan",
"durationMs": 18798.730999999974,
"styleRecalcs": 162,
"styleRecalcDurationMs": 12.185000000000002,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 18424.631999999998,
"heapDeltaBytes": -21625644,
"heapUsedBytes": 267551464,
"domNodes": -3330,
"jsHeapTotalBytes": 14667776,
"scriptDurationMs": 857,
"eventListeners": -651,
"totalBlockingTimeMs": 9,
"frameDurationMs": 33.750987432675046,
"p95FrameDurationMs": 66.70000000000073
},
{
"name": "vue-large-graph-pan",
"durationMs": 27276.66199999999,
"styleRecalcs": 181,
"styleRecalcDurationMs": 15.22699999999999,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 26944.355,
"heapDeltaBytes": -23138944,
"heapUsedBytes": 274537300,
"domNodes": -3330,
"jsHeapTotalBytes": 8900608,
"scriptDurationMs": 897.785,
"eventListeners": -651,
"totalBlockingTimeMs": 11,
"frameDurationMs": 50.8050279329609,
"p95FrameDurationMs": 100.09999999999854
},
{
"name": "vue-node-resize-workload",
"durationMs": 5577.142999999978,
"styleRecalcs": 20,
"styleRecalcDurationMs": 69.33900000000001,
"layouts": 20,
"layoutDurationMs": 22.44600000000001,
"taskDurationMs": 5433.139,
"heapDeltaBytes": 28884064,
"heapUsedBytes": 289482908,
"domNodes": 0,
"jsHeapTotalBytes": 0,
"scriptDurationMs": 90.45699999999998,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 33.94146341463414,
"p95FrameDurationMs": 66.70000000000073
},
{
"name": "vue-node-resize-workload",
"durationMs": 8042.741999999976,
"styleRecalcs": 20,
"styleRecalcDurationMs": 70.025,
"layouts": 20,
"layoutDurationMs": 21.487999999999992,
"taskDurationMs": 7937.629,
"heapDeltaBytes": -34292212,
"heapUsedBytes": 255392552,
"domNodes": -3330,
"jsHeapTotalBytes": 20434944,
"scriptDurationMs": 125.39900000000002,
"eventListeners": -653,
"totalBlockingTimeMs": 0,
"frameDurationMs": 47.35117647058824,
"p95FrameDurationMs": 100
},
{
"name": "vue-zoom-out-idle",
"durationMs": 11448.531000000003,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 10931.840999999999,
"heapDeltaBytes": 16436428,
"heapUsedBytes": 283603532,
"domNodes": 0,
"jsHeapTotalBytes": -20185088,
"scriptDurationMs": 151.005,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 39.21095890410958,
"p95FrameDurationMs": 83.29999999999927
},
{
"name": "vue-zoom-out-idle",
"durationMs": 11387.112000000001,
"styleRecalcs": 0,
"styleRecalcDurationMs": 0,
"layouts": 0,
"layoutDurationMs": 0,
"taskDurationMs": 10938.835000000001,
"heapDeltaBytes": 13222612,
"heapUsedBytes": 263698944,
"domNodes": 0,
"jsHeapTotalBytes": 3407872,
"scriptDurationMs": 157.69000000000005,
"eventListeners": 0,
"totalBlockingTimeMs": 0,
"frameDurationMs": 40.22190812720848,
"p95FrameDurationMs": 83.29999999999927
},
{
"name": "workflow-execution",
"durationMs": 482.1550000000343,
"styleRecalcs": 22,
"styleRecalcDurationMs": 12.209000000000001,
"layouts": 12,
"layoutDurationMs": 4.159999999999999,
"taskDurationMs": 144.754,
"heapDeltaBytes": -12879496,
"heapUsedBytes": 63424020,
"domNodes": 126,
"jsHeapTotalBytes": 5242880,
"scriptDurationMs": 15.427000000000001,
"eventListeners": 1,
"totalBlockingTimeMs": 0,
"frameDurationMs": 17.260714285714293,
"p95FrameDurationMs": 16.800000000000182
},
{
"name": "workflow-execution",
"durationMs": 117.34500000000025,
"styleRecalcs": 7,
"styleRecalcDurationMs": 8.042999999999997,
"layouts": 7,
"layoutDurationMs": 2.988,
"taskDurationMs": 91.08000000000001,
"heapDeltaBytes": -11103928,
"heapUsedBytes": 64859676,
"domNodes": 117,
"jsHeapTotalBytes": 10223616,
"scriptDurationMs": 6.650000000000001,
"eventListeners": 1,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.671428571428546,
"p95FrameDurationMs": 16.700000000000273
},
{
"name": "agent-composer-empty-conversation-typing",
"durationMs": 456.80700000002616,
"styleRecalcs": 5,
"styleRecalcDurationMs": 0.7240000000000001,
"layouts": 76,
"layoutDurationMs": 14.257000000000001,
"taskDurationMs": 405.01,
"heapDeltaBytes": 11540984,
"heapUsedBytes": 99876196,
"domNodes": 4,
"jsHeapTotalBytes": 12058624,
"scriptDurationMs": 197.64200000000002,
"eventListeners": 292,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.667857142857137,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "agent-composer-empty-conversation-typing",
"durationMs": 515.1519999999437,
"styleRecalcs": 5,
"styleRecalcDurationMs": 0.962,
"layouts": 76,
"layoutDurationMs": 15.28,
"taskDurationMs": 477.60200000000003,
"heapDeltaBytes": 12307444,
"heapUsedBytes": 100080160,
"domNodes": 4,
"jsHeapTotalBytes": 12320768,
"scriptDurationMs": 238.019,
"eventListeners": 292,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.665624999999977,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "agent-composer-long-conversation-typing",
"durationMs": 473.9280000000008,
"styleRecalcs": 5,
"styleRecalcDurationMs": 1.736,
"layouts": 76,
"layoutDurationMs": 13.440999999999999,
"taskDurationMs": 435.404,
"heapDeltaBytes": 114696,
"heapUsedBytes": 120754868,
"domNodes": 4,
"jsHeapTotalBytes": 524288,
"scriptDurationMs": 172.743,
"eventListeners": 292,
"totalBlockingTimeMs": 0,
"frameDurationMs": 16.668965517241396,
"p95FrameDurationMs": 16.700000000000728
},
{
"name": "agent-composer-long-conversation-typing",
"durationMs": 631.183999999962,
"styleRecalcs": 5,
"styleRecalcDurationMs": 2.198,
"layouts": 76,
"layoutDurationMs": 14.363000000000001,
"taskDurationMs": 584.119,
"heapDeltaBytes": 536924,
"heapUsedBytes": 118096364,
"domNodes": 4,
"jsHeapTotalBytes": 262144,
"scriptDurationMs": 223.77000000000004,
"eventListeners": 292,
"totalBlockingTimeMs": 0,
"frameDurationMs": 17.118918918918908,
"p95FrameDurationMs": 16.800000000000182
}
]
}🔬 E2E Coverage
Top 10 files by uncovered lines
🎭 Playwright: ✅ 2283 passed, 0 failed📊 Browser Reports
🎨 Storybook: ✅ Built — View Storybook🧩 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 🌾 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 61eca01a973ae0545243f7d084f7156c930add5b
Full docs: ADR-DEVEX-LINT-0015 |
Codecov Report✅ All modified and coverable lines are covered by tests. @@ Coverage Diff @@
## main #20092 +/- ##
==========================================
+ Coverage 86.49% 87.29% +0.80%
==========================================
Files 3731 2657 -1074
Lines 176878 148293 -28585
Branches 46813 38660 -8153
==========================================
- Hits 152984 129459 -23525
+ Misses 21833 18444 -3389
+ Partials 2061 390 -1671
Flags with carried forward coverage won't be shown. Click here to find out more.
... and 1082 files with indirect coverage changes 🚀 New features to boost your workflow:
|
|
@coderabbitai review |
✅ Action performedReview finished.
|
A resubscribe can replay an agent_ask the store has already retired. A run-approval card or the unrenderable-ask notice is removed on retirement, so the replay put it back. The store now drops an agent_ask whose id it has retired for that thread before routing it.
…' into kishore/agent-delete-approval-card
|
Independent exact-head review at
Focused exact-head Vitest is green: 145/145 across AgentMessage, message grouping, message parts, and REST client. I did not push or rebase this teammate-owned stack, and I left the existing Dante assignment unchanged. |
dante01yoon
left a comment
There was a problem hiding this comment.
Thanks, the card is small and follows RunApprovalCard well: the exact delete/keep, single-required, no-free-text check with fallback to the notice, unique/capped node list with an "N more nodes not listed" count (plural via vue-i18n), stable keys, i18n under agent.deleteApproval.*, and tests for parsing, bounds, dedupe, in-flight and resolved states. Checks were green on this head. I did not run vitest locally (no node_modules in my review worktree).
Requesting changes because the contract is again frontend-first: delete_approval is not in the generated AgentPendingAsk['kind'], the cloud spec change is the draft Comfy-Org/cloud#11870, and context.nodes has no schema in that spec at all, so zAgentAskNodeRef is a hand-written shape for server data with nothing to generate it from. Also one issue about a Delete card that lists no nodes, and a suggestion. The settledAskResolution concern on #19981 (client deciding which answer won) carries over to the "You chose to delete/keep them" readback here.
| * One `context.nodes` entry of a `delete_approval` ask. Parsed per entry, so a | ||
| * bad entry is counted as unlisted instead of dropping the whole card. | ||
| */ | ||
| export const zAgentAskNodeRef = z |
There was a problem hiding this comment.
issue: This declares the shape of server data (context.nodes entries of a delete_approval ask) locally. Could we get it into the spec so it can be imported from @comfyorg/ingest-types instead? Right now there is nothing to import: the generated AgentPendingAsk['kind'] is 'ask_user' | 'run_approval', and Comfy-Org/cloud#11870 (still a draft) adds delete_approval to the enum but leaves context as a free-form object, so id: string | number, type, title exist only in this file. Hand-declared copies drift from the real contract silently; in #14771 that produced two real bugs.
Same contract-first point as on #19981: could the delete_approval context (including the node ref shape) land in services/ingest/openapi.yaml first, then this card build on the generated type after the ingest-types sync? The card is inert until the backend flag is on anyway, so waiting for the spec shouldn't cost anything.
There was a problem hiding this comment.
Agreed. The spec change (a named ask kind enum, a schema for the delete context, and the server reporting the committed answer on the resolved frame) lands on the backend first. Then this PR switches to the generated types and drops the local shapes. It won't merge before that.
There was a problem hiding this comment.
Done, now that #20448 carries the spec. In 039f17e the local zAgentAskNodeRef is gone: the delete card parses context with the generated zAgentDeleteApprovalContext (node ids are strings, nodes is 1 to 50, hidden_node_count is read), delete_approval comes from the generated AgentAskKind that RENDERED_ASK_KINDS now satisfies, and a context the generated schema rejects falls back to the askUnavailable notice. The e2e delete frame is typed with satisfies AgentPendingDeleteApproval & { thread_id: string } in c001dd2.
| context: AskInput['context'] | ||
| ): Pick<DeleteApprovalPart, 'nodes' | 'hiddenNodeCount'> { | ||
| const raw: unknown = context?.nodes | ||
| if (!Array.isArray(raw)) return { nodes: [], hiddenNodeCount: 0 } |
There was a problem hiding this comment.
issue: When context.nodes is missing or not an array, this returns { nodes: [], hiddenNodeCount: 0 }, so the card shows an armed Delete button with no node list and no "N more nodes not listed" line (pinned by 'still asks when context.nodes is malformed, without a node list'). That breaks the guarantee the hiddenNodeCount doc states, that a Delete answer never covers nodes the user was not told about: here the user approves a destructive action against a set they were never shown, based only on the agent's prompt text. Could a delete ask without a readable node list fall back to the existing askUnavailable notice (the way a wrong option set does), or at least show a line saying the nodes could not be listed? The server re-checking ids protects against a mismatch, but not against the user approving blind.
There was a problem hiding this comment.
Fixed in 8016e81. A delete ask whose node list is missing, not an array, empty, or has no readable node now falls back to the askUnavailable notice, so Delete is never armed without the nodes it would remove. The pinned test now asserts the notice.
| const decision = computed(() => { | ||
| const resolution = part.resolution | ||
| if (!resolution) return undefined | ||
| if (!resolution.answered) return t('agent.deleteApproval.closed') |
There was a problem hiding this comment.
suggestion: Through the shared retirement default from #19981, the 5xx-after-retries path in useAgentSession.answerAsk retires this card with { answered: false }, so it reads "This request was closed without an answer" right next to a toast saying the answer may still have been applied. For a delete that is the worst place to be ambiguous: the nodes may already be gone. Could this path show an "outcome unknown, reload to check" state instead of "closed"?
…t is unknown - Resolutions carry a status: answered, closed or unknown. The card names the committed answer only as far as the agent_ask_resolved frame reports it. When the frame names no selection, or one that differs from what this client sent, it reads as answered without naming either, and this client's free text is never echoed as the committed answer. settledAskResolution takes the frame's answer as an object so a server-reported other_text drops in later. - A 202 confirms the ask was answered but not with what, so it no longer shows our selection. - After retries are spent on a 5xx, the card is retired as unknown and reads "We couldn't confirm your answer was received", matching the toast, instead of "closed without an answer". An unknown outcome gives way to a real answer but not to a bare closure.
…e is unknown - A delete_approval whose context.nodes is missing, not an array, empty, or has no readable node falls back to the askUnavailable notice, the same as an unknown kind, so a Delete answer can never cover nodes the user was not shown. - The card reads back only what the server reported: delete, keep, or just "Answered." when the frame named no selection. After retries are spent on a 5xx it says "We couldn't confirm your answer. Check the canvas before trying again." instead of "closed without an answer".
Main's new answerAsk test passes a bare id list; it now passes the AgentAnswerRequest body this branch types answerAsk with.
# Conflicts: # browser_tests/fixtures/data/agent/agentAskFrames.ts # browser_tests/tests/agent/agentRunApprovalConsent.spec.ts # src/locales/en/main.json # src/workbench/extensions/agent/components/agent/message/AgentMessage.test.ts # src/workbench/extensions/agent/components/agent/message/AgentMessageGroup.vue # src/workbench/extensions/agent/components/agent/message/agentMessageGroup.ts # src/workbench/extensions/agent/schemas/agentApiSchema.ts # src/workbench/extensions/agent/services/agent/agentMessageParts.test.ts # src/workbench/extensions/agent/services/agent/agentMessageParts.ts # src/workbench/extensions/agent/services/agent/agentRestClient.test.ts # src/workbench/extensions/agent/stores/agent/agentConversationStore.test.ts
…-Org#19981) ## Summary Renders the agent's `ask_user` question as a card on main, and sends `ask_kinds` on each turn so the server parks a turn only on an ask this panel can show. `paused` is the exception: the server raises it for every client, and it gets the notice below. Supersedes Comfy-Org#18298. The delete approval card is Comfy-Org#20092, stacked on this PR. - `AskUserCard.vue` shows the prompt and the options. A required single choice is radios (`AskUserRadioOptions`). Anything else is checkboxes capped at `max_selections` (`AskUserCheckboxOptions`) with a hint. With `allow_other`, an Other box counts as one more selection, matching the server's `validateSelection`. The parser dedupes option ids, caps options at 50, clips long text and caps `max_selections` at what can be chosen. An ask with nothing to choose, or with bounds the card can't satisfy (a minimum above what it can offer, or a maximum below one), gets the notice instead of a card that could only draw 422s. - The answer goes through `answerAsk` (`POST /agent/threads/:id/asks/:ask_id/answer`) with the generated `AgentAnswerRequest`. Outcomes by response: - 202: the card shows the answer the receipt committed. A status-only 202 reads as "Answered" and leaves the comparison to the frame. - 409: the card shows the stored outcome from `AgentAskConflict`, or "no longer open here" when the body has none. - 422: nothing was committed, so the card goes back into service and the error overlay asks the user to change their choices or stop the turn. - 403/404/501: refusals, so the card is retired. - Any other failure, 400 included, reads as unconfirmed. - When a committed answer from the receipt, the 409 body or the frame differs from what this tab sent, in ids or in free text, the user gets one "answered somewhere else" toast. - A resolved card stays read-only and shows what the server committed. A cancelled or expired ask reads "closed without an answer". - Types come from `@comfyorg/ingest-types`. `AskUserResolution` is a discriminated union, and only `answered` carries `selected`/`otherText`. - Parse policy (differs from draft Comfy-Org#20891): `kind` is `AgentAskKind | 'unknown'`, via `.catch('unknown')`. A kind this client predates, or `paused`, shows the `askUnavailable` notice instead of failing the frame or the transcript parse. The answer receipt keeps `selected` optional. Comfy-Org#20891 switches both to the strict generated schemas; whichever PR merges second has to pick one policy. - Coordination: Comfy-Org#16949 and Comfy-Org#18949 each add their own `handleAnswerFailure`, and this PR adds a third. The three agree on 400 (unconfirmed). On the 202, this PR treats a receipt that carries `selected` as the committed answer and defers a status-only 202 to the frame. Comfy-Org#18949's notice for a status-only 202 on a recovered card can sit alongside that. This PR's `retireAsk(askId, resolution, owner?)` takes a required resolution, and `answerAsk` takes an `AgentAnswerRequest`. Whichever lands second merges the three handlers and updates those call sites. Comfy-Org#18949 also restores only `run_approval` asks and should widen recovery to every pending ask kind. - Boundaries: main now classifies every `src` file (Comfy-Org#20802), so the card's `Button`, `Input`, `Checkbox` and `RadioGroup` imports cross from `agent` into `design-system`. This PR approves `agent` as a `design-system` consumer in both records and regenerates `boundaries.json`. It is the first dependency approval in the catalog, so it needs a design-system owner's review. - Known limitation: the server keeps `pending_ask` only while the ask is pending, so a resolved card disappears when the thread is loaded again (a reload, or opening another chat and coming back). The agent README notes this. ## Verification - Vitest: `pnpm exec vitest run src/workbench/extensions/agent`, 163 files, 3,416 tests. Each new test fails under the mutation it targets, for example: the watchdog dropping the 202 receipt on a live turn, a 422 keeping the rejected answer, the old empty-answer guard, reporting a status-only 202, 400 as a refusal, a one-way id comparison, clamping unsatisfiable bounds and the single optional choice hint. - Playwright: `agentRunApprovalConsent.spec.ts` (consent) and `agentAskUserQuestion.spec.ts` (question). The question spec posts once to its ask and reads back the committed answer, and screenshots each card state (radio, in flight, closed, retired, unconfirmed) plus the notice. The eight Linux baselines moved with the spec. - `pnpm typecheck`, `pnpm typecheck:browser`, oxlint, eslint, oxfmt and knip are clean. --------- Co-authored-by: github-actions <github-actions@github.com>
|
@dante01yoon this was approved before the latest push, which only merged main after #19981 landed and cleaned up commit messages. The conflicts were with the ask-card changes from #19981; the delete-approval strings are unchanged. No other changes. Could you re-approve? |
There was a problem hiding this comment.
Verdict: ready for human review
This head is 2a6c8becd3, now against main after #19981 landed. One new item needs a disposition.
skishore23's re-approval request: confirmed. The push changes commit messages and adds merges, and nothing else in this PR.
- The force-push rewrote the history, and the rewritten
5415df5bd6has the same tree (f8e599453f) as the head I last reviewed,c823e90b2b. e5419ea2d1merges the final #19981 and2a6c8becd3mergesmain. The PR's diff against its merge base withmainhas the same 14 files and the same hunks as atc823e90b2b; only blob ids and line offsets inmain.jsonandagentApiSchema.tsdiffer.- CI at this head: 111 success, 18 skipped, no failures.
The push dismissed dante01yoon's approval. It was given on the identical tree, so a re-approval needs no new review from him.
The rewrite also addresses my second P3 note from round 8. None of the 80 commits now carries a Claude co-author trailer, so no squash-message instruction is needed.
One non-blocking note remains:
- P3 The description is out of date. "Stacked on #19981" no longer applies now that #19981 has landed and the base is
main. The Verification section still saysagentAskConsent.spec.ts"now covers every ask card" and that "All six cases pass". At this head the file has four tests (three run approval, one delete approval), and the ask_user cases are in #19981'sagentAskUserQuestion.spec.ts.
Review record
Inline findings, rounds 1 and 2:
- P1 This drops the server's
hidden_node_count(round 1): fixed in039f17e400; reply - P2 No fixture sends
hidden_node_count, so nothing fails on the defect intoAskNodeRefs(round 1): fixed in039f17e400; reply - P3 Each term of the hidden count has its own case (duplicates, unreadable, over the cap), but no case... (round 1): superseded: the client no longer adds count terms; the readback sum it moved to was fixed in round 2; reply
- P3 The
optionIds.length === DELETE_APPROVAL_OPTIONS.lengthguard has no falsifier: removing it leaves every... (round 1): partial in round 2 (subset row missing), completed in1e19e3611b; reply - P2 Nothing at unit level keeps a resolved delete card on screen (round 1): fixed in
039f17e400; reply - P3 With the
toAskPart(...).toBeDefined()assertion gone, this table iteratesRENDERED_ASK_KINDSand... (round 1): fixed in039f17e400; reply - P3 The blank-title fallback (
title?.trim() || type?.trim()) is only proven by the component test 'lists the... (round 1): fixed in039f17e400; reply - P3
toTurnId('msg-delete')would avoid a newas TurnIdassertion (docs/guidance/typescript.md treats... (round 1): fixed in26345c6f9c; reply - P3 The expected answer is derived as
button.toLowerCase(), which only works because the English label... (round 1): fixed in26345c6f9c; reply - P3 This repeats two tests that already fail for the same defect: the parser row 'nodes not an array' in... (round 1): fixed in
26345c6f9c; reply - P3 This file asserts the internal
{ kind, part }group shape that onlyAgentMessage.vuereads (round 1): fixed in26345c6f9c; reply - P3 The added q-3 ingest is never observed (round 1): fixed in
26345c6f9c; reply - P3 The contract's
AgentAskNodeRef.idis a string, and the generatedzAgentAskNodeRefrejects{ id: 12 }(round 1): fixed in039f17e400,c001dd26ed; reply - P3 the file is named for run approval and its header describes the run_approval contract, but it now holds... (round 1): fixed in
26345c6f9c,c001dd26ed; reply - P3 The keep iteration costs a second full page load on the same path as delete:
AgentMessageGroup.vuemaps... (round 1): fixed inc001dd26ed; reply - P3 No visual baseline covers the new card (destructive button, the scrolling list, truncated names and ids) (round 1): declined in round 2, then added: element screenshot in
edc8757ec3with the Linux baseline fromf595ec7600; reply - P3
expect(answers()).toHaveLength(0)runs right after the card renders, with no flush, so a self-answer... (round 1): fixed inc001dd26ed; reply - P3 The step title ("locks the card") and the
held/releasenames already say this (round 1): fixed inc001dd26ed; reply - P3 The shared
answers()recorder decodes the path the same way without a comment, andisDeleteAnswer... (round 1): fixed inc001dd26ed; reply - P3 this is the third inline
agent_ask_resolvedbuilder in the file (round 1): fixed inc001dd26ed; reply - P3 The poll on line 504 already settled the count at one and checked the body, and line 524 removes the only... (round 1): fixed in
c001dd26ed; reply - P3 This describes
toAskNodeRefs's parsing policy, not the schema, and it is out of date: when every entry... (round 1): fixed in039f17e400; reply - P3 The interface and
toAskNodeRef(title?.trim() || type?.trim()) say this, and "as the host listed it"... (round 1): fixed in26345c6f9c; reply - P3
AskNodeRefshared a stem with the generatedAgentAskNodeRef(round 1): fixed in26345c6f9c,039f17e400; reply - P3 This restates the interface name (round 1): withdrawn in round 2: the base now gives every exported part type a one-line JSDoc, which #20823 allows; reply
- P3 At this head the invariant is false (the server's
hidden_node_countis ignored, seetoAskNodeRefs) (round 1): fixed in26345c6f9c; reply - P3 A copy of
AskUserPart's docblock with one word changed, and "decision" is not right for closed, retired... (round 1): fixed in26345c6f9c; reply - P3 The function name and
resolution === undefinedsay this (round 1): fixed in26345c6f9c; reply - P3 One rule ("every ask card except run approval keeps its resolution") is written two ways:... (round 1): fixed in
26345c6f9c; reply - P3
nodes: 50lands inASK_USER_LIMITS, whose docblock still says "anask_userframe", and... (round 1): partial in round 2 (docblock wording), fixed when the base trimmed the docblock and the PR renamed it toASK_LIMITS; reply - P3 This narrates lines 315-322, and "anything not listed counted as hidden" is incomplete while... (round 1): fixed in
039f17e400; reply - P3
offersDeleteOrKeepand the 'renders nothing for $name' table say this (round 1): fixed in26345c6f9c; reply - P3 the constant on line 325, a length check and
every/includesspell one comparison (round 1): fixed in26345c6f9c; reply - P3 This repeats the
isAskPartkind list over group kinds, and this PR had to edit both (round 1): fixed in26345c6f9c; reply - P3
AskUserAnswer.vuemaps the same statuses through a typedRecord<AskUserResolution['status'], string>,... (round 1): fixed in26345c6f9c; reply - P3 This shell class string now appears verbatim in
RunApprovalCard.vue,AskUserCard.vueand here, and the... (round 1): deferred by the author as a follow-up across all three ask cards; accepted; reply - P3
Buttonalready hasloading, which setsdisabledandaria-busy(round 1): deferred with the wrapper above; accepted (parity withRunApprovalCard); reply - P3
AskNodeListhas one consumer, all its strings areagent.deleteApproval.*, and the props rename the... (round 1): fixed in26345c6f9c; reply - P3
agent.deleteApproval.nodeId("#{id}") gives translators a key with no words, whileComposer.vuein... (round 1): fixed in26345c6f9c; reply - P3 the list keeps this accessible name after the user chose Keep, so a screen reader announces "Nodes to... (round 1): fixed in
26345c6f9c; reply - P3 "Answered: delete them." reads oddly for a one-node delete, and "Answered." has a period where the... (round 1): fixed in
26345c6f9c; reply - P3 the
: voidannotation is inferred (round 2): fixed in1e19e3611b; reply - P3 In dark theme
--component-node-backgroundand--secondary-backgroundare bothcharcoal-600, so this... (round 2): fixed in1e19e3611b; reply - P2 The fixture has one listed node and
hidden_node_count: 1, so passingpart.nodes.lengthtomoreNodes... (round 2): fixed in1e19e3611b; reply - P3 these tests build the part through the parser and then
assert(part.type === 'deleteApproval'), which... (round 2): fixed in1e19e3611b; reply - P2
toHaveTextContent(string)matches a substring, so "Answered: delete them" satisfies the 'an answer the... (round 2): fixed in1e19e3611b; reply - P2 The readback plural adds listed and unlisted nodes, but no readback fixture had unlisted nodes (round 2): fixed in
1e19e3611b; reply - P3 These four rows pin constraints of the cloud-owned
zAgentDeleteApprovalContext(string ids,.max(50),... (round 2): fixed in1e19e3611b; reply - P3 'a third option' catches a superset, but not a subset: with "every id is delete or keep" and no length... (round 2): fixed in
1e19e3611b; reply
Review-body findings:
- P1 No screenshots or screencast of the new card (round 1, round 2, round 3): resolved. Stories in
0c24e4cbeb; an element screenshot with a Linux baseline inedc8757ec3andf595ec7600; the image is embedded in the description at56ccdbf2e5. - P2 Description said a malformed node list still shows the buttons, and overstated what Delete covers (round 1): fixed in the rewritten body. The blank-id attribution (round 2) and the free-text condition were fixed at
56ccdbf2e5. - P3 #18949 narrows
dropResolvedAskstorun_approval(round 1): answered by the author and pinned by the refetch store test inc60d4f66b9, which still fails at4ba5ea3142ifretireAskPartskeeps onlyask_usercards. - P3 AI attribution in commits and the body footer (round 1): footer removed; commits since round 2 are clean. The earlier trailers were removed by the round 9 history rewrite.
- P3 Body not on the PR template; test list incomplete; readback claim outdated (round 1): fixed in the rewritten body. The spec description went stale again after this round's merges (P3 note 1).
- P3 Backend gating claim not checkable (round 1, round 4): fixed at
56ccdbf2e5. - P3 No Storybook story for the card (round 1): fixed in
0c24e4cbeb. - P3 Screenshot not embedded in the description (round 4): fixed at
56ccdbf2e5. - P3 #20891 strict ask parsing (round 2, round 4): addressed at
56ccdbf2e5. The body states this PR keeps #19981's tolerant policy, and the author left a coordination note on #20891. - P3 Comments added or left stale under #20823 (round 2): gone; the PR adds no comments.
- P3 Shared
AskCardwrapper and per-buttonloading(round 1): accepted as a follow-up. - P3 A 422 showed ask_user copy on the delete card (round 6): addressed by #19981's
e48eb49559. - Cleared in round 6: the delete e2e moved to #19981's shared
holdAnswerand dropped itswaitForRequestguard.answeringstays set until the resolution arrives, so the lock check does not depend on catching the POST mid-flight.
Existing reviewers: dante01yoon's contract-first finding, fixed in 039f17e400; his malformed-list, 5xx-wording, and answer-readback findings, addressed in round 1. christian-byrne's fail-closed list and black-box e2e requests: addressed; his exact-head review found nothing new. CodeRabbit's five findings: addressed in round 1.
-
christian-byrne's gate disposition (round 8): the no-behaviour-change point is confirmed and moot after dante01yoon's approval; the base overlap is confirmed, and a local merge with
dd3c8fb3e1is clean, with agent unit tests andvue-tscpassing; the three unresolved threads are contested as a blocker because all are outdated, answered, and owned by the approving reviewer. -
dante01yoon approved at
c823e90b2b(round 8), replacing his Oct 5 changes-requested review. -
skishore23's re-approval request (round 9): confirmed. The rewritten
5415df5bd6has the same tree asc823e90b2b, and the merges of #19981 andmainleave the PR's hunks unchanged. -
P3 Description asked for, but did not contain, a squash-message instruction to drop the Claude co-author trailers (rounds 6 to 8): addressed by the history rewrite at
2a6c8becd3; no commit carries the trailer. -
P3 Stale Verification section (rounds 7 to 9) and the stale "Stacked on #19981" line (round 9): open, non-blocking.
Checks at 2a6c8becd3: CI 111 success, 18 skipped, no failures. At c823e90b2b: CI 111 success, 19 skipped. At 4ba5ea3142, the four delete-card unit files passed locally (290 tests). playwright test --list shows four tests in agentAskConsent.spec.ts. Changing retireAskParts to keep only ask_user cards fails the parts retire test and the store refetch test.
|
Disposition for exact head Live gate analysis:
This pass stops here without an approval or changes-requested review. |
There was a problem hiding this comment.
Verdict: ready for human review
The head is unchanged at 2a6c8becd3. This round answers christian-byrne's gate disposition for this head. I checked each point.
- Confirmed: the four required checks (
test,lint-and-format,e2e-status,website-e2e) are green at this head. - Confirmed, and moot: the no-behaviour-change grant cannot apply to a feature PR, as I noted in round 8. This PR needs a normal approval, and the push dismissed dante01yoon's.
- Confirmed as a merge-gate state, not a code finding: the three unresolved threads are dante01yoon's Oct 4 threads. The author answered each with its fixing commit, and dante01yoon approved the same tree at
5415df5bd6. He can resolve them when he re-approves. - Confirmed, and not blocking:
mainchangedagentRestClient.test.tsin59965ddd23(#21169) after this head. That hunk adds oneerror mappingcase forAgentApiError.operation; the PR's hunk in the same file only updates theask_kindsrequest expectations. I mergedmainat59965ddd23into this head locally (not pushed). The merge has no conflicts, andvue-tsc --noEmitreports no errors on the merged tree. - On that merged tree,
vitest run src/workbench/extensions/agentpassed once and failed twice on one case:useAgentSession.test.ts(g18), "a history fetch that never answers is abandoned at the recovery deadline", withsignal.abortedalready true at 59,999 ms. That file passed five runs out of five on its own, and plainmainat59965ddd23, without this PR, failed the same case in both of two full-suite runs. It is a load-sensitive flake thatmainalready has, in a file outside this PR's diff.
One non-blocking note remains, unchanged from round 9:
- P3 The description is out of date. "Stacked on #19981" no longer applies now that the base is
main. The Verification section still saysagentAskConsent.spec.ts"now covers every ask card" and that "All six cases pass", but the file has four tests (three run approval, one delete approval), and the ask_user cases are in #19981'sagentAskUserQuestion.spec.ts.
Review record
Inline findings, rounds 1 and 2:
- P1 This drops the server's
hidden_node_count(round 1): fixed in039f17e400; reply - P2 No fixture sends
hidden_node_count, so nothing fails on the defect intoAskNodeRefs(round 1): fixed in039f17e400; reply - P3 Each term of the hidden count has its own case (duplicates, unreadable, over the cap), but no case... (round 1): superseded: the client no longer adds count terms; the readback sum it moved to was fixed in round 2; reply
- P3 The
optionIds.length === DELETE_APPROVAL_OPTIONS.lengthguard has no falsifier: removing it leaves every... (round 1): partial in round 2 (subset row missing), completed in1e19e3611b; reply - P2 Nothing at unit level keeps a resolved delete card on screen (round 1): fixed in
039f17e400; reply - P3 With the
toAskPart(...).toBeDefined()assertion gone, this table iteratesRENDERED_ASK_KINDSand... (round 1): fixed in039f17e400; reply - P3 The blank-title fallback (
title?.trim() || type?.trim()) is only proven by the component test 'lists the... (round 1): fixed in039f17e400; reply - P3
toTurnId('msg-delete')would avoid a newas TurnIdassertion (docs/guidance/typescript.md treats... (round 1): fixed in26345c6f9c; reply - P3 The expected answer is derived as
button.toLowerCase(), which only works because the English label... (round 1): fixed in26345c6f9c; reply - P3 This repeats two tests that already fail for the same defect: the parser row 'nodes not an array' in... (round 1): fixed in
26345c6f9c; reply - P3 This file asserts the internal
{ kind, part }group shape that onlyAgentMessage.vuereads (round 1): fixed in26345c6f9c; reply - P3 The added q-3 ingest is never observed (round 1): fixed in
26345c6f9c; reply - P3 The contract's
AgentAskNodeRef.idis a string, and the generatedzAgentAskNodeRefrejects{ id: 12 }(round 1): fixed in039f17e400,c001dd26ed; reply - P3 the file is named for run approval and its header describes the run_approval contract, but it now holds... (round 1): fixed in
26345c6f9c,c001dd26ed; reply - P3 The keep iteration costs a second full page load on the same path as delete:
AgentMessageGroup.vuemaps... (round 1): fixed inc001dd26ed; reply - P3 No visual baseline covers the new card (destructive button, the scrolling list, truncated names and ids) (round 1): declined in round 2, then added: element screenshot in
edc8757ec3with the Linux baseline fromf595ec7600; reply - P3
expect(answers()).toHaveLength(0)runs right after the card renders, with no flush, so a self-answer... (round 1): fixed inc001dd26ed; reply - P3 The step title ("locks the card") and the
held/releasenames already say this (round 1): fixed inc001dd26ed; reply - P3 The shared
answers()recorder decodes the path the same way without a comment, andisDeleteAnswer... (round 1): fixed inc001dd26ed; reply - P3 this is the third inline
agent_ask_resolvedbuilder in the file (round 1): fixed inc001dd26ed; reply - P3 The poll on line 504 already settled the count at one and checked the body, and line 524 removes the only... (round 1): fixed in
c001dd26ed; reply - P3 This describes
toAskNodeRefs's parsing policy, not the schema, and it is out of date: when every entry... (round 1): fixed in039f17e400; reply - P3 The interface and
toAskNodeRef(title?.trim() || type?.trim()) say this, and "as the host listed it"... (round 1): fixed in26345c6f9c; reply - P3
AskNodeRefshared a stem with the generatedAgentAskNodeRef(round 1): fixed in26345c6f9c,039f17e400; reply - P3 This restates the interface name (round 1): withdrawn in round 2: the base now gives every exported part type a one-line JSDoc, which #20823 allows; reply
- P3 At this head the invariant is false (the server's
hidden_node_countis ignored, seetoAskNodeRefs) (round 1): fixed in26345c6f9c; reply - P3 A copy of
AskUserPart's docblock with one word changed, and "decision" is not right for closed, retired... (round 1): fixed in26345c6f9c; reply - P3 The function name and
resolution === undefinedsay this (round 1): fixed in26345c6f9c; reply - P3 One rule ("every ask card except run approval keeps its resolution") is written two ways:... (round 1): fixed in
26345c6f9c; reply - P3
nodes: 50lands inASK_USER_LIMITS, whose docblock still says "anask_userframe", and... (round 1): partial in round 2 (docblock wording), fixed when the base trimmed the docblock and the PR renamed it toASK_LIMITS; reply - P3 This narrates lines 315-322, and "anything not listed counted as hidden" is incomplete while... (round 1): fixed in
039f17e400; reply - P3
offersDeleteOrKeepand the 'renders nothing for $name' table say this (round 1): fixed in26345c6f9c; reply - P3 the constant on line 325, a length check and
every/includesspell one comparison (round 1): fixed in26345c6f9c; reply - P3 This repeats the
isAskPartkind list over group kinds, and this PR had to edit both (round 1): fixed in26345c6f9c; reply - P3
AskUserAnswer.vuemaps the same statuses through a typedRecord<AskUserResolution['status'], string>,... (round 1): fixed in26345c6f9c; reply - P3 This shell class string now appears verbatim in
RunApprovalCard.vue,AskUserCard.vueand here, and the... (round 1): deferred by the author as a follow-up across all three ask cards; accepted; reply - P3
Buttonalready hasloading, which setsdisabledandaria-busy(round 1): deferred with the wrapper above; accepted (parity withRunApprovalCard); reply - P3
AskNodeListhas one consumer, all its strings areagent.deleteApproval.*, and the props rename the... (round 1): fixed in26345c6f9c; reply - P3
agent.deleteApproval.nodeId("#{id}") gives translators a key with no words, whileComposer.vuein... (round 1): fixed in26345c6f9c; reply - P3 the list keeps this accessible name after the user chose Keep, so a screen reader announces "Nodes to... (round 1): fixed in
26345c6f9c; reply - P3 "Answered: delete them." reads oddly for a one-node delete, and "Answered." has a period where the... (round 1): fixed in
26345c6f9c; reply - P3 the
: voidannotation is inferred (round 2): fixed in1e19e3611b; reply - P3 In dark theme
--component-node-backgroundand--secondary-backgroundare bothcharcoal-600, so this... (round 2): fixed in1e19e3611b; reply - P2 The fixture has one listed node and
hidden_node_count: 1, so passingpart.nodes.lengthtomoreNodes... (round 2): fixed in1e19e3611b; reply - P3 these tests build the part through the parser and then
assert(part.type === 'deleteApproval'), which... (round 2): fixed in1e19e3611b; reply - P2
toHaveTextContent(string)matches a substring, so "Answered: delete them" satisfies the 'an answer the... (round 2): fixed in1e19e3611b; reply - P2 The readback plural adds listed and unlisted nodes, but no readback fixture had unlisted nodes (round 2): fixed in
1e19e3611b; reply - P3 These four rows pin constraints of the cloud-owned
zAgentDeleteApprovalContext(string ids,.max(50),... (round 2): fixed in1e19e3611b; reply - P3 'a third option' catches a superset, but not a subset: with "every id is delete or keep" and no length... (round 2): fixed in
1e19e3611b; reply
Review-body findings:
- P1 No screenshots or screencast of the new card (round 1, round 2, round 3): resolved. Stories in
0c24e4cbeb; an element screenshot with a Linux baseline inedc8757ec3andf595ec7600; the image is embedded in the description at56ccdbf2e5. - P2 Description said a malformed node list still shows the buttons, and overstated what Delete covers (round 1): fixed in the rewritten body. The blank-id attribution (round 2) and the free-text condition were fixed at
56ccdbf2e5. - P3 #18949 narrows
dropResolvedAskstorun_approval(round 1): answered by the author and pinned by the refetch store test inc60d4f66b9, which still fails at4ba5ea3142ifretireAskPartskeeps onlyask_usercards. - P3 AI attribution in commits and the body footer (round 1): footer removed; commits since round 2 are clean. The earlier trailers were removed by the round 9 history rewrite.
- P3 Body not on the PR template; test list incomplete; readback claim outdated (round 1): fixed in the rewritten body. The spec description went stale again after this round's merges (P3 note 1).
- P3 Backend gating claim not checkable (round 1, round 4): fixed at
56ccdbf2e5. - P3 No Storybook story for the card (round 1): fixed in
0c24e4cbeb. - P3 Screenshot not embedded in the description (round 4): fixed at
56ccdbf2e5. - P3 #20891 strict ask parsing (round 2, round 4): addressed at
56ccdbf2e5. The body states this PR keeps #19981's tolerant policy, and the author left a coordination note on #20891. - P3 Comments added or left stale under #20823 (round 2): gone; the PR adds no comments.
- P3 Shared
AskCardwrapper and per-buttonloading(round 1): accepted as a follow-up. - P3 A 422 showed ask_user copy on the delete card (round 6): addressed by #19981's
e48eb49559. - Cleared in round 6: the delete e2e moved to #19981's shared
holdAnswerand dropped itswaitForRequestguard.answeringstays set until the resolution arrives, so the lock check does not depend on catching the POST mid-flight.
Existing reviewers: dante01yoon's contract-first finding, fixed in 039f17e400; his malformed-list, 5xx-wording, and answer-readback findings, addressed in round 1. christian-byrne's fail-closed list and black-box e2e requests: addressed; his exact-head review found nothing new. CodeRabbit's five findings: addressed in round 1.
-
christian-byrne's gate disposition (round 8): the no-behaviour-change point is confirmed and moot after dante01yoon's approval; the base overlap is confirmed, and a local merge with
dd3c8fb3e1is clean, with agent unit tests andvue-tscpassing; the three unresolved threads are contested as a blocker because all are outdated, answered, and owned by the approving reviewer. -
dante01yoon approved at
c823e90b2b(round 8), replacing his Oct 5 changes-requested review. -
skishore23's re-approval request (round 9): confirmed. The rewritten
5415df5bd6has the same tree asc823e90b2b, and the merges of #19981 andmainleave the PR's hunks unchanged. -
P3 Description asked for, but did not contain, a squash-message instruction to drop the Claude co-author trailers (rounds 6 to 8): addressed by the history rewrite at
2a6c8becd3; no commit carries the trailer. -
P3 Stale Verification section (rounds 7 to 10) and the stale "Stacked on #19981" line (rounds 9 and 10): open, non-blocking.
-
christian-byrne's second gate disposition (round 10): required checks and the grant point confirmed; the unresolved threads confirmed as gate state owned by dante01yoon; the overlap with
59965ddd23confirmed, and a local merge is clean and typechecks. The only test failure on the merged tree,useAgentSession.test.ts(g18), also fails on plainmainunder suite load.
Checks at 2a6c8becd3: CI 111 success, 18 skipped, no failures. At c823e90b2b: CI 111 success, 19 skipped. At 4ba5ea3142, the four delete-card unit files passed locally (290 tests). playwright test --list shows four tests in agentAskConsent.spec.ts. Changing retireAskParts to keep only ask_user cards fails the parts retire test and the store refetch test.
…pproval card (#19981, #20092) (#21308) Manual backport of #19981 and its stacked follow-up #20092 to `core/1.57`. Two commits, in merge order, so 1.57 users get the real "Ask user" card and the delete-approval card. ## Why On 1.57 the agent panel cannot render an `ask_user` question, so a Desktop turn that asks one hangs with the spinner running until the user presses stop (https://linear.app/comfyorg/issue/PM-2127). The backport of both PRs to this line was approved by Christian on 2026-10-18. ## What differs from the originals - Code, tests and snapshots are byte-identical to the merged commits. - Dropped: the `docs/architecture/domains` boundary edits from #19981. On 1.57 the two `records/*.domain.json` files do not exist and `boundaries.json` has no `agent` or `design-system` rules, so there is nothing to approve. - One conflict in `agentApiSchema.ts` imports: 1.57 does not import `zAgentAttachmentRejected`; kept the line's imports plus the new ask schemas. `@comfyorg/ingest-types` on 1.57 already exports `zAgentAskKind`, `zAgentPendingAskBase` and `zAgentAskResolvedData`. ## Verification - `pnpm typecheck:app`, `pnpm typecheck:browser`, oxlint, eslint, stylelint, oxfmt: clean (pre-commit). - `vitest run src/workbench/extensions/agent`: 164 files pass. One known order-dependent test, `(g18) a history fetch that never answers is abandoned at the recovery deadline`, fails when run alone on this line, on unchanged `core/1.57` and on `main` alike, so it is not caused by this backport. Do not merge one commit without the other: #20092 is stacked on #19981. --------- Co-authored-by: skishore23 <shimikeri.kishore@gmail.com> Co-authored-by: github-actions <github-actions@github.com>
…approval card (#19981, #20092) (#21309) Manual backport of #19981 and its stacked follow-up #20092 to `cloud/1.57`. Two commits, in merge order, so 1.57 users get the real "Ask user" card and the delete-approval card. ## Why On 1.57 the agent panel cannot render an `ask_user` question, so a Desktop turn that asks one hangs with the spinner running until the user presses stop (https://linear.app/comfyorg/issue/PM-2127). The backport of both PRs to this line was approved by Christian on 2026-10-18. ## What differs from the originals - Code, tests and snapshots are byte-identical to the merged commits. - Dropped: the `docs/architecture/domains` boundary edits from #19981. On 1.57 the two `records/*.domain.json` files do not exist and `boundaries.json` has no `agent` or `design-system` rules, so there is nothing to approve. - One conflict in `agentApiSchema.ts` imports: 1.57 does not import `zAgentAttachmentRejected`; kept the line's imports plus the new ask schemas. `@comfyorg/ingest-types` on 1.57 already exports `zAgentAskKind`, `zAgentPendingAskBase` and `zAgentAskResolvedData`. ## Verification - `pnpm typecheck:app`, `pnpm typecheck:browser`, oxlint, eslint, stylelint, oxfmt: clean (pre-commit). - `vitest run src/workbench/extensions/agent`: 164 files pass. One known order-dependent test, `(g18) a history fetch that never answers is abandoned at the recovery deadline`, fails when run alone on this line, on unchanged `cloud/1.57` and on `main` alike, so it is not caused by this backport. Do not merge one commit without the other: #20092 is stacked on #19981. --------- Co-authored-by: skishore23 <shimikeri.kishore@gmail.com> Co-authored-by: github-actions <github-actions@github.com>
…nning (Comfy-Org#21320) ## Summary A 409 from the ask answer endpoint can come back with `selected` missing or `null` (`AgentAskConflict.selected` is nullish). `toAskResolution` turns that into an answered resolution with `selected: []`, and the 409 branch always ran the superseded check. So a user who picked an option got a false "answered somewhere else" notice and an `agent_ask_answer_superseded` error report, even though nothing said another answer won. Fix: only run the superseded comparison when the conflict body actually carries a selection array. This is the same guard the 202 path already has (`receipt.selected !== undefined`). The card is still retired with the conflict's status, as before. Found by the cursor-review panel on the 1.57 backports of Comfy-Org#19981/Comfy-Org#20092 (Comfy-Org#21308, Comfy-Org#21309). The code there is identical to main, so this fix goes to main first and is backported after those land. ## Tests - New: `does not report a 409 with no selection as another answer winning` and `... with a null selection ...` in `useAgentSession.test.ts`. Both fail on main before the fix (`reportError` called with `agent_ask_answer_superseded`) and pass after. - Existing `shows the answer a 409 reports as already committed` (409 that names another selection) still reports superseded. - `useAgentSession.test.ts`: 253 pass. The one failure is the known order-dependent `(g18)` test, which fails on main without this change (https://linear.app/comfyorg/issue/FE-3485).
Summary
Renders the agent's
delete_approvalask as its own card, so a delete of nodes the user added waits for an explicit Delete or Keep. Stacked on #19981.DeleteApprovalCard.vueis built likeRunApprovalCard.vue. It shows the host's prompt, the nodes it would remove (title, else type, and#id), then Keep (secondary) and Delete (destructive). The answer goes through the sharedanswerAskpath as{ selected: ["delete" | "keep"] }, so it inherits feat(agent): render ask_user questions and advertise ask_kinds #19981's handling: the committed 202/409 answer, 422 back into service, refusals retired.contextis parsed with the generatedzAgentDeleteApprovalContext. The card lists the server's nodes (at most 50 by contract, unique by id) and says "N more nodes not listed" fromhidden_node_count, so the user always knows how many nodes a Delete covers.delete/keepas one required choice, or with free text allowed;contextthe generated schema rejects (missing, empty, or a malformed entry);toDeleteTargetsbecause the generatedidis a plain string.AskUserAnswer.RENDERED_ASK_KINDSgainsdelete_approval, soask_kindsadvertises it on each turn. Strings live underagent.deleteApproval.*.delete_approvalinask_kindsbut does not raise this ask yet. Theask_kindsdescription in this repo'spackages/ingest-types/src/types.gen.tssays "delete_approvalraises no card yet", so the card is inert until that lands.Verification
Vitest:
pnpm exec vitest run src/workbench/extensions/agent(163 files, 3,434 tests):agentMessageParts.test.ts: parsing,hidden_node_count, notice fallbacks, option shape, retirement;AgentMessage.test.ts: card, node list, answers, in flight, unlisted count, every resolved readback;agentConversationStore.test.ts: settling drops a pending delete card, and a retired one stays read-only through a refetch;agentRestClient.test.ts:ask_kindsas literals.retireAskPartskeep onlyask_usercards still fails the retention tests.Playwright,
browser_tests/tests/agent/agentAskConsent.spec.ts(renamed fromagentRunApprovalConsent.spec.ts; it now covers every ask card):fixtures/data/agent/agentAskFrames.ts, typed against the generatedAgentAskFrame. The turn harness is feat(agent): render ask_user questions and advertise ask_kinds #19981'sagentAskTurnFixture.Screenshot:
agent-delete-approval-card-cloud-linux.png, the pending card, recorded through the New Browser Test Expectations run:Storybook:
Agent/AgentMessagehasDeleteApproval,DeleteApprovalPending,DeleteApprovalLongList(50 listed, 30 unlisted) andDeleteApprovalResolved.pnpm typecheck,pnpm typecheck:browser, oxlint, oxfmt and knip are clean.