Repository navigation
feat(ui): render durable: false receipt on the queue card (#11695) - #11768
fathirramadhan-web wants to merge 1 commit into
Conversation
…se receipt on the queue
|
👋 Hi! This PR's description is missing some required sections from our PR template. Workflow runs won't be auto-approved until the description is updated. Missing sections:
Please update your PR description to include these sections, then push or re-save the description. The workflows will be approved on the next cycle. |
4 similar comments
|
👋 Hi! This PR's description is missing some required sections from our PR template. Workflow runs won't be auto-approved until the description is updated. Missing sections:
Please update your PR description to include these sections, then push or re-save the description. The workflows will be approved on the next cycle. |
|
👋 Hi! This PR's description is missing some required sections from our PR template. Workflow runs won't be auto-approved until the description is updated. Missing sections:
Please update your PR description to include these sections, then push or re-save the description. The workflows will be approved on the next cycle. |
|
👋 Hi! This PR's description is missing some required sections from our PR template. Workflow runs won't be auto-approved until the description is updated. Missing sections:
Please update your PR description to include these sections, then push or re-save the description. The workflows will be approved on the next cycle. |
|
👋 Hi! This PR's description is missing some required sections from our PR template. Workflow runs won't be auto-approved until the description is updated. Missing sections:
Please update your PR description to include these sections, then push or re-save the description. The workflows will be approved on the next cycle. |
Design Review (Fable 5, fork) —
|
First Principles Review (Fable 5, fork) —
|
GPT 5.6 Review (fork) —
|
Opus 4.8 Review (fork) —
|
|
🤖 Kiro Crew [operator: NicholasRBowers#a942f9ca]: This PR has been inactive for 7+ days with failing CI. I've assessed the blockers and they appear resolvable — I'll push fixes directly to this branch as a co-author. Assessment: the change itself is fine; the reds are one un-propagated i18n key plus a stale review round. One thing I will also drop, because it is not yours: the If you'd prefer I don't touch this PR, add the |
|
Thank you for this, @fathirramadhan-web — the queue-card indicator and its tests are cleanly done. One thing surfaced while looking at landing it on Making it render would need a new backend producer (emitting I'm leaving this PR open for a maintainer to weigh in on that backend direction. Your authorship is preserved either way. Thanks again for the contribution! |
|
PR triage (2026-10-05): rebase needed, plus one overlap to coordinate Checked against Overlap with #10690 ( Status vs main. This isn't a duplicate, and nothing on main supersedes it. Main has no Recommendation: rebase and continue development. I found no breaking change and no P0 bug: the new field is optional, and the marker appears only when |
dodgeblaster
left a comment
There was a problem hiding this comment.
Thanks for this, @fathirramadhan-web — the queue-card marker, the reducer threading, and the three-branch vitest (false / absent / true) are cleanly written and are the right shape for the UI half of #11695. Requesting changes, because as it stands the PR cannot land and, more fundamentally, the marker it adds cannot render on real data today. Pointers below, ordered by what blocks it.
1. The marker never renders on main — the field it keys on is not emitted (the core blocker)
QueueStack.tsx shows the label only when m.meta?.durable === false, and chatSlice.ts preserves that field off the queue payload. But on current main (41987ae8) nothing writes durable onto any queue_push frame or enqueue receipt. The three producers carry exactly slot, content, ts, queue_id:
src/kiro_crew/dashboard/chat_delivery.py:1103(broadcast_ws("queue_push", push)), with the comment at:1105-1108: "It is not a field on this frame: a caller-visible flag was carried here and read by nothing."src/kiro_crew/dashboard/chat_api/chat_send.py:923(sub-agent hold branch)src/kiro_crew/dashboard/chat_turn/steer_queue.py:406(requeued steer — a third site not covered here)
The only consumer of the durable verdict is warn_if_not_durable (src/kiro_crew/dashboard/slot_queue_repository.py:192), which logs at WARNING and whose docstring records the decision: "Not a receipt field. A caller-visible durable boolean on the enqueue acknowledgments has no reader, so it is not shipped; the on-screen queue-card marker belongs with its consumer."
So this frontend-only diff renders a marker that is never reachable in production. Making it render requires re-adding the field at the producer sites together with this consumer — which is a backend emission + spec change (docs/system-specs/modules/session.md §queue durability currently documents the field as intentionally absent). That is a product/design decision #11677's author deferred on purpose, and it is still open on #11695. It is not something this PR can decide unilaterally: it needs a maintainer to settle (a) where the flag lives — WS frame, JSON receipt, or both; (b) whether the requeued-steer path at steer_queue.py:406 gets it too; and (c) the marker's wording and at-rest-vs-hover behaviour (which also fixes the i18n key). Until that is settled, the field and the marker must land in one change — re-adding the field alone would recreate the read-by-nothing state that got it removed.
2. CI is red and the PR is conflicting
- CONFLICTING / DIRTY, head
98a47d61is ~2416 commits behindmain. Needs a rebase onto currentmain. - Frontend Lint & Type Check, Frontend Tests (3), Frontend Coverage Merge, Coverage Gate all fail. The i18n key
components.queueStack.wont_survive_a_restartwas added toen.jsononly, socatalogParity.test.tsfails once per sibling locale andi18n:checkfails its pseudolocale hard-zero. Add the key to all 12 sibling catalogs and regenerate:node scripts/gen-pseudolocale.mjs, committing the result. - Screenshot Evidence fails — attach before/after stills of the badge once it can actually render (which depends on item 1).
3. Drop the unrelated .gitignore hunk
The .gitignore change is headed # Auto-added by github-engineer and adds *.pyc, *.pyd, *.obj, *.so, nem_stats.c, *.exp, *.lib. None of that relates to a queue-card receipt (nem_stats.c in particular looks like local tooling committing itself). Please remove it, or say if it was deliberate.
4. Coordinate the one overlap
#10690 (feat(members), open) rewrites the same QueueStack.tsx row — <span className="truncate flex-1">{m.content}</span> becomes a peerAuthored ? stripSentByPrefix(...) form — and edits QueueStack.test.tsx. Both changes can stand; whichever lands second must keep both the sent-by stripping and the queue-non-durable marker on that row.
Net: the UI work here is good and I'd like to see it land, but it is gated on a maintainer's backend/spec decision on #11695 (item 1), and separately needs a rebase, the i18n catalogs + pseudolocale, screenshots, and the .gitignore hunk removed. Leaving #11695 open for that decision; this is a fork PR so it's review-only on our side — the rebase and any merge are yours and the maintainers'.
|
[operator: bolichen97] Continuing this in #31441; thanks for the start. |
Problem / Motivation
When a queued prompt carries
durable: falsein its metadata (meaning it will notsurvive a gateway restart), the queue card in the dashboard shows no visual
distinction from a durable prompt. The user has no way to know which queued
messages are ephemeral until the restart discards them.
Refs #11695
Why it matters
Users who queue prompts with
durable: false(e.g. via the API or a subagent) seethem listed identically to persistent entries. A restart silently drops those
prompts with no prior indication, which is confusing and can cause perceived data
loss.
What changed
Symptom: no visual cue for non-durable queued prompts.
Root cause: the
durableflag from the backend queue payload was never threadedthrough to the store or rendered in the UI.
Changes:
chatSlice.ts— extendedSlotQueueItem,hydrateQueuedBubbles, andappendQueuedMessageto preservedurable: falsein the messagemetaobject.Only the explicit
falsevalue is propagated; absent/true leaves no extra field.QueueStack.tsx— whenm.meta?.durable === false, render an italic label(
wont_survive_a_restart) next to the prompt text on the queue card, with amatching
titletooltip.en.json— added thecomponents.queueStack.wont_survive_a_restarti18n key.Backwards compatibility
Compatible: the
durablefield onSlotQueueItemis optional; existing queueentries without it render exactly as before. No API contract changes.
Tests
QueueStack.test.tsx— three new cases in adurable: false receipt markerdescribe block:
queue-non-durablemarker whenmeta.durableisfalse.durableis absent (default).durableis explicitlytrue.queued()test helper was extended to accept arbitrarymetaoverrides.Manual verification
N/A — the rendered indicator is a plain
<span>with a test-id; the three testcases cover every branch (
false, absent,true).Related Issues
Fixes #11695
Pattern harvest
Not generalizable: one-off missing UI indicator for an already-available metadata
field.
Checklist
Contribution License Agreement