Skip to content

fix(crew): one status line for a crewmate's current step, still indicator - #18256

Merged
CrysisDeu merged 1 commit into
mainfrom
fix/crewmate-live-activity-row-18238
Oct 9, 2026
Merged

CrysisDeu merged 1 commit into
mainfrom
fix/crewmate-live-activity-row-18238

Conversation

@CrysisDeu

@CrysisDeu CrysisDeu commented Oct 8, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: a crewmate's chat shows the one step the crewmate is on right now, above a working indicator that never moves, and nothing of the steps once the turn is over.

Since #17794 a crewmate DM keeps every tool call and thinking row of the running turn, so a long patrol stacks a dozen 🔧 lines between the crewmate's bubbles (#18238, screenshot on the issue). The owner's rule for the live turn (#18238): the chat answers "what is the crewmate doing now", never "what has it done" — and (owner, on round 1 of this PR) the ghost-pose working indicator must not hop up and down while it does.

Why it matters

Every patrol cycle a crewmate runs today turns its DM into a tool-call log the Work log already holds, burying the few bubbles the user is meant to read; the reader has to scroll past the machinery to find them. With round 1's transcript-row approach the indicator hopped a row on every step, which reads as jitter for the whole turn. Members whose job is to patrol quietly cannot have a readable chat without both fixed.

Not a goal

  • Expandable details on the status line (input / output / diff). The Work log and the Sessions page hold those; this is a status, not a pill.
  • Changing what the Sessions page or the Work log draws (unchanged; the rows stay in the transcript).
  • The ordinary (non-crewmate) chat; filterCrewmateChat and the new line are reached only under ChatPane's crewmate prop.

What changed (motivation → approach → change)

Observed: the live turn's progress rows were kept in the crewmate chat (filterCrewmateChat(messages, live)), so a long turn rendered as a list, and any per-step mount/unmount above the footer moved the indicator. Root cause: "what the crewmate is doing" was modelled as transcript rows. Change: it is a status line, and the status is the one the DM header pill already computes.

The turn in flight is filtered like any other turn (filterCrewmateChat is back to "what the crewmate says"; the live parameter and the opensTurn scan are gone). The current step is one fixed-height status line directly above the working indicator (the ghost-pose carousel in ChatFooter), mounted for the whole live turn, words only:

data-activity status line reads
tool the open tool call's shared status label (toolStatusLabel: purpose / derived title / raw title, same preference and language guard as the sidebar row and the header pill)
thinking the indicator's own "Thinking…" — the call has returned (its output is in the tool log), the model is reasoning, or the open call is nothing_to_do (#16429; the header pill reads the same)
writing / working / delegated / compacting / stopping the header pill's phase words (PILL_ACTIVITY_KEY)
not live no line
  • pages/members/useSlotActivity.ts (new): the ONE reading of "what the crewmate is doing right now" — the slot's live status record (slotStatusDetail), stream state and the tool log's returned flag folded by the existing pure resolvePillActivity. Extracted from MembersPage's identity-pill selectors, which now call it; the chat line calls it too, so the header and the chat can never name one moment differently (First Principles R5). The nothing_to_do call reads as thinking on BOTH surfaces (no per-caller option — First Principles R6), matched on the tool log entry's trusted identity (tool_name + mcp_server, the QUIET_END_* constants), never on its title. Both callers pass the same delegatedOnly (Design R6).
  • pages/members/pillActivity.ts: PILL_ACTIVITY_KEY moved here (exported) from MembersPage so both renderers share the phase words; a tool activity also carries fullText (the label before the 40-char clamp the pill needs).
  • components/chat/CrewmateLiveActivity.tsx (new): one h-5, truncating row in the footer column; text from the hook's PillActivity — the FULL label where the pill clamps (this column has the room; CSS truncate is the only cut, title carries the whole text — UX R6); fades in on each change (reduced-motion respected). No spinner — the ghost under it carries the motion (UX R5). The role="status" / aria-live="polite" region is the stable wrapper, not the keyed span. role="status", data-testid="crewmate-live-activity".
  • components/ChatPane.tsx: renders the line in belowRows above ChatFooter whenever the crewmate is live; the "hasn't said anything yet" hint is hidden while the crewmate is live (crewmateLive), since its current step is on screen.
  • store/chatSlice.ts: re-exports QUIET_END_SERVER / QUIET_END_TOOL from chat/selectors (the facade is the only way out of src/store; chatSlice.ownership.test and chatSlice.publicSurface.test pin it).

Why a status line and not the transcript's own tool row. Round 1 kept one live transcript row and swapped it per step; the ghost under it hopped −25px/+25px on every swap (old row unmounts instantly, new row slides in), and a tool→thinking swap jumped it another 23px. Sampled with the harness below: swap1 [289,292,301,310,313,314 …] from a 314 baseline. A fixed-height line mounted for the whole turn holds the ghost still: 391 on every sample through two calls, a return, thinking, the nothing_to_do call, up to the turn's end.

Why the pill's seam and not the transcript. R2–R5 derived the step from transcript rows plus the tool log, which was a second spelling of what MembersPage's header pill already computes from slotStatusDetail (First Principles R5). Reading one seam removes the duplicate completion logic (meta.done, ✅/🚫 siblings, completedToolIds) entirely: the status record already knows the open call, and the tool log already knows it returned.

Spec: the "While a turn runs" paragraph of docs/system-specs/modules/crew-mode.md, and the Crew Members + Crewmate chat rows of docs/feature-map/README.md, rewritten in the same commit. No product default changes hands: the decision that the crewmate chat shows only what the crewmate says is the existing crew-mode spec; this PR changes how the running turn's single live step is drawn, recorded on #18238 by the owner.

Backwards compatibility

Compatible: filterCrewmateChat drops an optional parameter its one caller (ChatPane) stops passing; the header pill renders the same resolvePillActivity result through the extracted hook (its tests are unchanged and green); every transcript row is still persisted and still drawn on the Sessions page and the Work log; the facade re-exports are additive.

Tests

  • pages/members/useSlotActivity.test.tsx (new): idle when not running; names the open call by the shared label; thinking once the call returned (tool log output); nothing_to_do reads as thinking by identity on every surface, and a shell whose title says nothing_to_do is an ordinary step; a tool step carries fullText beside the clamped text.
  • crewmateBubbles.test.ts: the transcript never draws progress rows during a running turn; a live tool row between two replies does not reshape the run.
  • ChatPane.crewmate.test.tsx: the line shows the open call's label (data-activity=tool), one line, no wake text, no quiet hint; a returned call → thinking / "Thinking…", label gone; nothing_to_do in flight → "Thinking…", no hint; the line carries no SVG (words only); a long step shows its full label and carries it as title.
  • chatSlice.ownership.test.ts + chatSlice.publicSurface.test.ts: the two facade re-exports registered.
  • MembersPage.test.tsx / MembersPage.sideChat.test.tsx (205 tests, unchanged) pass against the extracted hook.
  • Run locally once (touched files only): all green; tsc --noEmit, eslint on touched files, scripts/docs-lint.sh clean. Full suites are CI's.

Manual verification

Recorded from the real MembersPage mounted by the capture/members-page.tsx harness (gateway-free, /api/** from fixtures), status frames arriving through the same setSlotStatusDetail / sseToolActivity / sseToolResult reducers the WebSocket layer dispatches. The ghost carousel's y was read at every step: 391 throughout. The fifth still shows the header pill and the chat line naming the same step from the one seam. A live-gateway run is not needed for this surface: the line reads only the persisted rows the fixtures reproduce byte-for-byte.

Screenshots / video

Light theme, crewmate DM pane, 646×560 crop. The recording: a running turn ("Working" → "Thinking…") → a tool call fills the line → the next call swaps the text in place → its return turns the line to "Thinking…" → another call → thinking → a nothing_to_do call ("Thinking…", same height) → the turn ends and line and indicator leave together. Stills: the open call, "Thinking…" after a return, "Thinking…" during nothing_to_do, after the turn, and a full frame with the header pill and the line agreeing.

Attached below as github.com/user-attachments assets (one GIF, five PNGs).

Related Issues

Fixes #18238. Follows #17794 (introduced the live rows) and #16429 (nothing_to_do).

Pattern harvest

Rule candidate: review-prompt
Pattern: "a live status drawn as transcript rows — anything mounted/unmounted above a pinned indicator moves it; model a changing status as one fixed-height element whose content changes".

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

Crewmate DM: the status line names the open call in full, swaps text per step, reads Thinking… after a return, during thinking and for nothing_to_do, leaves with the turn; the ghost never moves

tool: the open call label in full above the ghost indicator, words only

thinking: the call returned, the line reads Thinking… at the same height

thinking: nothing_to_do in flight never names itself

turn ended: no line, no indicator, only the crewmate bubbles

full frame: the header identity pill (clamped) and the chat status line (full) name the same step from the one shared reading

@CrysisDeu
CrysisDeu requested review from a team and KiroCrewCaptain as code owners October 8, 2026 21:56
@CrysisDeu
CrysisDeu requested a review from bolichen97 October 8, 2026 21:56
@CrysisDeu
CrysisDeu force-pushed the fix/crewmate-live-activity-row-18238 branch from 65da1b0 to dae8182 Compare October 8, 2026 22:18
@CrysisDeu CrysisDeu changed the title fix(crew): one live activity row in a crewmate chat, none for nothing_to_do fix(crew): one status line for a crewmate's current step, still indicator Oct 8, 2026
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Oct 8, 2026
@github-actions

github-actions Bot commented Oct 8, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — 🟡 CONCERNS

UX-level review of b36bedc4f39e8d092873041282f611c190164b4f — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

UX-Verdict: CONCERNS

When a quiet turn ends, the status line just disappears, and nothing says whether the crewmate finished, stopped or failed. "Working" and "Thinking…" also read as the same state.

Watch

  • The end of a turn looks the same as a stall. The diff renders {crewmate && crewmateLive && <CrewmateLiveActivity …/>}, so when the turn ends the line unmounts. A patrol turn that said nothing leaves the chat looking exactly as it did before the turn started. The blind reader, on shot-05: "No new reply appeared, so I can't tell whether it finished, stopped or failed." This happens on every silent patrol cycle, which is the case this PR exists for. It is friction, not task failure. Smallest fix: when a turn ends with no new bubble, leave one muted line (e.g. "Finished — nothing to report"). Spelled out as a fix after Suggestions.
  • "Working" and "Thinking…" can't be told apart. In shots 1–3 they sit in the same slot, and the reader said: "I can't tell what the difference is between 'Working' and 'Thinking...'". Both come from PILL_ACTIVITY_KEY and the thinking label. The line now shows these phase words for the whole turn, so the ambiguity is visible far more often than it was in the header pill. Low severity, because the line is not interactive.
  • The stopping, compacting and delegated states are not in any screenshot. In stopping and compacting the line deliberately renders an empty h-5 row above the footer's own text. That is probably fine, but nobody has looked at it. A still of each would confirm there is no orphaned gap.

Suggestions

  • In ChatPane.tsx, after a crewmate's live turn ends with no new bubble, render a one-line muted end state in the slot CrewmateLiveActivity occupied. Use the transcript's last-turn result ("Done", "Stopped", "Failed"), not a blank.

[UX-REVIEWED] b36bedc

Description read

[DESCRIPTION-READ] d05862b36c13f4888589985cbe0f970b87fe9523ef54a40187b03f6d9c352846

That sha256 names everything this verdict read: the PR description, plus the 9 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 8, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of b36bedc4f39e8d092873041282f611c190164b4f — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Check that the header pill now reading an in-flight nothing_to_do call as "Thinking…" is the change the owner wanted, not a side effect of the shared hook.

[DESIGN-REVIEWED] b36bedc

Description read

[DESCRIPTION-READ] 310f5bd8325da24aa3007c5fb4c28e4f672898261ca4d4e233ec8381d09ce2aa

That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 8, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ✅ PASS

Premise-level review of b36bedc4f39e8d092873041282f611c190164b4f — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

First-Principles-Verdict: PASS

Check that the owner's ruling on #18238 reverses #17794's "the running turn's tool calls show" for crewmate chats; the deleted live-row tests depend on that ruling.

What this change ships

Inventory (9 items) — 9 justified

Intent: while a crewmate is working, its chat should show only the one step it is on now, and the working indicator under it should hold still. This is a FIX.

  1. While a turn runs, the crewmate chat no longer lists that turn's tool calls and thinking — justified
  2. A single status line naming the current step sits above the working indicator for the whole turn — justified
  3. The header pill reads "Thinking…" while a nothing_to_do call is open, where it used to name the call — justified
  4. The "hasn't said anything yet" hint is hidden while the crewmate is working — justified
  5. The status line has no text during compacting or stopping, because the footer under it already says so — justified
  6. The status line shows the whole step label where the pill cuts it at 40 characters (fullText, one consumer) — justified
  7. The header pill and the chat line now share one reading of the current step (useSlotActivity, two consumers) — justified
  8. The pill's phase words (PILL_ACTIVITY_KEY) are now shared by both places that draw them (two consumers) — justified
  9. The QUIET_END_* constants are re-exported through the chat store's public entry point; the hook needs them, and the store can only be reached that way — justified

[FIRST-PRINCIPLES-REVIEWED] b36bedc

Description read

[DESCRIPTION-READ] 3bb2728c2f743d7c76501bee133e3f6e5e570bfc18f63d4025bafe3329fe7563

That sha256 names the PR description this verdict read. Recompute it with .github/scripts/pr-description-capture.sh; if it differs, the description changed after this verdict was formed and any finding drawn from the description is unproven.

@github-actions

github-actions Bot commented Oct 8, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of b36bedc4f39e8d092873041282f611c190164b4f and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] b36bedc

False positive or not applicable? A repository writer can comment:
/ai-review override gpt b36bedc4f39e8d092873041282f611c190164b4f: <one-sentence reason>

@github-actions

github-actions Bot commented Oct 8, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed b36bedc4f39e8d092873041282f611c190164b4f — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] b36bedc

Verdict parsed from the review's SHA-scoped output markers for commit b36bedc4f39e8d092873041282f611c190164b4f.

False positive or not applicable? A repository writer can comment:
/ai-review override fable b36bedc4f39e8d092873041282f611c190164b4f: <one-sentence reason>

@CrysisDeu
CrysisDeu force-pushed the fix/crewmate-live-activity-row-18238 branch from dae8182 to 5526700 Compare October 8, 2026 23:02
@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • BLOCK — cannot evaluate: missing screenshots and recording of the crewmate status line

fixed in R3 5526700a4e: the GIF and four stills (current, done, empty during nothing_to_do, turn ended) are attached to the description as user-attachments assets and verified anonymously fetchable (HTTP 302) before this push; the R2 asset answered 404 to the lane's unauthenticated fetch. The dimmed state is gone: a thinking burst now turns the spinner into a check mark on the same text-muted label (shape, not a 40% fade), as the Suggestions line asked.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • No rendered evidence for a new user-visible surface

fixed in R3 5526700a4e: live-activity-line.gif plus four stills are attached as user-attachments assets (verified HTTP 302 anonymously before the push; the R2 asset answered 404 to the lane). The recording covers the first call appearing, a step swap, the thinking state, the nothing_to_do case and the turn's end.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • The line still moves the indicator once per turn

fixed in R3 5526700a4e: the h-5 slot is now mounted for the whole time crewmateLive is true — data-state=empty before the first call and during a nothing_to_do call — so the indicator does not move at either point. Sampled ghost y is 391 on every step from an empty line through two swaps, the done state and the quiet-end call.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author

rebutted: it does. The owner's rule on #18238 (comments of 2026-10-08, Zezhen) is "show only what the crewmate is doing right now, never the list of steps; nothing_to_do shows nothing; nothing after the turn", and on round 1 of this PR he added that the indicator must not move — which the transcript-row form cannot satisfy (measured in the body). The #17794 paragraph and tests pinned the list form that rule replaces; the new paragraph in crew-mode.md and the crewmateLiveActivity tests pin the status-line form.

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 8, 2026
@CrysisDeu
CrysisDeu force-pushed the fix/crewmate-live-activity-row-18238 branch from 5526700 to 060fc2c Compare October 8, 2026 23:37
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Oct 8, 2026
@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • The done check implies success.

fixed in R4 060fc2c9fe: the done state and the check mark are gone. The line has two states only — current (spinner + the running call's label) and thinking (spinner + the indicator's own "Thinking…"), which it enters the moment the call's ✅ / 🚫 row arrives or a thinking burst follows. No outcome is implied for any step.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • The status line's first reading was only "a guess".

rebutted: a visible "Now:" prefix would add a new i18n key (propagated to every locale for one word) and would make the crewmate's line differ from the transcript's own tool pill, which carries no prefix either. R4 frames the line as a status by behaviour instead: it alternates between the running call's label and the indicator's own "Thinking…" wording (the same string the ChatFooter reads out), under a turning spinner, in the footer column rather than the message column. Shot-02 of the same reader was already "sure it's a status"; if the next blind read still guesses, this reopens.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • CrewmateLiveActivity empty state (shot-04): the reader saw an "empty gap"

fixed in R4 060fc2c9fe: the line is never blank. Before the first call, after a call completes, and during a nothing_to_do call it reads "Thinking…" (the existing pages.chat.chatFooter.thinking string) with the spinner, so the reserved row reads as intentional.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • The done state still shows a finished step during thinking

fixed in R4 060fc2c9fe: the subtraction asked for. current is removed from the helper's return (it now returns the running 🔧 row or null), the Check/done branch is deleted from CrewmateLiveActivity.tsx, and a call that completed (✅ / 🚫 sibling) or was followed by a thinking burst yields null. The one addition over "show empty" is that the reserved row shows the indicator's existing "Thinking…" wording instead of a blank, answering the UX lane's gap finding on the same head; no new string, no new state beyond current / thinking.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • Check that the spinner stops once a call's ✅ arrives with no thinking after it

fixed in R4 060fc2c9fe: crewmateLiveActivity now treats a call as over when its ✅ / 🚫 completion row (same tool_call_id) has arrived, independent of any thinking burst; the line then reads "Thinking…". The spinner keeps turning in that state on purpose — it marks the crewmate as working, which it is — but the completed call's label is no longer shown.

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 8, 2026
@CrysisDeu
CrysisDeu force-pushed the fix/crewmate-live-activity-row-18238 branch from 060fc2c to 223ff8f Compare October 9, 2026 00:01
@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • Completed calls remain current despite the documented completion behavior

fixed in R5 223ff8fe36: crewmateLiveActivity now returns null when the newest call's row carries the runner's persisted meta.done (what a reload hydrates), and takes a completedIds set that ChatPane fills from selectSlotToolLog (an entry with output or rejected), since meta.done is persisted but not broadcast live. The ✅ / 🚫 sibling rule stays as a third signal. Two tests pin the meta.done and tool-log paths; spec and feature-map row name all three signals.

@CrysisDeu
CrysisDeu force-pushed the fix/crewmate-live-activity-row-18238 branch from c674cc8 to 1fcc349 Compare October 9, 2026 00:58
@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • hideQuietEnd makes the chat line and the header pill name the nothing_to_do step differently

fixed in R7 1fcc349a28: the subtraction asked for. SlotActivityOptions.hideQuietEnd is deleted; the quiet-end read lives inside useSlotActivity with no option, so the header pill and the chat line both read "Thinking…" while a nothing_to_do call is open. One call site shape, no fork; the hook test pins it for the default call.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • The step label is cut short with room to spare.

fixed in R7 1fcc349a28: a tool activity now carries fullText (the shared label before the pill's 40-char clamp); CrewmateLiveActivity renders it, truncates by CSS only, and carries the whole text as title. The pill keeps the clamp it needs for its 24rem slot. The aria-live="polite" / role="status" region is now the stable wrapper, not the keyed span (your Suggestions line). Evidence re-recorded: the still shows the full label.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • Check that the chat's status line and the header pill show the same words during a delegated-only turn

fixed in R7 1fcc349a28: ChatPane passes delegatedOnly: !!paneSlot?.subagents_running && !paneSlot?.running, the same expression the pill computes from its live slot, so both read "delegated" on a workers-only turn.

…ator

Since #17794 a crewmate DM kept every tool call and thinking row of
the running turn, so a long patrol stacked a dozen 🔧 lines between the
crewmate's bubbles. The owner's rule: the chat answers "what is the
crewmate doing now", never "what has it done" -- and the working
indicator must not hop while it does.

The turn in flight is now filtered like any other; the current step is
ONE status line directly above the working indicator
(`CrewmateLiveActivity`, from `crewmateLiveActivity(messages, live)`):
the newest tool call of the running turn, labelled as the transcript's
pill would label it. Each new call swaps the text in place, a thinking
burst after the call dims the line, and the line goes with the turn.
A fixed-height line whose text changes is what keeps the indicator
still: transcript rows mounting and unmounting above it moved the
ghost by a row on every step. The "hasn't said anything yet" hint stays
hidden while the crewmate is live.

A `nothing_to_do` call (#16429) shows nothing even in flight: matched
on the row's trusted identity (`meta.tool_name` + `meta.mcp_server`,
the `QUIET_END_*` constants), not on `meta.ends_turn`, which the runner
stamps only after the applier ran.

Spec paragraph in crew-mode.md and the feature-map row updated in the
same commit.

Fixes #18238
@CrysisDeu
CrysisDeu force-pushed the fix/crewmate-live-activity-row-18238 branch from 1fcc349 to b36bedc Compare October 9, 2026 01:27
@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • CrewmateLiveActivity shows a duplicate status during stopping / compacting

fixed in R8 b36bedc4f3: for those two kinds the line renders text-free (the ChatFooter directly under it already spells them out) and keeps its h-5 so nothing below moves. A test pins exactly one "Compacting…" on screen.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • The end of a turn says nothing.

rebutted (repeat of the R5 ruling on 223ff8fe36, same answer): a quiet end leaving nothing new in the chat is the product rule this surface was built on (crew-mode spec; #16429 nothing_to_do; the owner's #18238 direction "after the turn, nothing"). The end IS shown where it already lives — the DM header pill switches to "Idle · just now" the moment the slot stops, through the same hook — and a failed or stopped turn still draws its error row or stop card. A "Finished · see Work log" line would be the closing bubble the rule forbids.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • "Working" and "Thinking…" look the same to a newcomer.

rebutted (out of goal): both words are the header pill's existing PILL_ACTIVITY_KEY phases, unchanged by this PR and shown identically in the pill; "Working" is the phase before the first status frame, "Thinking…" the reasoning phase. Rewording a shared vocabulary is a cross-surface change, not this PR's one line. As your own finding says, not a blocker here.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • The status line was read correctly, but only as "a guess".

accepted as read: the reader identified it correctly ("a description of the current step", not clickable). The same text in the header pill is by design and now by construction — one hook, two renderers — so the pill (clamped, 24rem) and the chat line (full width, beside the indicator) cannot drift; the chat line is the one a reader following the conversation sees without looking up.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • Item 1 — inherited (decision not on record)

rebutted: the decision is the repository owner's, recorded on the issue this PR fixes. #18238 was opened against #17794's live rows; the owner's rule ("show only what the crewmate is doing right now, never the list; nothing_to_do shows nothing; nothing after the turn") is in the issue's first two comments, he approved the mock in this PR's R2 recording, and he added the "indicator must not move" constraint that the status-line form answers. The crew-mode spec's standing rule — the crewmate chat draws only what the crewmate says — is unchanged; #17794's live rows were a two-day-old exception to it, and this PR returns to the rule with a status line in place of the rows. No default, loop, command or capability changes hands, which is why this stayed at CONCERNS rather than BLOCK. If readiness still wants the override, the owner can post it; I am not posting one on his behalf.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • Item 4 — rides along: the header pill now reads an in-flight nothing_to_do call as thinking

accepted, and it is this lane's own R6 subtraction: R6 asked to delete the hideQuietEnd fork so the pill and the line read one thing; with no fork, the quiet-end read applies to both. The pill saying "Nothing to report" for a call whose whole purpose is to end quietly was the inconsistency; "Thinking…" for the beat before the turn ends is the #16429 contract on every surface.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • The end of a turn looks the same as a stall.

rebutted (third repeat; rulings on 223ff8fe36 and 1fcc349a28 stand): a quiet end leaving no new bubble is the product rule this surface was built on (crew-mode spec; #16429; owner's #18238 "after the turn, nothing"), and a "Finished — nothing to report" line is exactly the closing bubble that rule forbids. The end is shown in the DM header pill ("Idle · just now", same hook); a stalled or failed turn still draws its error row or stop card, so the two are not the same to a reader who looks at the pill or the chat's last row.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • "Working" and "Thinking…" can't be told apart.

rebutted (repeat of the 1fcc349a28 ruling, out of goal): both words are the header pill's existing PILL_ACTIVITY_KEY phases, unchanged here. Rewording a shared vocabulary is a cross-surface change, not this PR's one line; the pill reads them identically.

@github-actions github-actions Bot added readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running labels Oct 9, 2026

@buluoray buluoray left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approving b36bedc.

What this brings users: a crewmate's DM no longer turns into a tool-call log while it works. During a running turn the chat shows one status line, the step the crewmate is on right now, directly above the working indicator. The indicator no longer jumps on every step. When the turn ends the line goes away and only the crewmate's own messages are left. The header identity pill and the chat line now get their text from the same useSlotActivity hook, so they can't name the same moment differently.

Logic checked against source:

  • The line is mounted only while crewmate && crewmateLive, the same liveness ChatFooter uses. After the turn ends, is stopped or errors, it unmounts together with the indicator. Stale status cannot show on a resting slot because resolvePillActivity returns idle when the slot isn't busy.
  • A tool call counts as finished by tool_call_id against the tool log's output, so parallel calls can't be mixed up. An approved call keeps its name until its result arrives. nothing_to_do is recognised by tool_name + mcp_server, not by its title, so a shell command whose title mentions it is still drawn as an ordinary step.
  • The step rows are filtered out only when the chat renders. Nothing is removed from the transcript, so the Sessions page and the Work log draw what they did before. isCrewmateChatRow is unchanged, so error rows, stop cards and empty-turn notices still appear.
  • Stopping and compacting leave the line empty but keep its height, so the footer's own wording isn't repeated and the indicator doesn't move.
  • No new user-facing strings: the phase words reuse the existing PILL_ACTIVITY_KEY catalog entries, and the line is styled with text-muted only.

Upgrade and audience fit: the ordinary (non-crewmate) chat is unaffected. filterCrewmateChat and the status line are reached only with the crewmate prop. A non-crewmate pane calls useSlotActivity(''), which reads nothing and is never rendered; its empty-hint condition still works out to !running, and wrapping the footer in a fragment changes nothing in the output. The one change outside the crewmate chat is intended and noted in the PR: the header pill now reads an in-flight nothing_to_do call as "Thinking…".

Verification: reverting the quiet-end check, reverting the empty line for stopping/compacting, and keeping step rows in the filter each made a named test fail; using the clamped text instead of the full label made the long-label test fail. CI is green on this head: 92 passed, 9 skipped, 0 failed.

Non-blocking follow-ups:

  1. No test checks that the line is gone after the turn ends. Removing crewmateLive from the gate in ChatPane ({crewmate && crewmateLive && <CrewmateLiveActivity …/>}) keeps every test green, and without the gate the pane would show a permanent "Idle" line. Please add a ChatPane case that ends the turn and asserts crewmate-live-activity is absent — that is the "nothing after the turn" half of #18238.
  2. The quiet-hint gate change (!(crewmate ? crewmateLive : running)) isn't pinned for the case where only paneSlot.running is set (a patrol turn started by the server while the pane's stream is still idle). Putting the gate back to !running keeps the suite green; a test with that slot state would cover it.
  3. slotStatusDetail isn't cleared when a turn ends. If a new turn starts before its first status frame arrives, the line (and the header pill, as before this PR) can briefly show the previous turn's last phase, for example "Writing…". Not new, but the chat line makes it more visible; consider resetting the detail when a turn starts.

@CrysisDeu
CrysisDeu merged commit c7b17cf into main Oct 9, 2026
102 checks passed
@CrysisDeu
CrysisDeu deleted the fix/crewmate-live-activity-row-18238 branch October 9, 2026 03:13
@github-actions github-actions Bot removed the readiness: passed Eligible automated validation passed for the current revision label Oct 9, 2026
iamwhatever pushed a commit that referenced this pull request Oct 12, 2026
…ckering

Builds on #18256. The live reading flips to Thinking the moment a tool call
returns, so fast calls flashed past a line that mostly said Thinking. The
chat status line now holds each step for 600ms and a finished tool's title
for 1.5s before Thinking. The newest step wins when a hold ends, and a tool
that ran inside a hold still gets its turn; stopping/compacting swap at once.
The line is keyed by slot: the Members page rebinds one ChatPane to another
crewmate without remounting it, so a rebind starts from that crewmate's own
reading, never from the previous one's held step. The header pill is
unchanged. The crew-mode spec and feature map describe the hold.
iamwhatever pushed a commit that referenced this pull request Oct 12, 2026
…ckering

Builds on #18256. The live reading flips to Thinking the moment a tool call
returns, so fast calls flashed past a line that mostly said Thinking. The
chat status line now holds each step for 600ms and a finished tool's title
for 1.5s before Thinking. The newest step wins when a hold ends, and a tool
that ran inside a hold still gets its turn; stopping/compacting swap at once.
The line is keyed by slot: the Members page rebinds one ChatPane to another
crewmate without remounting it, so a rebind starts from that crewmate's own
reading, never from the previous one's held step. The header pill is
unchanged. The crew-mode spec and feature map describe the hold.
iamwhatever added a commit that referenced this pull request Oct 12, 2026
…ckering (#18528)

Builds on #18256. The live reading flips to Thinking the moment a tool call
returns, so fast calls flashed past a line that mostly said Thinking. The
chat status line now holds each step for 600ms and a finished tool's title
for 1.5s before Thinking. The newest step wins when a hold ends, and a tool
that ran inside a hold still gets its turn; stopping/compacting swap at once.
The line is keyed by slot: the Members page rebinds one ChatPane to another
crewmate without remounting it, so a rebind starts from that crewmate's own
reading, never from the previous one's held step. The header pill is
unchanged. The crew-mode spec and feature map describe the hold.

Co-authored-by: Joe Guo <zejiangg@amazon.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

bug(crew): crewmate DM lists every live tool call; should show only the latest, none after the turn

2 participants