Skip to content

feat(schedule): show a crewmate's schedules as Crewmate in the list, with a by-crewmate filter - #22136

Merged
chenmingwei23 merged 1 commit into
mainfrom
feat/18772-schedule-crewmate-type
Oct 20, 2026
Merged

chenmingwei23 merged 1 commit into
mainfrom
feat/18772-schedule-crewmate-type

Conversation

@iamwhatever

@iamwhatever iamwhatever commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: On the Schedule page, a crewmate's own schedule reads as "Crewmate" with that crewmate's face and name, and the list can be narrowed to one crewmate's schedules.

Today the Schedule list shows every LLM job as "Agent · ". A crewmate's own schedule (a job with member_id, which runs in that crewmate's chat with its own memory) and a job that opens a new session on a custom agent look the same. A crewmate schedule whose crewmate has run once even shows only its durable id, because the row's member_id changes from the name to that id after the first run.

Why it matters

The two kinds of job deliver to different places: a crewmate schedule lands in that crewmate's chat with its private memory, a custom-agent job opens a new session on shared memory. #18772 asks the list to tell them apart so the user knows where a job's output goes, and to let a crewmate's jobs be read as one block.

Not a goal

What changed (motivation → approach → change)

The list now tells a crewmate's own schedule apart from a custom-agent job.

Row Type cell
Crewmate schedule (member_id set) Crewmate, then the crewmate's face and name, linking to /members?member=<name>
Crewmate schedule with no roster row (loading, failed, or crewmate gone) Crewmate, then the raw id, no link
Custom-agent job unchanged: Agent · template name
Script / command unchanged

The Type cell's hover tooltip names the crewmate for a crewmate schedule (the raw id when no roster row resolves), where it used to name the agent.

The page reads the crewmate from the members roster it already loads (membersRosterQuery, GET /api/members). Creation stores the crewmate's durable id in a row's member_id (bind_cron_memory), which is the roster row's slug, so crewmateForJob matches on the slug alone, as wakesCrew does; a slug two roster rows share names nobody, as the roster itself treats it. No backend change.

A by crewmate dropdown beside the search box narrows the list to one crewmate's schedules. It shows only when some listed job belongs to a crewmate, applies only while its crewmate still owns a listed job, and while it is set, folder collapse is bypassed the same way the search does. The search box also matches the crewmate's name and display name. If the roster read fails while a crewmate schedule is listed, an inline notice with Retry says so, rather than the links and filter quietly disappearing. The change is in pages/SchedulePage.tsx. Three new strings are translated in all 12 locales (i18n/locales/*.json, notes in i18n/en.context.json), en-XA is regenerated, and learn-cron-dashboard.md describes the row field, the Type cell and the filter.

Backwards compatibility

Compatible: no API changes; the page reads GET /api/crons and GET /api/members as they are on main.

Removes nothing: the Type cell keeps every existing case and adds one for rows with member_id.

Tests

  • website/src/test/SchedulePage.crewmateType.test.tsx: a crewmate job's Type cell reads Crewmate and links to /members?member=radar; a template job is unchanged and unlinked; a slug two roster rows share names nobody (raw id, unlinked); a crewmate filter whose crewmate no longer owns a listed job is dropped; a failed roster read shows the notice, keeps the raw ids, and recovers on Retry; search finds a crewmate job by its display name; a gone crewmate shows the raw id, unlinked; the filter narrows to one crewmate; no filter shows when no job belongs to a crewmate.

The frontend file is red on main's SchedulePage.tsx and green on this head, including with KIROCREW_ACT_STRICT=1.

Manual verification

Looked at the real Schedule page on a loopback dev server with a stubbed API (two crewmate schedules, one stored by durable id and one by name, a custom-agent job and a script job, with the members roster stubbed), dark and light, with the filter closed, open and set. The frames below are those captures.

Screenshots / video

Before: crewmate schedules read as Agent, with the raw id.

before: crewmate jobs show Agent and a raw member id

After: crewmate schedules read as Crewmate with face and name; the filter sits beside the search.

after: Crewmate rows with face and name, by-crewmate dropdown

After: filtered to Radar.

after: list narrowed to Radar's schedule

After: the members roster fails to load. The notice says so with Retry, and crewmate rows show the raw id, unlinked.

after: roster read failed, inline notice with Retry above the list, crewmate rows show raw ids

More: filter open, light theme

after: by-crewmate dropdown open

after: light theme

Related Issues

Closes #18772

Slice 3 of #18772 (the list), the last one: slice 2 (the form's "Runs as" switch, #22055) is on main.

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

@github-actions

github-actions Bot commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of 7745f8d7128f0f6d544930d8e9f66bc12d77b417 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Nothing to check.

[DESIGN-REVIEWED] 7745f8d

Description read

[DESCRIPTION-READ] 25fd8e6dd105881c9b29549dffe6a9d3eee77efba6829c240277e689c806df6c

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 19, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — 🟡 CONCERNS

UX-level review of 7745f8d7128f0f6d544930d8e9f66bc12d77b417 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

UX-Verdict: CONCERNS

The list now shows "Crewmate" next to "agent", but nothing on screen tells a first-time user how the two differ. Explaining that difference is the reason this PR exists.

Watch

  • "Crewmate" vs "agent" doesn't say where output goes. The PR's stated purpose is that "the user knows where a job's output goes." The type cell only adds the word "Crewmate", with no hint that the output lands in that crewmate's chat. The blind reader said: "both seem to mean an AI helper… can't tell what the difference is." This affects every visit to a mixed list and defeats the PR's goal. Fix: make the type cell's tooltip (its title) say where output goes, e.g. "Runs in Radar's chat", or add the same as a small hint line.
  • "No owning session" now sits under a named owner. Every crewmate row shows a crewmate (the owner) next to "No owning session". The reader concluded that "'owning' seems to mean two things." This is a low-impact contradiction, but it appears on every row. Fix: hide or reword that line on rows that have a member_id.
  • Crewmate name link was only guessed. The underlined "Radar"/"Scribe" link was read only as "probably opens that helper's page — a guess". Fix: give the <Link> a tooltip such as "Open Radar's crewmate page".
  • Rows show raw ids before names load. While the roster is still loading, CrewmateTypeCell shows job.member_id (e.g. "member-radar-7f…"), then swaps to the face and name. The reader saw one helper under three names. Fix: show a skeleton in that cell while the roster is pending, and keep the raw id for the error or gone-crewmate case only.

Suggestions

  • Re-capture shot-04 with "Backup check" in it. The reader worried it "disappeared when only the helper list failed to load". The code has no path that hides it, so the frame is misleading evidence.

[UX-REVIEWED] 7745f8d

Description read

[DESCRIPTION-READ] 0ac2b6aed11edf78b87a3cb91009c996eafcab9be9ebe6b23df816cfcbf2cc6d

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 19, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ✅ PASS

Premise-level review of 7745f8d7128f0f6d544930d8e9f66bc12d77b417 — 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 #18772 asks for a separate dropdown: this PR also makes search match crewmate names, so typing "Radar" already narrows the list.

What this change ships

Inventory (8 items) — 8 justified

Intent: let a user tell a crewmate's own schedule (it runs in that crewmate's chat with its private memory) apart from a custom-agent job, and read one crewmate's jobs together. ADDITION.

  1. A crewmate's schedule now reads "Crewmate" instead of "Agent" in the Type cell, with the crewmate's face and name — justified
  2. That name links to the crewmate's page on Crewmates — justified
  3. The Type cell's hover text names the crewmate instead of the agent — justified
  4. A crewmate whose roster row doesn't resolve shows its raw id, with no link — justified
  5. A new "by crewmate" dropdown beside the search narrows the list, shown only when a listed job belongs to a crewmate — justified
  6. Search now also matches the crewmate's name and display name — justified
  7. While the crewmate filter is set, folder collapse is bypassed and the empty state's Clear also resets it — justified
  8. If the members roster fails to load while a crewmate job is listed, an inline notice with Retry appears — justified

[FIRST-PRINCIPLES-REVIEWED] 7745f8d

Description read

[DESCRIPTION-READ] fe9390b31cc28c6129aae52215154e9c38ac0fa984cdd8e6c89a829bace296a8

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 19, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed 7745f8d7128f0f6d544930d8e9f66bc12d77b417 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 7745f8d

Verdict parsed from the review's SHA-scoped output markers for commit 7745f8d7128f0f6d544930d8e9f66bc12d77b417.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 7745f8d7128f0f6d544930d8e9f66bc12d77b417: <one-sentence reason>

@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 19, 2026
@github-actions

github-actions Bot commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of 7745f8d7128f0f6d544930d8e9f66bc12d77b417 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 7745f8d

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 7745f8d7128f0f6d544930d8e9f66bc12d77b417: <one-sentence reason>

@iamwhatever
iamwhatever force-pushed the feat/18772-schedule-crewmate-type branch from d86f6ef to ac470ee Compare October 19, 2026 09:40
@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 19, 2026
@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=6fb06890b13d second spelling of the crewmate identity map

Outcome: fixed.
Removed member_name from GET /api/crons, _crewmate_names_by_member_id and its test; handlers/cron_board.py is now identical to main.
The page resolves a job's crewmate from the members roster it already loads (membersRosterQuery): roster slug (the persisted member_id) first, then the stored name; a slug two rows share names nobody, as members.py treats it. Fixed in ac470ee.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=5d3077f90764 duplicate of the GET /api/members slug/name map

Outcome: fixed.
Same change as span 6fb06890b13d: the server field and helper are deleted and the PR body no longer claims the page cannot map the id. Fixed in ac470ee.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=25357b4f38a7 undeclared tooltip change

Outcome: fixed.
The Type cell's tooltip naming the crewmate (or the raw id when no roster row resolves) is now part of the body's Type-cell table. Fixed in ac470ee.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=b9f5f186bc32 crewmate name does not look like a link

Outcome: fixed.
The name in the Type cell is dotted-underlined (underline decoration-dotted underline-offset-2) and still brightens on hover. Fixed in ac470ee.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=fac6d546661f Crewmate vs Agent split has no visible meaning

Outcome: rebutted.
The words follow the job form that creates these jobs, merged in #22055: its "Runs as" switch names the two kinds Custom agent and Crewmate, with hints saying where each runs.
The list line carries the crewmate's face and name, linking to that crewmate's page; adding an explanation per row would repeat the form on every row.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=5bd86edf182a screenshots of a removed crewmate and an empty filtered list

Outcome: rebutted.
Both states are pinned by tests rather than frames: 'shows the raw id, unlinked, for a crewmate that is gone' and 'names nobody when two roster rows share the stored slug'.
The empty filtered state cannot occur from the UI: the dropdown only offers crewmates that own a listed job. Its notice is the existing FilteredEmpty used by the search.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=0c64588541a5 function-local import in the new helper

Outcome: fixed.
The helper is deleted (see the first-principles disposition), so the import is gone with it. Fixed in ac470ee.

@iamwhatever
iamwhatever force-pushed the feat/18772-schedule-crewmate-type branch from ac470ee to bbcf5d4 Compare October 19, 2026 09:51
@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=0559b9bc5561 name fallback disagrees with wakesCrew

Outcome: fixed.
The name fallback is removed: crewmateForJob matches the roster slug alone, as wakesCrew does, and a slug two rows share names nobody.
Creation stores the durable id: bind_cron_memory runs on every add path and sets job.member_id from derive_execution(target_member=...). Only a record written by an older build can still hold a name; it shows the raw value, unlinked, exactly as wakesCrew leaves it unattributed. Fixed in bbcf5d4.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=6b61208bf8d8 crewmateFilter never resets when its crewmate drops out

Outcome: fixed.
The filter is applied through activeCrewmate = crewmateOptions.find(...), which drives the predicate, the select value, filterActive and the empty-state query; a crewmate with no listed job leaves the list unfiltered and the trigger on "All crewmates".
Pinned by 'drops a crewmate filter whose crewmate no longer owns a listed job', red without the change. Fixed in bbcf5d4.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=e7ece97dd9d9 empty state shows the raw crewmate name

Outcome: fixed.
FilteredEmpty now gets crewDisplayName(activeCrewmate) for the crewmate filter. Fixed in bbcf5d4.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=56c518a64ffe no frame of the raw-id fallback or the filtered-empty state

Outcome: rebutted.
The raw-id fallback is pinned by 'shows the raw id, unlinked, for a crewmate that is gone' and 'names nobody when two roster rows share the stored slug'.
The crewmate-filtered empty list cannot be reached: the dropdown offers only crewmates that own a listed job, and a filter whose crewmate drops out is now released (span 6b61208bf8d8). It can show only combined with a search, where the existing FilteredEmpty notice applies unchanged.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=2177528ee80c Agent vs Crewmate is not explained

Outcome: rebutted.
Third round on this seam (fac6d546661f last round). The two words are the kinds the merged job form (#22055) asks the user to choose between, with a hint for each; the list repeats the choice the user made.
Wording or a per-row explanation is the owner's call at review; this PR keeps the form's vocabulary.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=3cd6f1dc3d3b crewmate link is a guess

Outcome: rebutted.
The reviewer notes the reader's reading is correct. The name is dotted-underlined, brightens on hover, and carries the crewmate's face, the same pattern as other crewmate links in the app.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=ae6087bf48c9 "No owning session" under a crewmate's schedule

Outcome: rebutted.
That line is the Name column's existing owning-session fact (which chat session can manage the job through cron tools), unchanged by this PR and true for crewmate jobs created from the page.
Rewording it is a separate change to the Name column, outside this PR's goal.

@github-actions github-actions Bot added merge conflict Branch has merge conflicts with its base — author must resolve before merge and removed merge conflict Branch has merge conflicts with its base — author must resolve before merge labels Oct 19, 2026
@iamwhatever
iamwhatever force-pushed the feat/18772-schedule-crewmate-type branch from 8d28090 to 0a4d180 Compare October 19, 2026 15:30
@iamwhatever

Copy link
Copy Markdown
Contributor Author

Rebased again onto main 6257f69 because GitHub reported a second merge conflict. New head 0a4d180. The conflict was in website/src/i18n/en.context.json again: main added app.release_notes_not_published_yet at the end, where this PR's three pages.schedulePage.* keys sit; kept both. Locally: src/i18n and src/test/SchedulePage* 1105/1105 pass, tsc clean.

The previous head's red Harness Parity Gate (ACP_BACKENDS_MEMBER_CAPABILITIES [claude]: allowlisted but now proven) is not from this PR. It is frontend-only, and main 6257f69 fails the same gate.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention and removed readiness: action required A blocking check or review needs attention merge conflict Branch has merge conflicts with its base — author must resolve before merge readiness: checking Automated validation is still running labels Oct 19, 2026
@github-actions github-actions Bot added the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Oct 19, 2026
@iamwhatever

Copy link
Copy Markdown
Contributor Author

Rebased onto main ed975d7 because GitHub reported a merge conflict. New head 7745f8d. The conflict was again only in website/src/i18n/en.context.json: main added app.focus_mode_on_announce at the end, where this PR's three pages.schedulePage.* keys sit; kept both, no other change. This base includes #22178, which cleared the Harness Parity and Coverage reds on main. Locally: src/i18n and src/test/SchedulePage* 1105/1105 pass, tsc and eslint clean.

@iamwhatever
iamwhatever force-pushed the feat/18772-schedule-crewmate-type branch from 0a4d180 to 7745f8d Compare October 19, 2026 21:46
@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision and removed readiness: action required A blocking check or review needs attention merge conflict Branch has merge conflicts with its base — author must resolve before merge readiness: checking Automated validation is still running labels Oct 19, 2026
@iamwhatever

Copy link
Copy Markdown
Contributor Author

Review-ready at head 7745f8d (single commit on main ed975d7, which includes #22178). pr_status.py --reviewers gpt,opus reports CLEAN: 103 checks, 0 failing, PR Readiness passed, mergeable, GPT and Opus fresh with no blocking findings, no unresolved threads, no disposition-rule violations.

This is the last slice of #18772 (Closes #18772); slice 2 merged as #22055.

Open for the maintainer, no code change pending:

  • UX wording: whether the list should explain Crewmate vs Agent, a link tooltip naming the Crewmates page, and an "N of M shown" cue when filtered.
  • The frozen Problem section says member_id holds the name until the first run; it is captured at creation (6265010139).

Not merged, approved, or set to auto-merge.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=600892b71dd9 Crewmate vs agent does not say where output goes

Outcome: needs-a-decision.
The merged job form (#22055) says it in its switch hint: a crewmate's job runs in that crewmate's chat. The list cell does not repeat it.
@maintainer: same UX wording thread as earlier rounds (latest spans b4f6523ddd16, 4cb4239b10a4). Code stops here; the choice is yours at review.
Question: add "Runs in 's chat" to the Type cell tooltip (one catalog key in 13 locales), or leave the form as the one place that explains it?

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=e6120e5c455e "No owning session" sits under a named owner

Outcome: needs-a-decision.
"No owning session" is the existing Session column for jobs not tied to a chat session; this PR does not change that column.
@maintainer: same UX wording thread as earlier rounds (latest spans b4f6523ddd16, 4cb4239b10a4). Code stops here; the choice is yours at review.
Question: reword that existing label for crewmate rows in a follow-up, or keep it?

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=cc4e25d3bd8b crewmate name link was only guessed

Outcome: needs-a-decision.
The reader's guess was right: the link opens /members?member=<slug>.
@maintainer: same UX wording thread as earlier rounds (latest spans b4f6523ddd16, 4cb4239b10a4). Code stops here; the choice is yours at review.
Question: add "Open 's crewmate page" as the link tooltip, or keep the current one?

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=e186d9f0aff9 rows show raw ids before names load

Outcome: needs-a-decision.
True for the short time before GET /api/members resolves; the cell then switches to the face and name. The raw value is also what a failed roster shows, next to the Retry notice.
@maintainer: same UX wording thread as earlier rounds (latest spans b4f6523ddd16, 4cb4239b10a4). Code stops here; the choice is yours at review.
Question: show a skeleton in the cell while the roster loads, or keep the raw id?

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • span=d4d148934c29 re-capture shot-04 with "Backup check" in it

Outcome: wontfix.
The reader's worry is answered in code: the roster-failure path only swaps names for raw ids and adds the notice; no row is filtered out. reports a failed crewmate roster, keeps raw ids, and recovers on retry pins it.
I am not re-capturing the frame this round, so the screenshots stay as reviewed.

@iamwhatever

Copy link
Copy Markdown
Contributor Author

Yes. The slice plan agreed on #18772 (comment 6258847665) names "the 'by crewmate' filter" as its own piece.
Search matches any substring in the job name or display name, so "Radar" also hits a custom-agent job named "Radar triage". The filter matches the roster slug exactly and lists only crewmates that own a listed job.

@iamwhatever

Copy link
Copy Markdown
Contributor Author
  • CONCERNS punchline: nothing on screen says how Crewmate and agent differ

Answered with span 600892b71dd9: the job form's switch hint explains both kinds; whether the list repeats it is left for the maintainer.

@chenmingwei23
chenmingwei23 merged commit 36f9f02 into main Oct 20, 2026
103 checks passed
@chenmingwei23
chenmingwei23 deleted the feat/18772-schedule-crewmate-type branch October 20, 2026 00:35
@github-actions github-actions Bot removed the readiness: passed Eligible automated validation passed for the current revision label Oct 20, 2026
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.

Schedule page: a job opens a session on a template, or delivers to a crewmate

2 participants