Skip to content

fix(mobile): platform notices read as named lines, and Crew notices as seat lines - #981

Merged
Jacksondr5 merged 3 commits into
j5/mainfrom
j5code/mobile-notice-lines
Nov 9, 2026
Merged

Jacksondr5 merged 3 commits into
j5/mainfrom
j5code/mobile-notice-lines

Conversation

@Jacksondr5

@Jacksondr5 Jacksondr5 commented Nov 9, 2026 •

Copy link
Copy Markdown
Owner

Problem

Mobile hadn't picked up #916's shared classifier. A silence line showed participant ids ("agent:j5:a2a:… turn ended without replying"). Crew notices and the delivery notices mobile couldn't read showed as raw cards of XML. A playbook step or native question showed as the person's own message on the right. Closes #918, and covers piece 2 of #836 (notice lines in the Captain's timeline).

What changed

Mobile now draws the timeline as web does (register D38):

  • Platform notices. Each notice with a summary is one grey line on the left: an icon for its kind, the summary with participants named, and the time. A line with a detail opens to it. This covers silence, not-delivered, exchange-dropped, exchange-withdrawn and the handoff reminder.
  • Crew notices. These use crewNoticeLines, crewApprovalSummary and crewDeclinedSummary from client-runtime:
    • The approval line opens to what the person changed, and to a link for each seat.
    • The decline line opens to what was requested.
    • Each seat that needs the person gets its own line, red or amber, which opens to the error and "Open seat thread".
    • What a Crew member was told is a line ("Introduction from J5") that opens to the text.
  • Hidden Crew notices. A clean launch report or clean finish takes no row, as on web (case 104). The filter runs after deriveThreadFeedPresentation, because that function also treats user messages as turn boundaries. Filtering earlier would merge a Captain's turn into the one before it. isHiddenCrewNotice and the filter moved from web into client-runtime crewNotices.ts, and web wraps the shared version.
  • Prompts. A playbook step or native question goes to upstream's Automation row under J5's label, preview and icon. That row now sits on the left with a border instead of the user bubble's fill, for upstream's automation runs too, as on web.
  • Unread notices. A notice nothing can read goes to upstream's row, on the left when the platform wrote it. The raw card is gone.

UI changes

Android. Before is trunk 067d732e1a; after is ff77dd9a42, with the same stored events, theme and viewport. The Automation, playbook-step and question rows were re-captured at 53624d47ef after the label fix.

State Before After
A stopped agent's notice, with its name A stopped agent's notice, with its name, before A stopped agent's notice, with its name, after
Crew approval Crew approval, before Crew approval, after
Changed roster, opened Changed roster, opened, before Changed roster, opened, after
Declined Crew roster Declined Crew roster, before Declined Crew roster, after
Seat introduction and playbook step, folded Seat introduction and playbook step, folded, before Seat introduction and playbook step, folded, after
Seat introduction, opened Seat introduction, opened, before Seat introduction, opened, after
Clean seat finish hidden above its Captain turn Clean seat finish hidden above its Captain turn, before Clean seat finish hidden above its Captain turn, after
Placed agent's question, folded Placed agent's question, folded, before Placed agent's question, folded, after
Scheduled Automation row, folded Scheduled Automation row, folded, before Scheduled Automation row, folded, after
Scheduled Automation row, opened Scheduled Automation row, opened, before Scheduled Automation row, opened, after

After only:

A stopped agent's notice, opened, after

Crew approval opened to its seat links, after

Declined Crew opened to the requested seats, after

Playbook step, opened, after

Placed agent's question, opened, after

Ordinary agent letter, folded, after (regression check)

Ordinary agent letter, opened, after (regression check)

Live testing

Pass for every state exercised, by the wave 4 Android tester (Sol 6.1), with real Claude Haiku turns. Tested at ff77dd9a42:

  • Crew approval, with its seat links and a changed roster;
  • a declined roster;
  • seat introductions;
  • a real Crew playbook step;
  • hidden clean finishes, checked against the Captain turn's grouping;
  • a real placed agent's native question;
  • a real scheduled automation;
  • a stopped agent's silence notice, named, with its detail opening;
  • ordinary agent letters and the person's own messages, as regression checks.

That run showed the Automation row's label truncating ("Automati…", "Playbook step …"). 53624d47ef fixed it: a J5 prompt's pill is full width with its label capped, and upstream's label never shrinks. The tester re-tested those rows at 53624d47ef and they pass. Nothing else changed between the two heads.

Not exercised:

  • a plain "ended without replying" silence, because Haiku replied and closed its Exchange each time;
  • the not-delivered, Exchange-dropped and Exchange-withdrawn notices;
  • failed and never-created seat lines;
  • dark theme, iOS, desktop and remote/relay.

These rows are static, so there is no video.

Upstream impact

  • apps/mobile/src/features/threads/ThreadFeed.tsx:
    • case 97: the notice call gets the timeline, and ThreadAutomationMessageRow gets prompt, moves left and loses its fill;
    • case 104: presentedFeed is filtered after derivation.
  • Restyling the Automation row is D38's recorded divergence. D38's line "Mobile draws silences and a labelled card for J5's other notices until it follows" is now stale. Its rewrite goes in the wave 4 docs PR.

Checklist

  • One concern: the description has no "also"
  • Tests cover the changed behavior: timelinePresentation.test.ts (names, side, prompt) and crewNotices.test.ts (the filter keeps turn boundaries), plus the seam rows
  • UI changes: before/after above
  • Live testing: the section above says what ran live and what did not
  • Upstream-owned files: each one is recorded in FORK.md (cases 97 and 104) in this PR
  • Upstream product: the Automation row's move is D38, already decided; its mobile wording goes in the wave 4 docs PR
  • Surfaces: mobile is the client here; web shares the moved filter unchanged; the A2A timeline isn't provider-shaped
  • Docs: D38's mobile sentence goes in the wave 4 docs PR

Built by Claude Opus 5.5 in Claude Code, as builder-timeline in the mobile catch-up Crew.

🤖 Generated with Claude Code

…s seat lines

Mobile follows #916's classifier (#918, #836 piece 2): each J5 notice is one grey
line on the left that opens to its detail, with participants named; a Crew notice
shows its approval or decline and one line per seat that needs the person, each
opening the seat's thread; playbook steps and native questions take upstream's
Automation row under J5's heading. Crew notices with nothing for the person are
dropped after the feed is derived, through the filter web now shares.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@vercel

vercel Bot commented Nov 9, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated
j5-code Ready Ready Preview Nov 9, 2026 8:21am UTC

Request Review

@coderabbitai

coderabbitai Bot commented Nov 9, 2026 •

Copy link
Copy Markdown

Explain Change →

No actionable comments were generated in the recent review.

Walkthrough

Mobile J5 timelines now present classified prompts, notices, and Crew details. A shared client-runtime helper filters hidden Crew notices from derived rows, and web and mobile feed code use that helper.

Changes: J5 timeline presentation and filtering

Layer / File(s) Summary
Shared Crew notice filtering
packages/client-runtime/src/j5/crewNotices.ts, packages/client-runtime/src/j5/crewNotices.test.ts, apps/web/src/j5/crew/crewNotices.logic.ts, apps/web/src/j5/crew/crewNotices.logic.test.ts, apps/mobile/src/features/threads/ThreadFeed.tsx, apps/mobile/src/j5/upstreamSeams.test.ts, FORK.md
The client runtime identifies and filters hidden Crew notices, preserves the input array when no rows are removed, and caches filtered results by input-array identity. Web delegates to the shared helper. Mobile filters after appending guest pending messages.
Mobile prompt and notice classification
apps/mobile/src/j5/a2a/timelinePresentation.ts, apps/mobile/src/j5/a2a/timelinePresentation.test.ts
Mobile extracts J5 prompt presentation, uses system authors as a left-side fallback, and classifies notices using participant labels. Tests cover prompt and notice presentation results.
Mobile timeline and automation rendering
apps/mobile/src/j5/a2a/timeline.tsx, apps/mobile/src/features/threads/ThreadFeed.tsx, apps/mobile/src/j5/upstreamSeams.test.ts, FORK.md
Timeline notices render summaries, details, and Crew seat links. Automation rows use J5 prompt labels, previews, icons, and assistant markdown styling.

Estimated code review effort: 3 (Moderate) | ~25 minutes

Possibly related issues

  • Jacksondr5/j5code#337 — The PR changes mobile platform notices from unlabeled user bubbles to named notice lines and renders Crew notices as seat lines.
  • Jacksondr5/j5code#911 — The shared filter hides clean Crew notices in mobile and web timelines; this PR does not change server-side Captain wake behavior.

Possibly related PRs

  • Jacksondr5/j5code#916 — The PR consumes the shared classifier contracts and Crew notice helpers added by this PR.

Suggested reviewers: bryantderosier

Merge Risk: Minimal · up to 53624

This change presents mobile platform and Crew notices as readable named lines and routes unreadable notices to the upstream row. No concrete merge-blocking risk was found. Android UI verification is still pending with the tester.

Pre-merge checks: Passed 3 · Inconclusive 1

Failed Failed checks: 1 inconclusive

Check name Status Explanation Resolution
Out of Scope Changes check Inconclusive Most changes connect to #918 or to the notice-line work the PR cites (#836 piece 2):
- ThreadFeed.tsx case 97 routes J5 notices and prompts through renderJ5TimelineNotice, which #918 requires.
- `…
Provide the D38 decision text, or the wave 4 product record, that covers upstream's non-J5 automation runs. If D38 does not cover them, limit the Automation row restyle to J5 prompts and note the deviation in the PR. Confirm that the hidden…
Passed Passed checks: 3 passed

Check name Status Explanation
Title check Passed The title names the main change: mobile timeline notices are now shown as named lines, and Crew notices as seat lines. The description and objectives confirm this is the primary focus. The title uses …
Linked Issues check Passed Only #918 is a directly linked active issue. Its coding requirements are covered at the reviewed head (53624d4):

1. New fields and kinds [#918]: apps/mobile/src/j5/a2a/timelinePresentation.ts reads…
Description check Passed The description follows the repository template. It has Problem, What changed, UI changes, Live testing, Upstream impact, and Checklist sections, and each is substantively filled in. The Problem secti…
Full details: Title check

Explanation

The title names the main change: mobile timeline notices are now shown as named lines, and Crew notices as seat lines. The description and objectives confirm this is the primary focus. The title uses a conventional fix(mobile) prefix and stays within one sentence. Its 81-character length is slightly long but still readable.

Full details: Linked Issues check

Explanation

Only #918 is a directly linked active issue. Its coding requirements are covered at the reviewed head (53624d4):

  1. New fields and kinds [#918]: apps/mobile/src/j5/a2a/timelinePresentation.ts reads presentJ5UserMessage for notices and reads notice.detail and notice.participantIds. Prompts are read for label and preview (j5Prompt). timeline.tsx maps silence, not-delivered, exchange-dropped and exchange-withdrawn to icons. handoff-nudge has no entry and falls back to bell.badge. playbook-step and native-question use J5_PROMPT_ICONS in the Automation row.
  2. Names, not ids [#918]: presentJ5TimelineNotice calls nameJ5NoticeParticipants(text, notice.participantIds, labels), and labels comes from the identity read on J5Timeline. Participant ids are replaced with names in both summary and detail.
  3. Crew lines [#918]: timeline.tsx uses crewNoticeLines, crewApprovalSummary and crewDeclinedSummary from @t3tools/client-runtime/j5/crew-notices. Each seat line is red or amber from line.tone. Its body text is line.detail, and its link is "Open seat thread".

The issue states that queued-removed and rolled back wording already needed no code. The PR does not change that behavior.

Automated tests: #918 does not name specific tests. timelinePresentation.test.ts and upstreamSeams.test.ts cover names, side and prompt behavior, as the PR states.

Limit: I did not read the client-runtime crew-notices module. The path packages/client-runtime/src/j5/crew-notices.ts is absent at the head. I therefore could not confirm the exact reason and detail field mapping of crewNoticeLines. The mobile code uses reason in the line summary and detail in the body, which matches the issue text.

Full details: Description check

Explanation

The description follows the repository template. It has Problem, What changed, UI changes, Live testing, Upstream impact, and Checklist sections, and each is substantively filled in. The Problem section names the defect and links #918 and #836. What changed explains the shape of the fix and why the filter runs after deriveThreadFeedPresentation. UI changes gives a before/after table with Android captures. Live testing states the verdict, the tested head, the one-head re-test, and what was not exercised. Upstream impact lists the two upstream-owned ThreadFeed.tsx edits with their FORK.md cases (97 and 104) and the D38 divergence. The Checklist is ticked, except one item marked as pending: the D38 docs sentence is deferred to a separate docs PR. The description also ends with the model and harness, as the template requests. Minor gaps: the Upstream impact section does not link a human decision or docs/j5/product/upstream.md register entry for the Automation row restyle, though the description says D38 already records that decision. The description also says 'Built by Claude Opus 5.5', which is a model name the template does not need, but that does not affect completeness.

Full details: Out of Scope Changes check

Explanation

Most changes connect to #918 or to the notice-line work the PR cites (#836 piece 2):

  • ThreadFeed.tsx case 97 routes J5 notices and prompts through renderJ5TimelineNotice, which #918 requires.
  • timeline.tsx and timelinePresentation.ts implement the name, field and kind changes that #918 lists.
  • FORK.md records the upstream-owned cases 97 and 104 that the PR edits.
  • The test changes support the mobile behavior.
  • Moving isHiddenCrewNotice and withoutHiddenCrewNotices to client-runtime, with the web wrapper, supports one shared filter for mobile and web. It is a refactor that the PR justifies.

One change lacks a connection in the linked issue text. The Automation row restyle applies to all upstream automation runs, not only J5 prompts. The row moves left, loses its user-bubble fill and gains a border. #918 does not request this. The PR cites D38 as the decision record, but that record is not in the evidence I could read. The PR also says the change matches web. I could not verify that web or D38 covers upstream's non-J5 automation runs. The hidden Crew notice filter (case 104) and the routing of unreadable notices to the upstream row are behavior changes that the PR links to #836 piece 2. This link is plausible, but #918 does not name it.

Resolution

Provide the D38 decision text, or the wave 4 product record, that covers upstream's non-J5 automation runs. If D38 does not cover them, limit the Automation row restyle to J5 prompts and note the deviation in the PR. Confirm that the hidden Crew notice filter and the unreadable-notice fallback belong to #836 piece 2.

Finishing Touches

Generate docstrings

  • Commit to this branch
  • Create a new PR
Generate unit tests (beta)

  • Commit to this branch
  • Create a new PR
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Warning

Some tools did not complete. Review the errors below.

LanguageTool

FORK.md

LanguageTool checks are incomplete because the per-file request limit of 5 was reached. Remaining text was skipped; findings from completed checks are retained.

Recent review info

Run configuration

Commits

Reviewing files that changed from the base of the PR and between 067d732 and 53624d4.

Files selected for processing: 10

  • FORK.md
  • apps/mobile/src/features/threads/ThreadFeed.tsx
  • apps/mobile/src/j5/a2a/timeline.tsx
  • apps/mobile/src/j5/a2a/timelinePresentation.test.ts
  • apps/mobile/src/j5/a2a/timelinePresentation.ts
  • apps/mobile/src/j5/upstreamSeams.test.ts
  • apps/web/src/j5/crew/crewNotices.logic.test.ts
  • apps/web/src/j5/crew/crewNotices.logic.ts
  • packages/client-runtime/src/j5/crewNotices.test.ts
  • packages/client-runtime/src/j5/crewNotices.ts

Included review availability: This review used your included allowance. 1 included review remains after this review. Your included PR review attempts over the past 7 days set your current allowance at 4 reviews per hour.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

@Jacksondr5 Jacksondr5 left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

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

Combined review: reviewer-opus and reviewer-sol (Claude Opus 5.5 and GPT Sol 6.1), head ff77dd9

Both of us reviewed the PR independently at b7f4ccf. Neither found anything that blocks merge. ff77dd9 only adds case 104 to the ThreadFeed.tsx file-index row, which fixes the ledger omission reviewer-sol noted.

Your three questions:

  1. presentedFeed identity. withoutHiddenCrewNotices returns its input when nothing is hidden, and caches the filtered array per input array, so its output is exactly as stable as the input. appendGuestPendingMessages returns its input when there's no guest pending, and the derivation goes through feedCache. The filter runs after deriveThreadFeedPresentation, so a hidden notice still bounds the turn it started. The lift from web is line for line, with a messageOf accessor added. Web's wrapper keeps its signature, and the "keeps the rows a hidden notice bounded" test moved with it.
  2. Upstream's automation row against D38. It now sits on the left with a border and no fill, for scheduled runs too, as web's case 7 draws it. That reads right under D38: anyone but the person folds on the left. The stale D38 sentence goes in the wave 4 docs PR, as the PR says.
  3. Surfaces.
    • Crew notices use the shared summaries and lines, as web does.
    • Playbook steps and native questions route through the automation pool, because getItemType reads resolveJ5FeedMessage.
    • Notice participants are named through the existing identity read.
    • A system message nothing reads returns null and stays upstream's row, on the left; the old raw cards are gone, as on web.
    • The icon names are typed as AppSymbolName, so typecheck guards the Android map.
    • Seat links go through useOpenLinkedThread.

Tests. 63 runtime classifier and filter tests pass, and 157 mobile presentation, seam, filePreview and dependency-graph tests.

The panel has no open code findings. Before merge it still needs the Android captures, the live run and green J5 CI. The PR says honestly that all three are pending.

@github-actions github-actions Bot added vouch:trusted PR author is trusted by repo permissions or the VOUCHED list. size:L 100-499 effective changed lines (test files excluded in mixed PRs). labels Nov 9, 2026
…s summary

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

@Jacksondr5 Jacksondr5 left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

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

Follow-up review: reviewer-opus, head 53624d4

The pill fix for the tester's truncated labels ("Automati…", "Playbook step …") is right.

  • J5 prompts. A playbook step or native question gets a full-width pill even when folded (width: bubbleMaxWidth), and its label is capped at max-w-[55%], so the summary's min-w-0 shrink takes the rest. That's web's case 7 shape.
  • Upstream's automation runs. They keep a content-sized pill, and their short "Automation" label no longer shrinks.
  • Records. The case 97 sentence is updated.

No open code findings from me. The tester's re-captured pills are the proof.

This branch was successfully deployed

1 active deployment
Preview — 53624d47 Deployed Nov 9, 2026 by vercel[bot]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

size:L 100-499 effective changed lines (test files excluded in mixed PRs). vouch:trusted PR author is trusted by repo permissions or the VOUCHED list.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Mobile: follow the shared A2A classifier after #916

1 participant