cmux NEXT: long thoughts fold by default; Codex's 140px threshold is a choice by teamleaderleo · Pull Request #21661 · manaflow-ai/cmux

teamleaderleo · 2026-10-18T13:55:41Z

Stacked on #21660, which is stacked on #21656. This branch includes both. Review only this PR's own commits: e5b2e49db7d6 and 7da4f34c3a36. Land this last.

Long thoughts in the agent chat now fold by default, as Leo asked on 2026-10-18. agentPane.debug.foldLongThoughts defaults to on, so in a new agent tab a long thought streams in the seven-line live viewport (#21660). Once the agent moves on, it folds to "Thought for 12s" (#21656). Turning the toggle off in Debug Settings > Pages still keeps every thought in full.

What counts as long:

Live viewport on (the default): a finished thought taller than the 140 px viewport folds, so a thought never grows after it streams. Before this, the threshold was max(400 px, transcript height). In an 800 px pane, a thought between 140 and 800 px streamed clipped, then expanded nearly a screen when it ended.
Live viewport off: a new Debug choice, "Long thought fold threshold" (agentPane.debug.thoughtFoldThreshold), decides.
- viewport (the default) keeps Agent chat: long thoughts fold once done, behind a Debug toggle #20948's rule: taller than the transcript, never under 400 px.
- codex folds anything taller than seven lines (140 px). The app sets data-thought-fold-threshold="codex" on <html>.

Freeze WINDOW areas: Packages/macOS/CmuxNext (Debug tunables in AgentPaneTunables.swift, including a changed default).

Testing

bun test src/agent-session/acpmux/conversation/ThoughtRow.test.tsx passes, 19 tests. They include:
- In an 800 px pane, a 300 px thought clipped while streaming folds once done. This test fails without 7da4f34c3a36's change, as e5b2e49db7d6 has it. Next to it, a 140 px thought stays in full and a 141 px one folds.
- A thought over a minute reads "Thought for 1m 5s".
- With the live viewport off, the 400 px and transcript-height rule holds, and codex folds past 140 px.
bun run check in webviews/ passes, and lint-ios-package-conventions.sh --namespace-fix is clean.
Added but not run here (Linux, no Swift; CI runs it): AgentPaneThoughtTunablesTests.swift checks each tunable's defaultValue, including the fold being on, plus the threshold's key and its codex attribute.
CI skips the webview tests until the base-red check:text-focus on feat-cmux-next is fixed.
Localization: no new user-facing strings in the page.

Changelog

Changed: A long reasoning trace in agent chat folds to "Thought for 12s" once the agent moves on

🤖 Generated with Claude Code

A long thought folded once done (#20948, Debug toggle agentPane.debug.foldLongThoughts) now reads how long the agent thought, from its first chunk to its latest, clamped at zero for out-of-order events. An undated thought, or one under a second, reads "Thought". #20948's "Thought" plus first line stays as the Debug choice agentPane.debug.thoughtHeader = preview, which the app sets on <html> as data-thought-header. New string tools.thoughtFor in all 21 pane locales (split from #20963). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

With the long-thought fold on (agentPane.debug.foldLongThoughts), a thought taller than 140px that is still streaming reads "Thinking" and holds a fixed 140px viewport (Codex's 8.75rem) with its newest line at the bottom and older lines fading out over 16px. Once done it folds as before; the fold measures the text's whole height, not the clip. Debug toggle agentPane.debug.liveThoughtViewport, on by default; off sets data-live-thought-viewport="off" and the thought streams in full (split from #20963). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…a choice agentPane.debug.foldLongThoughts is now on by default (Leo 2026-10-18): a finished long thought folds to "Thought for 12s" in new agent tabs. Off still keeps every thought in full. A new Debug choice, agentPane.debug.thoughtFoldThreshold, picks what "long" means: viewport (default, #20948: taller than the transcript, never under 400px) or codex (taller than seven lines, 140px), which the app sets on <html> as data-thought-fold-threshold="codex" (split from #20963). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

github-actions · 2026-10-18T13:55:49Z

Receipt for 81b7d63 (this PR's head; a new push replaces it)

Visual review: HQ feed card · gallery before/after runs
Checks: this head
App of this head (app.zip, kept 1 day): gh workflow run reload-build.yml -R manaflow-ai/cmux --ref feat-cmux-next -f tag=pr21661 -f ref=split/20963-thought-defaults -f platform=macos -f backend_mode=local

coderabbitai · 2026-10-18T13:55:49Z

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository: manaflow-ai/cmux/.coderabbit.yaml
Review profile: ASSERTIVE
Plan: Advanced
Run ID: 5c8215f3-7e9f-47f4-8f55-057ea3340fe1

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

🔍 Trigger review

Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

github-actions · 2026-10-18T13:58:38Z

red: web / Web status; do not merge until green; see lane-rules
red: web / react-apps-check; do not merge until green; see lane-rules
red: web / react-apps-check (2); do not merge until green; see lane-rules

teamleaderleo · 2026-10-18T13:58:50Z

Bead cx-vedm2 created for this PR by the cmux coordinator.

github-actions · 2026-10-18T13:59:01Z

Unclear: web / react-apps-check (2) failed: no known signature; failed step: Test webviews.

CI failed on 81b7d63930 (run 39173320257 attempt 2): 2 unknown.
macOS jobs did not run: compile and app tests were skipped: Fast static checks skipped.

Job	Verdict	Why
web / react-apps-check (2)	unknown	no known signature; failed step: `Test webviews`
web / react-apps-check	unknown	no known signature; failed step: `Check webview test shards`

Not re-run automatically: web / react-apps-check (2), web / react-apps-check are not machine failures.

Written by scripts/ci/classify_failures.py (ci-failure-attribution.yml); signatures are its SIGNATURES table. A machine verdict is the runner's fault, not this PR's; yours means the failing file is one this PR changes, also red on main that main's latest full suite fails the same way, seen on other PRs that it failed on another pull request's run lately.

github-actions · 2026-10-18T13:59:32Z

CI fast guards failed on 81b7d63930 (https://github.com/manaflow-ai/cmux/actions/runs/39173320257). It does not block the merge; a red guard merged into main breaks it for every open PR.
The log named no failed step; see the run.

Agents: python3 scripts/ci/guard_attribution.py fix applies the mechanical fixes locally. This comment is updated in place on each push.

teamleaderleo · 2026-10-18T14:03:39Z

react-apps-check (1) and (2) fail on the same base-red check:text-focus problem as #21656 and #21660 (CmuxNextDaemonRecovery/DaemonRecoveryControls.swift:98, from #20765). That failure stops the job before Test webviews, so the webview tests haven't run here and need a rerun once the base is fixed. The other 92 checks pass.

🤖 Generated with Claude Code

The fold threshold was max(400px, transcript height), so in an 800px pane a thought between 140px and the pane's height streamed clipped to the live viewport and then grew to nearly a screen when it ended. While agentPane.debug.liveThoughtViewport is on, a finished thought taller than the 140px viewport now folds, so nothing grows after it streams; agentPane.debug.thoughtFoldThreshold counts only with the live viewport off. Review finding on #21661. Also: a test for the minutes label ("Thought for 1m 5s"), and the Swift tests check each tunable's defaultValue rather than its stored value. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

cursor · 2026-10-18T14:12:41Z

Bugbot needs on-demand usage enabled

Bugbot uses usage-based billing for this team and requires on-demand usage to be enabled.

A team admin can enable on-demand usage in the Cursor dashboard.

teamleaderleo · 2026-10-18T15:38:30Z

Closing and reopening to start a fresh CI run on the current base: #21679 fixed the base check:text-focus red, and a plain rerun would reuse the old merge commit. Head unchanged.

…s check:text-focus fix)

blacksmith-sh · 2026-10-18T16:04:34Z

Found 4 test failures on Blacksmith runners:

Failures

Test	View Logs
`/tmp/tmp0t_8pcg9/summary-fail.json (live):/summary-fail.json (live):`	View Logs
`/tmp/tmph7ufrss2/summary-missing.json (live):/summary-missing.json (live):`	View Logs
`test/narrow-layouts.test.ts/narrow-layouts.test`	View Logs
`test/narrow-layouts.test.ts/narrow-layouts.test`	View Logs

^{Need help on this PR? Tag @codesmith-bot with what you need.}

… as a squash) # Conflicts: # Packages/macOS/CmuxNext/Sources/CmuxNextAgentPane/AgentPaneTunables.swift # Packages/macOS/CmuxNext/Tests/CmuxNextAgentPaneTests/AgentPaneThoughtTunablesTests.swift # webviews/src/agent-session/acpmux/conversation/ThoughtRow.test.tsx # webviews/src/agent-session/acpmux/conversation/ThoughtRow.tsx # webviews/src/agent-session/acpmux/conversation/conversation.css

cursor · 2026-10-18T21:22:34Z

Bugbot needs on-demand usage enabled

Bugbot uses usage-based billing for this team and requires on-demand usage to be enabled.

A team admin can enable on-demand usage in the Cursor dashboard.

… as a squash) # Conflicts: # Packages/macOS/CmuxNext/Sources/CmuxNextAgentPane/AgentPaneTunables.swift # Packages/macOS/CmuxNext/Tests/CmuxNextAgentPaneTests/AgentPaneThoughtTunablesTests.swift # webviews/src/agent-session/acpmux/conversation/ThoughtRow.test.tsx # webviews/src/agent-session/acpmux/conversation/ThoughtRow.tsx

cursor · 2026-10-18T22:41:19Z

Bugbot needs on-demand usage enabled

Bugbot uses usage-based billing for this team and requires on-demand usage to be enabled.

A team admin can enable on-demand usage in the Cursor dashboard.

github-actions · 2026-10-19T18:27:09Z

Shipped in:

cmux NEXT nightly 1.0.0-nightly.3929331184001 (83d36a611f)
cmux NEXT nightly 1.0.0-nightly.3930148023001 (9d185ebb26)

…rows Base built its own design for the transcript features this PR touched: long thoughts fold behind Debug toggles (#20948, #21656, #21661) and Codex-style tool runs build on ToolRun (#20969). This merge takes base's side in every conflict (ThoughtRow, TurnRows, turns, direct, model, the permission panel, the transcript gallery), keeps ToolRun.tsx and toolRunSummary.ts, drops the generated string tables (build output since #21970), and adds this PR's new catalog strings beside base's. The redesign on base's architecture follows in the next commits. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…or lists every step (red) On base's ThoughtRow fold (#20948, #21656, #21661): by default every done thought folds to its "Thought for Ns" line, a streaming one reads "Thinking", and a thought sent as one chunk lasts until the next step starts. The Debug choices viewport and codex keep folding only long ones. On base's tool runs (#20969): a live turn's Codex-style run still folds under ToolRun, while an open "Worked for" lists every step unless the new Debug toggle agentPane.debug.workedForFoldsRuns folds it again. transcript-live-runs gives its calls ids apart from their titles: a title equal to the call's id now reads as no title (toolKinds.ts isCallId). agent-pane-step-rows goes back to base's: without the app's fold attribute a thought shows in full. Red at this commit: bun test ThoughtRow.test.tsx TurnRows.test.tsx transcript-live-runs transcript-row-thought transcript-worked-steps: 25 pass, 10 fail. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

teamleaderleo and others added 3 commits October 18, 2026 13:48

teamleaderleo mentioned this pull request Oct 18, 2026

cmux NEXT: fold a long reasoning trace the way Codex does #20963

Closed

3 tasks done

teamleaderleo closed this Oct 18, 2026

teamleaderleo reopened this Oct 18, 2026

Merge feat-cmux-next into split/20963-thought-defaults (pick up #21679'…

38c43db

…s check:text-focus fix)

teamleaderleo temporarily deployed to cmux-tui-merge-tree October 18, 2026 15:54 — with GitHub Actions Inactive

teamleaderleo merged commit 5e7966c into feat-cmux-next Oct 18, 2026
157 of 173 checks passed

teamleaderleo deleted the split/20963-thought-defaults branch October 18, 2026 23:23

teamleaderleo mentioned this pull request Oct 20, 2026

cx-bwbfv: transcript rows say what happened (thoughts, message status, tool summaries, full Worked for) #20456

Closed

Conversation

teamleaderleo commented Oct 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Testing

Changelog

Uh oh!

github-actions Bot commented Oct 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

coderabbitai Bot commented Oct 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review skipped

Uh oh!

github-actions Bot commented Oct 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

teamleaderleo commented Oct 18, 2026

Uh oh!

github-actions Bot commented Oct 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Oct 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

teamleaderleo commented Oct 18, 2026

Uh oh!

cursor Bot commented Oct 18, 2026

Bugbot needs on-demand usage enabled

Uh oh!

teamleaderleo commented Oct 18, 2026

Uh oh!

blacksmith-sh Bot commented Oct 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Failures

Uh oh!

cursor Bot commented Oct 18, 2026

Bugbot needs on-demand usage enabled

Uh oh!

cursor Bot commented Oct 18, 2026

Bugbot needs on-demand usage enabled

Uh oh!

Uh oh!

github-actions Bot commented Oct 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

teamleaderleo commented Oct 18, 2026 •

edited

Loading

github-actions Bot commented Oct 18, 2026 •

edited

Loading

coderabbitai Bot commented Oct 18, 2026 •

edited

Loading

github-actions Bot commented Oct 18, 2026 •

edited

Loading

github-actions Bot commented Oct 18, 2026 •

edited

Loading

github-actions Bot commented Oct 18, 2026 •

edited

Loading

blacksmith-sh Bot commented Oct 18, 2026 •

edited

Loading

github-actions Bot commented Oct 19, 2026 •

edited

Loading