Repository navigation
cmux NEXT New Tab: the model menu keeps the harness rail once the page's chat starts behind it - #28966
Merged
Conversation
…e's deferred chat starts A pick or typed prompt on a New Tab starts its deferred chat behind the page. That session's start row counted as a message, so ComposerPickers locked the harness (harnessFixed) and the model menu dropped the rail to one harness's plain model list. A New Tab never fixes the harness now, and a session start row alone no longer counts as a message. Leo, 2026-10-31: "IM SAYING I CANT FUCKING SEE ANY HARNESS" Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-d750cf10922c785efa64
Contributor
|
cmux status (one comment, edited in place; each line is updated by the job that owns it) Review App of this head (app.zip, kept 1 day)
Gallery before/after for
States
Pick other states with a Shipped in:
|
Collaborator
Author
|
Owned by an agent session · session |
Collaborator
Author
|
Review (codex gpt-6-astra high, router-review): BLOCK at a654cf5 BLOCK at a654cf5 Visual verdict: Incomplete—no after screenshots were supplied.
No additional correctness defects found by static inspection. Tests could not run because Bun and dependencies are unavailable. |
teamleaderleo
temporarily deployed
to
ci-routing-agent
October 31, 2026 05:54 — with 
GitHub Actions
Inactive
GitHub Actions
Inactive
Collaborator
Author
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent-Session: https://cmux.com/deeplink/agent?id=aid-d750cf10922c785efa64
teamleaderleo
temporarily deployed
to
ci-routing-agent
October 31, 2026 06:27 — with
GitHub Actions
Inactive
GitHub Actions
Inactive
Collaborator
Author
|
Review (codex, router-review): SHIP at 1342cd3 Follow-ups: 1. Each lands as an issue in this repo that links this receipt; the lane that opened the PR fixes it forward. SHIP at 1342cd3 Visual verdict: no images; Leo tries the build
Follow-ups:
|
This branch was previously deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Leo, on the cmux NEXT New Tab composer (2026-10-30/31), verbatim:
His screenshots: with the New Tab field empty, the model menu shows the harness rail. With "helo" typed, the same menu shows only Opus 5.5 / Fable 5.1 / Sonnet 5.5 / Haiku 5.5 / More models / Fast mode, and no harness at all.
Cause
ComposerPickersfixes a chat's harness once it has a message:locked = !harnessSwitch && summary.harness && (turnCount > 0 || snapshot.rows.length > 0).lockedbecomesharnessFixed, andModelPicker.quickFamiliesthen returns no rail. A pick or typed prompt on a New Tab starts the page's deferred chat behind it (chat.new {deferred}). That session draws itssessionStartrow before any prompt is sent, sorows.length > 0read the New Tab as a started chat and the rail went away.A jsdom probe at tip (
815c758fd30) throughNewTabScreen+NewTabComposerChips:turnCount: 0, no rowsturnCount: 0, only itssessionStartrowRuled out for Leo's case:
current?.folder(#24942) hides the rail too, butcmux harness list --folder ~/Projects/leo-workspace/briefs/2026-10-20-gui-pushon cosmic-gray lists no folder profiles. The #26969 folding leaves several tabs, so it does not empty the rail.Fix
DefaultComposerChipspassesstartsChattoComposerPickers. A New Tab never locks the harness, because no prompt has gone from it.sessionStartrow alone is not a message, so a fresh chat with no prompt yet also keeps its harness switch. A chat with a real message keeps its own harness's list as before (Leo 2026-10-13, and the Fork-to rail from 2026-10-22 is unchanged).This changes nothing in the picker's design (design/model-picker.md): the rail placement, the symbols work in #27889 and the folder-profile rule from #24942 stay as they are.
Tests
webviews/test/new-tab-harness-rail.test.tsx(jsdom, realNewTabScreen/NewTabComposerChips/ChatComposerChips):Evidence on cmux-ci (Linux pool): with the fix, the new test plus the neighbouring picker suites (new-tab-deferred-pickers, new-tab-ask-model, model-menu-started-chat-fork, model-menu-compact, signed-out-rail, QuickModelMenu) pass 117/0, and
bun run typecheckreports no errors. With the fix reverted, the two middle tests fail and the other two pass.Screenshots
Waived by Leo for this PR, verbatim (2026-10-31, relayed by claude-sr2-coord): "WE DONT NEED FUCKIGN SCREENSHOTS JUST LET ME TRY THE SHIT OUT". The rail this brings back is the existing one, unchanged; Leo checks it in the nightly himself. Reviewers: the screenshot and viewed-the-after-image requirements do not apply to this PR, so judge it on correctness.
Prior art acknowledged: #27889 (symbols on top, open, separate), #28050 / #27692 (per-harness tabs, not landing), #24942 (folder profiles off the rail), #26969.
Agent session: agent session