The persisted draft now records the leading command chip's claim token
(DraftSnapshot.command). Restoring the draft asks the trigger pipeline's
new resolveClaim, which polls the optional InputTriggerSource.matchClaim
hook: the command source awaits its catalog and answers only a
leadingInput claim, never opening a popup, running an action, or
executing the command. The input rebuilds the chip once the claim
resolves, only at the revision the restore produced and merged into the
seed's history entry; otherwise the literal token stays and Enter still
adjudicates it.
The client tsdown presets found a package's own manifest by name: a glob of
every packages/*/*/package.json, parsed until the name matched. tsdown's
native config loader imports each package config without module cache, so
the module-level cache never hit and all 82 configs repeated the glob
(40-125 ms each) synchronously on tsdown's main thread, where it delayed
every other package build. It was 34.5% of the client pass's main thread.
- clientBundle, staticLinked and clientLibrary take the package directory
(import.meta.dirname at every call site) and read <dir>/package.json
once; the package name, loader id and externals come from that manifest.
- A missing, unreadable or nameless manifest throws an actionable error.
- The glob, the name search and the module-level caches are removed.
tsdown client pass 12.78 s -> 8.73 s (median of 3 interleaved runs each,
-31.7%); all 11,566 lib outputs byte-identical; all 82 configs resolve the
same externals and banners.
Refs #5852
TypeScript builds a synthetic signature for every JSX intrinsic element and
computes a module specifier for @types/react (index and jsx-runtime) per
importing file. computeModuleSpecifiers visits every symlinked location of
that file and scans all 491 tsconfig paths for each; pnpm gave 65 packages
their own node_modules/@types/react link. Client test-aggregate Check
followed 26.9 + 0.494 * N seconds for N locations.
- Remove @types/react and @types/react-dom from all 67 workspace manifests
and declare them once in the root devDependencies; packages keep declaring
the react and react-dom they import (ui-slots gains react).
- verify-package-dependencies rejects either package in any workspace
manifest and requires the root declaration, so a new package cannot add
the link back.
- packages/client/AGENTS.md states the rule.
Client aggregate Check 59.1 s -> 26.5 s; the 76 client-only package
projects 54-57 s -> ~44 s wall. File set (2,613) and diagnostics unchanged.
Refs #5852
Delete @deepseek-ai/dsh-invariants, every package's ./invariant companion
and its tests, the invariant gates and Vitest setup, the scoped-event
resolver generator, and the INVARIANT-coded listener rethrow paths.
Update docs, READMEs, Agent Notes, and add an upgrade guide.
* feat(i18n): key translation pairing records by heading section
Replace the whole-file blob hashes in *.i18n.yaml with one entry per
heading section that still holds language-specific content, keyed by the
English heading-slug path. Blocks identical across the pair (generated
table rows, code blocks, shared regions) stay out of the hash, so edits to
different sections and identical regenerations merge with Git's default
text merge, including on GitHub.
Remove the dsh-translation-pairing merge driver, its resolver command and
installer wiring; gen-translation-brief recovers last-confirmed text from
commit history. Migrate all active records, archive the superseded Agent
Note, and record the replayed conflict reduction in the new one.
* test(i18n): refresh translation prompt snapshot for the terminology note
* feat(i18n): exclude only code blocks and generated regions from pairing records
Replace byte-equality masking with an explicit rule: a record hash skips
fenced code blocks and BEGIN/END GENERATED regions, the two kinds of
content the pairing gate already requires to be identical on both sides.
Move generated data into regions written into both pages:
gen-config-catalog renders both languages and the record with per-package
regions labelled by code identifiers; gen-module-graph keeps its table in a
region; gen-doc-graphs splices the event matrix regions into the Chinese
page. Add a generic region splice beside the region grammar.
Publish --dsh-frame-top-clearance on html[data-platform='darwin'] so
body-portalled overlays inherit it; useAnchoredMaxHeight and the portalled
Menu widen their 12px viewport margin to it, Modal centralizes the vertical
padding its cards previously each computed, and the trigger menu keeps an
84px margin on every platform so it stops below the conversation header.
Reduce Transparency replaces the sidebar tint with a 90%-opaque fill.
Opening the trigger menu from the launcher button has to put the keyboard in
the editor for the arrows to drive it, and Lexical's focus restores its
selection a tick later — a track the controller read as an empty draft and
answered by closing the menu the gesture had just opened. A launcher-opened
menu now survives that single follow-up track; typing, picking, Escape or a
pointer press still takes over.
The access-confirmation golden follows the panel's new highlight: it parks on
the row the options mark as the current value rather than the top row.
- The model seat decides before consuming: a focused control that is not a row
(a retry button) keeps the browser's traversal, and Tab enters the list only
from the trigger.
- A launcher-opened menu is a fresh intent, so the dismissal that closed the
same token earlier no longer silences it.
- Comments and JSDoc now match the implementation: the near-end entry resumes
from the walk's last index, Shift+Tab leaves whenever the menu is open, and
the anchor hand-back names the control that opened the menu.
- The e2e focus assertions all go through `expect.poll`.
- Regression coverage for a rowless pane (the trigger keeps the keyboard) and
for a focused retry keeping its Tab, in the seat and the popup shell.
- The PR records the two public-surface changes (`SessionInput.focus()` in,
`bindComposerFocus` out) and the row-fill focus decision.
Review follow-ups on the hand-offs this branch added:
- Menu gives the keyboard back to the control that opened it (an anchor may
wrap several, as a split button does), falls back to a disabled-free anchor
button, leaves a Tab it cannot use to the browser, and no longer moves focus
while the owner keeps the list open.
- The model seat settles only real rows with Tab and keeps the keyboard on its
trigger when a pane's rows are disabled or absent.
- The popup shell leaves Tab to the browser when nothing can be settled (the
error strip's retry button stays reachable), and a dismissal caused by a
stale catalog returns the keyboard like the other paths.
- A frozen-tier track is the submit gesture's own bookkeeping and no longer
re-arms a menu the user dismissed.
- The launcher reuses the shared caret-preserving focus helper; the option
contract records that `active` is the row an accept gesture confirms; the
conversation README documents the composer's key and focus behavior.
The dialog/menu Escape layering this branch also carried is out of scope here
and has been removed again.
Arrow navigation was bound to `autoFocus`, so the dropdowns that open with the
keyboard on their trigger — the composer's permission seat, the settings
pickers, the workspace and preset chips — had no walk at all, and a walk that
read its position from `document.activeElement` alternated between the first
and last row as soon as a row refused focus. The walk now works either way and
resumes from the index it last focused.
The keyboard returns to the anchor on Escape and on a selection — unless the
owner moved it itself, as a presented file card does for its preview button —
and the focused row wears the hover fill instead of the browser's ring. An open
menu takes the first Escape from the dialog it sits in.
The composer's trigger menu keeps its own Tab, which settles or drills the
highlighted candidate; Escape and Shift+Tab leave it without settling, and a
dismissed menu stays dismissed until its query changes. Opening it from the
launcher button returns the keyboard to the editor first, since the menu is a
combobox over that editor.