A new prompt_cursor_style setting (follow, block, bar, underline) shapes
the caret while the shell waits at a prompt, whether tty7's inline editor
or the shell's own line editor draws it. `follow`, the default, keeps
cursor_style everywhere, so nothing changes until it is set. Any other
value leaves cursor_style to the programs the shell runs: bar here with
cursor_style block gives kitty and ghostty's bar at the prompt and a block
inside a TUI that never sets a shape, such as Claude Code. A vi-mode
prompt keeps the shell's own insert/normal shapes.
Settings shows both cursor shapes as dropdowns, kept in step with resets,
language switches and config edits made outside the window.
Closes#958
The pin icon read as a smudge at header size. Kept groups now carry a small
◆ beside their name — every kept group, label groups too, since with the
divider gone it is the one thing that tells kept from derived. On a folder
group the mark is still the click that unpins it, and the hover button that
pins an auto group shows the same character.
The divider is no longer drawn at rest. It keeps its (smaller) place in the
layout and its recorded bounds, so pinning a header by dragging it up and
handing a tab back by dropping it below behave exactly as before, and it
still shows as a line while it is the drop target.
* feat(agents): recognise Empryo and Prime Agent
Prime Agent (PrimeIntellect-ai/prime-agent) is a Pi fork: detected as
`prime-agent`, resumes with `--resume <id>`, forks with `--fork <id>`,
`--no-session` opts out, and reports status through the shared Pi
extension bridge at ~/.prime/agent/extensions/tty7/index.ts.
Empryo is detected as `empryo` and resumes with `empryo --session <id>`.
No hook installer yet: Empryo filters TTY7* variables out of hook
processes, so `tty7-app agent-hook` would stop at the missing marker.
* feat(agents): Antigravity status hooks
Antigravity CLI (`agy`) now installs a `tty7` hook set in
~/.gemini/config/hooks.json, next to the user's own sets:
PreInvocation -> prompt-submit, PostToolUse -> tool-complete,
Stop -> stop. Only the first PreInvocation of a turn (invocationNum 0)
counts as a new prompt, later ones keep the turn working; a Stop with
fullyIdle false is ignored. conversationId and workspacePaths feed the
session id and cwd. PreToolUse is left alone because it must answer
with a decision.
* fix(agents): index Prime Agent and Antigravity in settings, scope conversationId alias
- Add settings titles, search keywords (en/ja/zh) and Agents index entries
for the two new HookAgents; agent_rows_are_in_the_search_index requires
one per HookAgent::ALL.
- Read Antigravity's conversationId as the session id only for antigravity:
the alias table is last-write-wins, so a global alias could replace
another agent's session id.
- Drop the stray .empryo/ job records.
---------
Co-authored-by: kalpak <you@example.com>
Co-authored-by: l0ng-ai <24760907+l0ng-ai@users.noreply.github.com>
* feat(tabs): MoveTabLeft / MoveTabRight — keyboard tab reordering
The drag that reorders a tab now has a keyboard form: the active tab trades
places with its neighbour in visual order, wrapping past either end. Ships
unbound like the pane-swap pair; bindable from config.json and the Keybindings
page, and listed in the palette and the docs.
* fix(tabs): keep a keyboard tab move inside its sidebar group
On a left tab bar the move stepped along the flat visual order, so a
step out of a group reordered self.tabs without moving anything on
screen (still saved, still synced as TabMove, and visible later on a
top bar), and a wrap through another group landed the tab at its own
group's far end by accident. The move now reassigns only the slots of
the tab's own sidebar section, and wraps at that section's ends.
---------
Co-authored-by: l0ng-ai <24760907+l0ng-ai@users.noreply.github.com>
* feat(sidebar): keep pinned groups on the workspace and derive the rest
Replace the sidebar's hand-made groups with the model from #955: the sidebar
groups tabs by repo automatically, and you pin what you want to keep.
- Pinned groups (`PinnedGroup`: id, optional name, optional folder, fold) are
stored on the workspace in the machine tree, in display order, and a tab
points at one by `GroupId`. Everything else is an auto group worked out
every frame and never stored: by repo home, or by `user@host` for an SSH
pane — native or a shell that ssh'd onward — so `/home/ubuntu` on two
machines no longer lands under one header.
- A tab whose cwd *enters* a pinned folder joins it (deepest folder wins; a
repo home equal to the folder counts, which keeps worktrees with their
repo). It is edge-triggered through `EntryWatch`, so a tab dragged out
while still inside the folder stays out until it leaves and comes back, and
a tab restored at launch is not pulled in by where it already sits.
- Groups sync as one `WorkspaceSetGroups` / `GroupsChanged`, pushed only from
an edit and adopted from every pull, so a fresh window can never push an
empty set over the workspace's. The machine hands tabs naming a dropped
group back to auto grouping in the same mutation. Control dialect → v11.
- Config: `sidebar_grouping` (three modes) and `sidebar_collapsed_groups` give
way to one `sidebar_auto_grouping` toggle; folds live with the workspace.
- `tty7 tab ls` reports the pinned group a tab is in (name or folder leaf;
JSON carries id, name and folder).
* feat(sidebar): draw pinned groups above a divider, with their own gestures
The sidebar now reads as two halves: the groups you keep, in the order you
put them, then a divider, then the groups it works out (Arc-style).
- Pinned headers drag-reorder among themselves (their own reorder surface, so
a pinned header cannot be dropped among the derived ones); the order lands
on the workspace's group list, not on the tabs.
- An auto header carried above the divider is pinned when let go. With
nothing pinned yet the divider appears during that drag as a "Drop here to
pin" zone, since a hairline at the top of the list is nothing to aim at.
- A tab kept in a pinned group and dropped anywhere below the divider goes
back to auto grouping; the divider lights to say so.
- An empty pinned group stays, with a "+ New Tab" row that opens a tab in its
folder (or where ⌘T would, for a label group) and files it there.
- Folder groups carry a pin mark that unpins on click and a tooltip with the
folder; auto headers show pin and "+" on hover.
- Header menus: pinned — Rename, Set Folder… (local workspaces), Use Current
Tab's Folder, Clear Folder, New Tab, Unpin (folder groups), Delete. Auto —
Pin Group, New Tab. Nothing renames an auto group; nothing pins implicitly.
* feat(sidebar): open folders as pinned groups from Finder, the file tree and the palette
Every way into a pinned group the design calls for:
- Drop a folder from Finder or Explorer onto the sidebar to pin it (a local
workspace only — a dropped path is this machine's, and a folder group keeps
a directory on the workspace's host). Files are let fall.
- "Pin as Group" on a folder in the file tree, on local and remote workspaces
alike, since the tree and the group are both on the workspace's host.
- Palette "New Group" makes an empty label group and opens its name for
typing; "Open Folder as Group…" picks a folder with the system picker, pins
it and opens a tab in it. The picker browses this computer, so that one is
not offered on a remote workspace.
- Tab right-click "Move to Group" lists the pinned groups plus "New Group…",
which files the tab in a fresh label group with its name open for typing.
Pinning a folder already pinned hands back the group that keeps it rather
than making a second one to split its tabs with.
* fix(sidebar): let groups that arrive from elsewhere pull no tab into a folder
A window draws its first frames before its copy of the workspace's groups
lands, so every tab's entry watch recorded "in no folder" — and the groups
landing then read as each tab walking into its folder. A restored tab, or one
dragged out of its folder group, was pulled back in on every launch.
Groups adopted from a pull or from another window's `GroupsChanged` now start
every tab's watch over from where it is; only this window's own pin gathers
the tabs inside the folder, and says so tab by tab. A tab also goes up with
the group it names even when the window does not know that group yet, so a
sync in that same gap cannot send every kept tab back to auto grouping.
* docs(sidebar): describe pinned and auto groups, and log the change
Rewrite the sidebar page's grouping section around "grouped by repo
automatically; pin what you want to keep": the divider, folder and label
groups, the edge-triggered join, every way to pin, and the header menus. The
configuration reference swaps `sidebar_grouping` for `sidebar_auto_grouping`,
the CLI reference describes the GROUP column as the pinned group, and the
changelog gains an Unreleased entry (#955).
* fix(sidebar): file a tab opened by the CLI in a pinned folder into it
A tab that reaches a window as TabCreated — from `tty7 tab new` or another
window — started its entry watch as a restored tab, so opening one inside a
pinned folder left it in the auto group below. It is as new as a tab opened
here, and now joins the folder like one; every window that hears of it
reaches the same answer.
* test(machine): build the group sets in their initializers
Clippy's field_reassign_with_default on the two WorkspaceGroups the
set-groups test assembles.
* fix(sidebar): draw restored tabs in their auto group, and title by repo again
Auto groups are not stored, so after a restart every tab sat in Ungrouped
until its own repo probe came back, then jumped; before pinned groups the
stored repo key put it in place on the first frame. Each tab now carries
`last_auto`, the auto group it last resolved to, as a hint: stored with the
tab, sent up alongside its group in `TabSetGroup` whenever the live answer
moves, and used to draw the tab until the probe answers. The probe always wins
and rewrites the hint, and the hint never outranks a pinned group or the
folder-entry rule. Another window's hint only fills a gap, so two windows can
never bounce a disagreement between them.
The workspace's fallback title regained the repo majority it lost: the most
common pinned folder first, then the repo most unpinned tabs were last filed
under (a worktree counting toward its repo home), then a pane's cwd.
* refactor: drop what the new sidebar left unused, and two clippy findings
- `TerminalView::native_ssh_cwd` and its helper existed for the sidebar's old
folder grouping of native SSH panes; an SSH tab now groups by host, and
nothing else read it.
- The file tree's context menu takes `cx` instead of `danger` and the new
groups flag, back to the argument count it had on main.
- A title test builds its workspace in the initializer.
* feat(agents): quick launch for detected CLI agents (#955)
Every agent whose launch program is on PATH becomes a palette command,
"Agent: <name>", ordered by frecency and bindable as LaunchAgent:<slug>.
"New Agent Tab" (Cmd+Shift+A on macOS; unbound elsewhere, where
Ctrl+Shift+A is select-all) launches the most recently used one, and the
New Tab menu gains a single "Launch Agent..." row that opens the palette
pre-filtered to them.
A launch always opens a new pane (a tab in the active tab's cwd, or a
split when picked from the palette with Alt held) and types the command
into that pane's shell once it exists - immediately for a local pane, on
landing for a remote one via PendingSpawn::run_on_land - never into a
pane that was already there. Detection, status and resume then work as
for a hand-typed agent.
The command is the agent's bare binary unless the new `agent_launch`
config map overrides it. A wrapper named there is detected as its agent
without an `agent_commands` entry: the daemon folds the programs
`agent_launch` runs into its alias map (interpreters, shells and real
agent names excepted), reloaded when config.json changes instead of
once per process, and a mapped script run under its interpreter
(`bash ~/bin/cc`, `node cc.js`) is now recognised too.
A running agent's pane menu gets "Set Current Launch Args as Default",
which writes its launch argv - minus session flags and positional
prompts, joined with the settings' quoting - into `agent_launch`.
Remote workspaces cannot be asked for their PATH through the Host
trait, so they offer the agents previously seen running in that
workspace (WindowView::seen_agents).
* fix(agents): count only agents that start under a view, not ones reattached to
A view rebuilt over a running pane - every agent tab after an app
restart, or a workspace switched back to - saw its agent appear from
nothing and reported it as detected, bumping that agent's frecency once
per launch of the app. The terminal now remembers whether its link was
an attach, and such a view only reports agents once its shell has been
seen back at the prompt with no agent in front.
Also pins down that the agents seen in a remote workspace, its quick
launch list, persist in views.json with the rest of the workspace.
* fix(daemon): probe the foreground as soon as a new program takes it
The foreground probes ran on output only, at most once per 500ms
interval. A quick launch types the agent's command the moment the shell
is up, so the agent drew its whole first screen inside the interval of
the prompt it was typed at and then waited for a key: the pane never
learned it was running an agent until something else printed. Seen in
a dev instance, where a launched Claude Code stayed undetected at its
trust prompt.
The reader now asks the pty for its foreground process group on every
read (one ioctl) and probes immediately when it changes.
The + menu listed every shell the machine reports (nine on a stock macOS box)
above the SSH hosts. The Local section now names the default shell, always
first, then only shells that have actually been opened, by frecency, three rows
at most, the same way the SSH section caps its hosts.
Shell usage is recorded in a new `shell_frecency` config map, keyed by the
inventory label, bumped from both the menu row and the palette. Every shell is
now a palette command titled "Shell: {label}" (same word in every locale), and
an "Other Shells…" row opens the palette pre-filtered to them. That row is
hidden when the menu already names the whole inventory. Running a shell command
from the palette respects the ⌥/Alt split modifier like the menu row.
A tab can be put to sleep from its context menu or the command palette:
its panes are stopped (screens kept on disk), the tab keeps its place in
the sidebar, and selecting it wakes it through the same restore a reboot
runs, resuming a supported agent's session. The sleep mark lives on the
machine tree, so it survives app and daemon restarts.
Closes#762
A forward could only be removed, so pointing one local port at another
remote target meant deleting the rule and retyping the other one (#439).
Forward rules gain an `enabled` flag (serde default on, so saved profiles
and older peers keep every rule live) and the daemon a SetForwardEnabled
op for panes and workspaces. Off releases the listener and keeps the
entry; on rebinds it from the rule it was made from, and is refused by
name when another switched-on forward of the same owner holds the port.
The Ports panel and the Settings rules editor each get a switch; a
switch flipped in the panel on a rule from a saved host is written back
to that host.
Closes#439
Saved SSH hosts and their usage counts move out of config.json into
servers.json beside it, so config.json can be synced between machines
without carrying a server list (#911). An older config.json is split on
first load: servers.json is written first (0600), then only the two keys
are removed from config.json, leaving every other key as it was. When
both files hold hosts, servers.json wins and the stale copy falls out of
config.json at its next save. A servers.json that does not parse is kept
aside and blocks saves, as config.json does; hand edits hot-reload.
Settings -> Window & Tabs -> SSH tab title (`ssh_tab_title`) pins an SSH
tab to the profile name (saved host name, ~/.ssh/config alias, or the
address typed for a quick connect) or the hostname, on the OSC title's
rung of the existing label ladder: a renamed tab still wins, OSC titles
are still tracked, local panes are untouched (#726).
The README now opens on an animated WebP of the one-minute tour (GitHub
won't play an mp4 inline), linking the full-quality mp4. The docs home
page autoplays the mp4 in the hero slot. The old hero screenshot and the
separate tour poster block are gone.
Every placeholder frame in docs/ now shows a real capture of the current
build: 20 screenshots plus two short looping clips (prompt editor, pane
drag). The README and the docs home page link a one-minute tour covering
agent status across repos, one agent driving another through the CLI,
the prompt editor, diffs, pane dragging, and sessions surviving a quit.
The v5 restyle turned the tab rows, the search field and the workspace
switcher into full pills, which read too round at this density. Put
back CARD_RADIUS on the rows, rounded_lg on the search field and
rounded_md on the switcher (and its rename field), keeping the 28px
row height. PILL_RADIUS has no users left and goes with them.
- Rails share the window fill and are separated by a hairline only;
the current tab is a neutral selected step plus a semibold title
instead of the blue navigation wash (the unused role is removed).
- Default Light theme takes ink #0F1419 and accent #1F6BF0; the caret
follows the accent.
- Sidebar: 28px pill rows with 18px avatars (two-line rows keep card
corners), pill search field and workspace switcher, group headings
in body ink with the chevron shown only when folded.
- Right panel: word tabs (Info / Changes / Files) with an underline
bar on a closing hairline. The Changes count is dropped before any
label is cut, and the panel floor grows with the measured labels so
the chrome tiles are never pushed off the edge.
- Headings are no longer uppercased; popovers use a 10px radius.
- docs/design-system.md updated to match.
The interactive cursor-agent TUI fires beforeSubmitPrompt and stop; only
print mode (-p) lacks them. Mapping postToolUse to prompt-submit froze the
activity counter (opportunistic git refresh, agents --changed), shortened
turn timing and hid tool-less turns. Map it to tool-complete instead.
A portable ZIP install whose folder has both the .tty7-portable marker
and a data\ folder next to tty7-app.exe now uses <exe dir>\data as its
config directory, so settings, sessions, scrollback and the daemon's
socket travel with the program. --config-dir and TTY7_CONFIG_DIR still
win. The folder is an explicit opt-in because the marker has shipped in
every portable ZIP since the in-app updater arrived: keying on it alone
would move existing users onto an empty directory at their next update.
The portable directory is not the machine's config dir, so a portable
copy never adopts the installed tty7's legacy tree. An updater test pins
that data\ is not a managed root and survives an in-place update.
SelectWorkspace1-9 and the Window menu numbered workspaces by most
recent use, so switching to one moved it to slot 1 and reshuffled the
rest. Number them by the order this client first had them instead
(the append-only views list), skipping synced remote references until
they are opened; opening one moves it to the end so it takes the next
free number. The switcher keeps its MRU list and shows each row's
number.
Add a macOS-only `font_thicken` key (default true) and a Settings row
under Appearance > Terminal text. When off, AppleFontSmoothing is pinned
to 0 in this process's NSArgumentDomain before gpui's text system first
reads it, so glyphs render at the face's own weight. The volatile domain
is in-memory only: nothing is persisted and no other app is affected.
gpui caches the preference in a OnceLock, so a change applies after a
restart; no gpui fork change is needed.
A remote workspace's Files tree took uploads by drag-and-drop but had no
way to bring a file back. Right-click a file -> Download now reads it
through Host::read_file and saves it into this machine's Downloads
folder, named and numbered the way the SFTP panel's Download does.
Capped at the same ~63 MB as uploads (one control frame); an oversized
file is refused up front with the limit and nothing is transferred.
Local trees keep Reveal in that slot; folders are not offered.
Closes#723
Cursor's ~/.cursor/hooks.json is a flat hook map — `command` directly on
each entry under `hooks.<event>` — so it reuses the flat writer Crush
introduced. Two things are Cursor-specific:
- The file needs `"version": 1` at its root or Cursor ignores it. tty7
now writes it when it creates the file or finds it missing, never
overwrites a version the user set, and reports a versionless file that
holds our hooks as Outdated so refresh repairs it.
- Payloads name the session `conversation_id` (only sessionStart repeats
it as `session_id`), and most events carry `workspace_roots` instead of
`cwd`. Both are now read as aliases, which is what Copy Session ID and
resume were missing.
Events: sessionStart, beforeSubmitPrompt, postToolUse, stop, sessionEnd.
cursor-agent does not fire beforeSubmitPrompt today (a known gap on
Cursor's side), so postToolUse also reports prompt-submit: the first tool
call opens the turn and Cursor's stop, which the CLI does fire, closes it.
A turn with no tool calls never shows as working. None of Cursor's
permission hooks are installed, since those would block on the empty
stdout tty7's hook prints.
Add SelectNextTab / SelectPrevTab, which move to the neighbouring tab in
the order the strip or sidebar shows them, wrapping, with no popup.
Defaults: Cmd+Shift+] / Cmd+Shift+[ on macOS, Ctrl+PgDn / Ctrl+PgUp
elsewhere. The tmux preset's prefix n / p now bind these (tmux
next-window semantics) instead of the MRU switcher, which never
auto-committed there.
NextTab / PrevTab keep their config names and Ctrl+Tab, but are labelled
Recent Tab Switcher on the Keybindings page; the palette's Next/Previous
Tab entries now show the SelectNextTab chord they actually run.
The code panel's Preview/Edit and Wrap buttons only toggled state in their
click handlers, so there was no way to reach them from the keyboard. Register
both as actions, unbound by default like ToggleDocumentFill and the
DocumentWidth* actions, list them in the command palette and the Keybindings
page, and route the buttons through the same methods.
Instead of config keys for the initial state, the last-used state is
remembered (editor_soft_wrap / editor_markdown_preview), the way diff_view
and scm_graph_expanded already are. A file opened at a line target always
opens as source so the cursor is visible.
CodeBuddy Code takes Claude Code's hooks as-is: the same nested
`hooks.<Event>[].hooks[{type, command}]` shape in ~/.codebuddy/settings.json
(or $CODEBUDDY_CONFIG_DIR/settings.json), the same event names, and the same
session_id/cwd payload fields. So it rides the shared hook-map installer and
needs no payload aliases.
The event table follows Qoder rather than Claude: CodeBuddy has a
first-class PermissionRequest and Elicitation, so it gets no Notification
hook, and StopFailure ends a turn like Stop. CodeBuddy also emits
SessionStart with source "compact" mid-turn, which is filtered out the same
way Qoder's is so the pane does not drop back to idle.
Detection covers all three npm bins (codebuddy, codebuddy-code, cbc).
Resume is `codebuddy --resume <id>` and fork appends --fork-session; stale
session and worktree flags are dropped from the replayed launch argv, and
--no-session-persistence disables both commands.
Icon, en/ja/zh names and search keywords, and docs are updated.
Reopening the SFTP panel went back to the last browsed folder; a fresh
open now prefers the pane shell's cwd, falling back to the last browsed
folder, then the login directory. Following a pane switch with the panel
still open keeps each pane's last browsed folder.
Also pin the test config dir in the sftp gpui harness so filtered runs
do not trip the real-config guard, and fix docs/remote/sftp.mdx, which
said the panel opens on the remote home directory.
Closes#826
The install guide said only releases up to 26.8.2 needed it, but 26.8.3
through 26.9.2 were built the same way and import VCRUNTIME140.dll too.
Also correct the CI comment: cc-built code follows +crt-static by itself;
what can bring the import back is a prebuilt native library or RUSTFLAGS.
Every Windows binary tty7 has ever shipped imports `VCRUNTIME140.dll`. That
file is not part of Windows; it arrives with the "Visual C++ 2015-2022
Redistributable", which Visual Studio, the GitHub runners and most developer
machines install as a side effect of something else. On a machine that has
never installed it the loader fails before `main` — no window, no log, no
crash report, just
The code execution cannot proceed because VCRUNTIME140.dll was not found.
which is how 26.8.2 failed winget's install validation
(microsoft/winget-pkgs#415841).
Confirmed against the installed 26.x `tty7-app.exe` with
`dumpbin /dependents`: `VCRUNTIME140.dll` plus nine `api-ms-win-crt-*`
entries, and the same in `tty7.exe` and `tty7-updater.exe`. The UCRT half is
in-box from Windows 10 on and never was the problem; VCRUNTIME140 is the one
piece that has to come from the redistributable. The bundled ConPTY pair is
already CRT-static, so it was never implicated — only our own three binaries
are.
`-C target-feature=+crt-static` for the MSVC targets, in `.cargo/config.toml`
rather than in the release workflow, so CI's Windows `build & test` job
compiles under the same flag a release does and a dependency that cannot link
statically fails a pull request instead of a tag. The alternative — declaring
`Microsoft.VCRedist.2015+.x64` in the winget manifest, or shipping the DLLs
beside the exe — was rejected: it leaves the portable zip, the GitHub release
and every non-winget install path broken, this repository publishes no winget
manifest to carry the declaration, and the winget PR shows the declared
dependency did not actually resolve the failure. Static linking removes the
requirement instead of documenting it.
The regression is invisible to everyone who could catch it, because every
machine that builds tty7 has the redistributable, so
`assert-no-vcruntime.ps1` reads the PE import and delay-load tables directly
(no `dumpbin`, which would re-introduce the same "my machine has Visual
Studio" assumption) and fails on any VC++ redistributable import. It runs in
CI on the Windows debug build and, via `verify-windows-package.ps1`, over
both shipped payloads in release and nightly.
Verified on Windows 11 x86_64 with MSVC 14.44: a full
`cargo build --release --locked --target x86_64-pc-windows-msvc` links
cleanly, and `dumpbin /dependents` on the resulting `tty7-app.exe` shows 29
imports, all in-box — no `VCRUNTIME140.dll` and no `api-ms-win-crt-*` at all.
Same for `tty7.exe` and the `--features updater` `tty7-updater.exe`.
`cargo test --release -p tty7-core` under the flag: 1223 passed, 4 failed,
those four being the `remote_link`/`router` tests that fail on a clean tree
on this machine too.
Fixes#902
Claude-Session: https://claude.ai/code/session_01JRqYZ9E153WpSHGS2AW3BM
Alt+1..9 are vim's tab keys, and tty7 takes all nine for Go to Tab.
Two things stood between the reporter and getting them back.
**Nothing in the app could leave an action unbound.** Backspace on a
Keybindings row that has recorded nothing *reset* the row — dropped the
override so the action gets its shipped chord back. On a row nobody had
overridden, which is every row the first time it is looked at, that is a
no-op: pressing it over Alt+1 left Alt+1 sitting exactly where it was,
which reads as the default restoring itself. `config.json` has spelled
"no chord" as `[]` since #868, but no gesture wrote it.
Backspace now writes that empty list. The row falls to `—` and grows the
**Reset** button every overridden row has, which is the way back to the
default. The capture hint names the key, and the docs say what it is for.
**A keybinding line serde could not read failed the whole `Config`.**
`keybindings` is a hand-edited map and was strict, so `"ActivateTab1":
null` — or a number, or an object — quarantined `config.json` and started
the app on built-in defaults. Every rebinding in the file then read as
its shipped default, and the next settings write persisted those
defaults over what the user had written. It now reads one entry at a
time, like every other hand-edited nested key here: the lines that name
a shortcut bind, a line that does not is logged and skipped.
Tests, each failing on the unfixed code:
- `ui::app::keybinding_gpui_tests::backspace_on_a_row_unbinds_the_action_rather_than_restoring_its_default`
- `core::config::tests::a_keybinding_line_that_cannot_be_read_does_not_take_the_config_with_it`
and `ui::keymap::gpui_tests::alt_digits_can_be_moved_off_the_tab_actions_for_good`
pins the merge and a save/reload round trip: a list replaces the shipped
Alt+1, `[]` leaves nothing, and neither comes back after a restart.
Fixes#901
Claude-Session: https://claude.ai/code/session_01JRqYZ9E153WpSHGS2AW3BM
A chord added in config wins a tie only against another app-wide action.
Copy, paste, find, clear scrollback and insert newline are bound in the
Terminal context, which is deeper, so they keep their chord while a
terminal is focused. Also tells people upgrading that a shortcut recorded
in Settings up to 26.9.2 was saved as a string. It now adds instead of
replacing, and a list brings back the old meaning.
Claude-Session: https://claude.ai/code/session_01JRqYZ9E153WpSHGS2AW3BM
effective_bindings kept one chord per action and set_binding overwrote that
slot, so "NextTab": "cmd-shift-]" silently took Ctrl+Tab away.
A string in keybindings now adds a chord beside the action's default (or
preset) chord; "" still unbinds, as configs and the docs already rely on; a
list is the exact chord set, [] unbinds. Configured chords are installed after
every shipped one, so a chord the user names wins a tie with another action's
default. The Settings page lists every chord of an action, and recording a
shortcut writes the list shape (it sets the binding) and takes only the stolen
chord from the action that had it.
Claude-Session: https://claude.ai/code/session_01JRqYZ9E153WpSHGS2AW3BM
Crush fires only PreToolUse, and it mapped to prompt-submit. With no Stop
behind it the pane stayed on Working until Crush exited: every close asked
whether to cut Crush's work short, the tray and dot stayed on working, and
fork warned mid-turn. Map it to tool-complete, which still records
session_id/cwd for resume and bumps activity without moving the status.
l0ng-ai
migegeandl0ng-ai
l0ng-ai
Fabrice Aneche