feat: hold-to-talk shortcut action (push-to-talk dictation) - #433
Open
theoephraim wants to merge 2 commits into
Open
feat: hold-to-talk shortcut action (push-to-talk dictation)#433theoephraim wants to merge 2 commits into
theoephraim wants to merge 2 commits into
Conversation
Greptile SummaryThis PR adds a
Confidence Score: 4/5Safe to merge for non-production testing; a chord held at the moment the app quits normally can stay stuck system-wide until that key is manually pressed and released. The implementation is carefully structured — rebind-safe, auto-repeat-safe, and CaptureInterrupted force-releases every held chord. One gap remains from the previous review: the normal shutdown path (Hook::drop / stop()) does not drain HELD_CHORDS before the callback thread exits, so a chord held at quit time stays physically down on the OS until the user manually clears it. That path is not addressed in this PR. crates/openlogi-agent-core/src/hook_runtime.rs — the shutdown drain is the one path still missing cleanup for held chords. Important Files Changed
Sequence Diagram%%{init: {'theme': 'neutral'}}%%
sequenceDiagram
participant HW as Physical Button
participant Hook as OS Hook Thread
participant HC as HeldChords (thread-local)
participant Inject as openlogi-inject
Note over HW,Inject: Happy path - button held and released normally
HW->>Hook: "Button DOWN (pressed=true)"
Hook->>HC: press(button, combo)
HC-->>Hook: Some(combo) first press only
Hook->>Inject: "press_hold(&combo)"
Inject-->>HW: key-down + modifiers posted to OS
HW->>Hook: "Button UP (pressed=false)"
Hook->>HC: release(button)
HC-->>Hook: Some(combo) stored chord
Hook->>Inject: "release_hold(&combo)"
Inject-->>HW: key-up + modifiers posted to OS
Note over HW,Inject: Auto-repeat guard
HW->>Hook: Button DOWN again (auto-repeat)
Hook->>HC: press(button, combo)
HC-->>Hook: None already held no double press
Note over HW,Inject: CaptureInterrupted (OS kills tap)
HW->>Hook: CaptureInterrupted
Hook->>HC: drain()
HC-->>Hook: all held combos
Hook->>Inject: release_hold x N
Inject-->>HW: force key-up for every stranded chord
%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
sequenceDiagram
participant HW as Physical Button
participant Hook as OS Hook Thread
participant HC as HeldChords (thread-local)
participant Inject as openlogi-inject
Note over HW,Inject: Happy path - button held and released normally
HW->>Hook: "Button DOWN (pressed=true)"
Hook->>HC: press(button, combo)
HC-->>Hook: Some(combo) first press only
Hook->>Inject: "press_hold(&combo)"
Inject-->>HW: key-down + modifiers posted to OS
HW->>Hook: "Button UP (pressed=false)"
Hook->>HC: release(button)
HC-->>Hook: Some(combo) stored chord
Hook->>Inject: "release_hold(&combo)"
Inject-->>HW: key-up + modifiers posted to OS
Note over HW,Inject: Auto-repeat guard
HW->>Hook: Button DOWN again (auto-repeat)
Hook->>HC: press(button, combo)
HC-->>Hook: None already held no double press
Note over HW,Inject: CaptureInterrupted (OS kills tap)
HW->>Hook: CaptureInterrupted
Hook->>HC: drain()
HC-->>Hook: all held combos
Hook->>Inject: release_hold x N
Inject-->>HW: force key-up for every stranded chord
Reviews (2): Last reviewed commit: "feat(agent): hold push-to-talk chords fo..." | Re-trigger Greptile |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
Every action was a one-shot tap fired on button-down, so there was no way to bind a key that stays held — the shape push-to-talk needs. Append Action::HoldShortcut(KeyCombo), carrying the same recorded chord as CustomShortcut but with press/release semantics. Appended at the end: serde encodes the declaration index, so existing variants keep their wire positions and the goldens are unchanged. PROTOCOL_VERSION 8 -> 9 since Action crosses the IPC boundary. openlogi-inject gains phase-aware primitives (press_hold/release_hold) built from the existing per-platform helpers. Down order is modifiers-then-key and up order is the reverse: a chord whose modifiers lifted before its key would surface as a bare keypress. A HoldShortcut reaching execute() came from a dispatch path with no release edge (the HID++ gesture button, which reports an instantaneous click), so it degrades to a tap rather than jamming the chord down. Like CustomShortcut, this is hand-authored TOML for now — the recorder UI that would offer a Hold toggle does not exist yet, so the GUI only labels and renders it.
The hook already delivered both edges on all three platforms; the single-action path dropped the release with an `if pressed` guard. Wire the release edge up to Action::HoldShortcut: press synthesises the key-down, release the matching key-up. Held chords live in a thread-local keyed by button, matching HOLD's thread-per-device reasoning — press and release for a device arrive on the same thread, so the pair always balances. A stuck key-down jams that key for every app on the system, so the unbalanced paths get the care: - The release resolves from stored state *before* the binding maps are read. A rebind or unbind mid-hold would otherwise resolve to a different action (or none, returning PassThrough early) and strand the chord that actually went down. - A repeated button-down does not stack a second key-down the single release could never balance. - CaptureInterrupted force-releases every held chord: the button-ups are never coming. Extract the button arm into on_button. It was already at the pedantic line limit and the hold paths pushed it over; the callback was doing too much to read. Not runtime-tested on hardware — no Logitech device available here.
theoephraim
force-pushed
the
feat/hold-shortcut-action
branch
from
July 20, 2026 19:46
e46a128 to
c8e1cb0
Compare
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
1 participant
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.
feel free to toss this implementation... But was a first go at making it possible to hold a button to do something.
Summary
Adds a
HoldShortcutaction so a mouse button can hold a key combo down foras long as the button is physically held, instead of firing a one-shot tap.
The motivating use case is push-to-talk dictation: bind a thumb button to
the dictation hotkey and hold it to talk, exactly like a walkie-talkie —
press-and-hold to dictate, release to stop. Every existing action fired on
button-down only, so there was no way to bind a key that stays held for the
duration of the press.
Changes
Action::HoldShortcut(KeyCombo), carrying the samerecorded chord as
CustomShortcutbut with press/release semantics.Appended at the end of the enum so serde wire positions and goldens are
unchanged.
PROTOCOL_VERSION10 → 11 (Action crosses the IPC boundary).press_hold/release_holdprimitivesbuilt from the existing per-platform helpers. Down order is
modifiers-then-key, up order is the reverse, so a chord never surfaces as a
bare keypress while its modifiers are mid-transition.
HoldShortcut—press synthesises key-down, release the matching key-up. Held chords live in
a thread-local keyed by button (press/release for a device arrive on the
same thread, so the pair always balances). Unbalanced paths are handled so a
stuck key-down can't jam a key system-wide: the release resolves from stored
state before binding maps are read (survives a mid-hold rebind/unbind), a
repeated button-down won't stack a second key-down, and
CaptureInterruptedforce-releases every held chord. Button arm extractedinto
on_buttonto stay under the pedantic line limit.languages.
Like
CustomShortcut, this is hand-authored TOML for now — the recorder UIthat would offer a Hold toggle doesn't exist yet, so the GUI only labels and
renders it. A
HoldShortcutreaching a dispatch path with no release edge(the HID++ gesture button, which reports an instantaneous click) degrades to
a tap rather than jamming the chord down.
Testing
devenv tasks run openlogi:check(fmt + clippy -D warnings + workspace tests)PROTOCOL_VERSIONbump covered bywire_formattestenvironment. Needs a real press/release verification on macOS/Linux/Windows.