feat(desktop): extend deeplinks + add Raycast extension scaffold

[sungdark]

Summary

• extend desktop deeplink actions with pause/resume/toggle pause
• add deeplink actions for switching microphone and camera
• add a Raycast extension scaffold in apps/raycast that dispatches Cap deeplinks for recording controls

Deeplink actions added

• pause_recording
• resume_recording
• toggle_pause_recording
• switch_microphone
• switch_camera

Closes #1540

Greptile Summary

Extended desktop deeplink actions to support pause/resume/toggle recording and device switching (microphone/camera), then added a Raycast extension scaffold to control Cap via these deeplinks.

• Added 5 new DeepLinkAction variants in Rust: PauseRecording, ResumeRecording, TogglePauseRecording, SwitchMicrophone, SwitchCamera
• Created new Raycast extension at apps/raycast with 7 commands that dispatch Cap deeplinks
• Critical issue: fireSimpleAction in cap.ts incorrectly wraps unit variants in { type: action } instead of passing the plain string, which will cause deserialization failures for pause/resume/stop/toggle commands

Confidence Score: 2/5

• Critical bug in fireSimpleAction will break 4 of 7 Raycast commands
• The Rust backend implementation is solid, but the TypeScript deeplink dispatch function has a critical serialization bug that breaks unit variant deserialization. Commands using struct variants (start_recording, switch_camera, switch_microphone) will work, but commands using unit variants (pause_recording, resume_recording, stop_recording, toggle_pause_recording) will fail due to incorrect JSON format
• apps/raycast/src/lib/cap.ts requires immediate fix before merge

Important Files Changed

Filename	Overview
apps/desktop/src-tauri/src/deeplink_actions.rs	Added new deeplink action variants and handlers for pause/resume/toggle and device switching
apps/raycast/src/lib/cap.ts	Core deeplink dispatch logic - contains critical bug in fireSimpleAction that will break unit variant actions
apps/raycast/src/pause-recording.ts	Pause recording command - affected by fireSimpleAction bug
apps/raycast/src/resume-recording.ts	Resume recording command - affected by fireSimpleAction bug
apps/raycast/src/stop-recording.ts	Stop recording command - affected by fireSimpleAction bug
apps/raycast/src/toggle-pause-recording.ts	Toggle pause command - affected by fireSimpleAction bug

Sequence Diagram

sequenceDiagram
    participant User
    participant Raycast
    participant cap.ts
    participant Desktop App
    participant Rust Handler
    
    User->>Raycast: Trigger command
    Raycast->>cap.ts: fireSimpleAction("pause_recording")
    cap.ts->>cap.ts: JSON.stringify({ type: "pause_recording" })
    cap.ts->>Desktop App: cap-desktop://action?value=...
    Desktop App->>Rust Handler: Parse deeplink
    Rust Handler->>Rust Handler: serde_json::from_str<DeepLinkAction>
    Note over Rust Handler: FAILS - expects "pause_recording"<br/>but receives {"type":"pause_recording"}
    Rust Handler-->>User: Error: Failed to parse deeplink
    
    Note over cap.ts,Rust Handler: FIX: Pass action string directly,<br/>not wrapped in object

Loading

_{Last reviewed commit: a595e0e}

_{(2/5) Greptile learns from your feedback when you react with thumbs up/down!}

[tembo]

DeepLinkAction uses Serde's default external tagging, so unit variants need to be JSON strings (e.g. "pause_recording"), not { type: "pause_recording" }.

Suggested change

	await dispatchAction({ type: action });
	await dispatchAction(action);

[tembo]

If captureName is empty, this will dispatch an action that the desktop app can't resolve (and the user gets a pretty opaque failure). Consider validating before dispatch.

Suggested change

	async function onSubmit(values: Values) {
	const captureMode =
	values.captureType === "screen"
	? { screen: values.captureName }
	: { window: values.captureName };
	async function onSubmit(values: Values) {
	const captureName = values.captureName.trim();
	if (!captureName) {
	await showHUD("Cap: capture name required");
	return;
	}
	const captureMode =
	values.captureType === "screen" ? { screen: captureName } : { window: captureName };

[greptile-apps]

_{11 files reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}

[greptile-apps]

Serde deserializes unit enum variants from plain JSON strings, not objects. { type: action } should be just action

Suggested change

	export async function fireSimpleAction(action: string) {
	await dispatchAction({ type: action });
	await showHUD(`Cap: ${action}`);
	export async function fireSimpleAction(action: string) {
	await dispatchAction(action);
	await showHUD(`Cap: ${action}`);
	}

Prompt To Fix With AI

This is a comment left during a code review.
Path: apps/raycast/src/lib/cap.ts
Line: 13-15

Comment:
Serde deserializes unit enum variants from plain JSON strings, not objects. `{ type: action }` should be just `action`

```suggestion
export async function fireSimpleAction(action: string) {
  await dispatchAction(action);
  await showHUD(`Cap: ${action}`);
}
```

How can I resolve this? If you propose a fix, please make it concise.

[richiemcilroy]

We are closing the deeplinks and Raycast bounty (#1540) without awarding it, and closing the pull requests opened against it. The full reasoning is on that issue: the remaining actions it asked for are deliberately limited to debug builds because deeplinks are a remote control surface any web page can invoke, and widening that in production is a security decision we do not want to take through a bounty.

This is not a reflection on your work. Leaving it unreviewed for this long is our failure, and we are sorry for the wasted effort. If you want to contribute again, the open bug reports are the most useful place and we will review promptly.