fix: lost control of tab when window inactive

2026-01-20 19:03:15 +08:00
parent 37332cbb9d
commit 1a0c533cb4
2 changed files with 121 additions and 103 deletions
--- a/packages/extension/src/agent/RemotePageController.ts
+++ b/packages/extension/src/agent/RemotePageController.ts
@@ -11,7 +11,7 @@ import type {
 	ScrollHorizontallyOptions,
 	ScrollOptions,
 } from '../messaging/protocol'
-import { rpcClient } from '../messaging/rpc'
+import { type RPCClient, createRPCClient } from '../messaging/rpc'

 /**
 * RemotePageController is a proxy that implements the PageController interface.
@@ -21,27 +21,39 @@ import { rpcClient } from '../messaging/rpc'
 * though events in the remote context are not currently bridged.
 */
 export class RemotePageController extends EventTarget {
+	private rpc: RPCClient
+
+	constructor() {
+		super()
+		// Capture the active tab ID at construction time to avoid issues when tab loses focus
+		const tabIdPromise = chrome.tabs.query({ active: true, currentWindow: true }).then(([tab]) => {
+			if (!tab?.id) throw new Error('No active tab found')
+			return tab.id
+		})
+		this.rpc = createRPCClient(tabIdPromise)
+	}
+
 	// ======= State Queries =======

 	/**
 	 * Get current page URL
 	 */
 	async getCurrentUrl(): Promise<string> {
-		return rpcClient.getCurrentUrl()
+		return this.rpc.getCurrentUrl()
 	}

 	/**
 	 * Get last tree update timestamp
 	 */
 	async getLastUpdateTime(): Promise<number> {
-		return rpcClient.getLastUpdateTime()
+		return this.rpc.getLastUpdateTime()
 	}

 	/**
 	 * Get structured browser state for LLM consumption.
 	 */
 	async getBrowserState(): Promise<BrowserState> {
-		return rpcClient.getBrowserState()
+		return this.rpc.getBrowserState()
 	}

 	// ======= DOM Tree Operations =======
@@ -50,14 +62,14 @@ export class RemotePageController extends EventTarget {
 	 * Update DOM tree, returns simplified HTML for LLM.
 	 */
 	async updateTree(): Promise<string> {
-		return rpcClient.updateTree()
+		return this.rpc.updateTree()
 	}

 	/**
 	 * Clean up all element highlights
 	 */
 	async cleanUpHighlights(): Promise<void> {
-		return rpcClient.cleanUpHighlights()
+		return this.rpc.cleanUpHighlights()
 	}

 	// ======= Element Actions =======
@@ -66,42 +78,42 @@ export class RemotePageController extends EventTarget {
 	 * Click element by index
 	 */
 	async clickElement(index: number): Promise<ActionResult> {
-		return rpcClient.clickElement(index)
+		return this.rpc.clickElement(index)
 	}

 	/**
 	 * Input text into element by index
 	 */
 	async inputText(index: number, text: string): Promise<ActionResult> {
-		return rpcClient.inputText(index, text)
+		return this.rpc.inputText(index, text)
 	}

 	/**
 	 * Select dropdown option by index and option text
 	 */
 	async selectOption(index: number, optionText: string): Promise<ActionResult> {
-		return rpcClient.selectOption(index, optionText)
+		return this.rpc.selectOption(index, optionText)
 	}

 	/**
 	 * Scroll vertically
 	 */
 	async scroll(options: ScrollOptions): Promise<ActionResult> {
-		return rpcClient.scroll(options)
+		return this.rpc.scroll(options)
 	}

 	/**
 	 * Scroll horizontally
 	 */
 	async scrollHorizontally(options: ScrollHorizontallyOptions): Promise<ActionResult> {
-		return rpcClient.scrollHorizontally(options)
+		return this.rpc.scrollHorizontally(options)
 	}

 	/**
 	 * Execute arbitrary JavaScript on the page
 	 */
 	async executeJavascript(script: string): Promise<ActionResult> {
-		return rpcClient.executeJavascript(script)
+		return this.rpc.executeJavascript(script)
 	}

 	// ======= Mask Operations =======
@@ -110,21 +122,21 @@ export class RemotePageController extends EventTarget {
 	 * Show the visual mask overlay.
 	 */
 	async showMask(): Promise<void> {
-		return rpcClient.showMask()
+		return this.rpc.showMask()
 	}

 	/**
 	 * Hide the visual mask overlay.
 	 */
 	async hideMask(): Promise<void> {
-		return rpcClient.hideMask()
+		return this.rpc.hideMask()
 	}

 	/**
 	 * Dispose and clean up resources
 	 */
 	dispose(): void {
-		rpcClient.dispose().catch(() => {
+		this.rpc.dispose().catch(() => {
 			// Ignore errors on dispose
 		})
 	}
--- a/packages/extension/src/messaging/rpc.ts
+++ b/packages/extension/src/messaging/rpc.ts
@@ -13,96 +13,102 @@ import type {
 } from './protocol'

 /**
- * Get the active tab ID for the current sidepanel context.
- * In MV3, we need to explicitly target the tab.
+ * Create an RPC client bound to a specific tab.
+ * The tabId is captured at creation time to ensure messages are sent to the correct tab
+ * even if the user switches tabs or the page loses focus.
 */
-async function getActiveTabId(): Promise<number> {
-	const [tab] = await chrome.tabs.query({ active: true, currentWindow: true })
-	if (!tab?.id) {
-		throw new Error('No active tab found')
+export function createRPCClient(tabIdPromise: Promise<number>): RPCClient {
+	return {
+		// State queries
+		async getCurrentUrl(): Promise<string> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:getCurrentUrl', undefined, tabId)
+		},
+
+		async getLastUpdateTime(): Promise<number> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:getLastUpdateTime', undefined, tabId)
+		},
+
+		async getBrowserState(): Promise<BrowserState> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:getBrowserState', undefined, tabId)
+		},
+
+		// DOM operations
+		async updateTree(): Promise<string> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:updateTree', undefined, tabId)
+		},
+
+		async cleanUpHighlights(): Promise<void> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:cleanUpHighlights', undefined, tabId)
+		},
+
+		// Element actions
+		async clickElement(index: number): Promise<ActionResult> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:clickElement', index, tabId)
+		},
+
+		async inputText(index: number, text: string): Promise<ActionResult> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:inputText', { index, text }, tabId)
+		},
+
+		async selectOption(index: number, optionText: string): Promise<ActionResult> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:selectOption', { index, optionText }, tabId)
+		},
+
+		async scroll(options: ScrollOptions): Promise<ActionResult> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:scroll', options, tabId)
+		},
+
+		async scrollHorizontally(options: ScrollHorizontallyOptions): Promise<ActionResult> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:scrollHorizontally', options, tabId)
+		},
+
+		async executeJavascript(script: string): Promise<ActionResult> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:executeJavascript', script, tabId)
+		},
+
+		// Mask operations
+		async showMask(): Promise<void> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:showMask', undefined, tabId)
+		},
+
+		async hideMask(): Promise<void> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:hideMask', undefined, tabId)
+		},
+
+		// Lifecycle
+		async dispose(): Promise<void> {
+			const tabId = await tabIdPromise
+			return pageControllerRPC.sendMessage('rpc:dispose', undefined, tabId)
+		},
 	}
-	return tab.id
 }

-/**
- * RPC client for calling PageController methods in ContentScript.
- * Each method sends a message and waits for the response.
- */
-export const rpcClient = {
-	// State queries
-	async getCurrentUrl(): Promise<string> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:getCurrentUrl', undefined, tabId)
-	},
-
-	async getLastUpdateTime(): Promise<number> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:getLastUpdateTime', undefined, tabId)
-	},
-
-	async getBrowserState(): Promise<BrowserState> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:getBrowserState', undefined, tabId)
-	},
-
-	// DOM operations
-	async updateTree(): Promise<string> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:updateTree', undefined, tabId)
-	},
-
-	async cleanUpHighlights(): Promise<void> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:cleanUpHighlights', undefined, tabId)
-	},
-
-	// Element actions
-	async clickElement(index: number): Promise<ActionResult> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:clickElement', index, tabId)
-	},
-
-	async inputText(index: number, text: string): Promise<ActionResult> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:inputText', { index, text }, tabId)
-	},
-
-	async selectOption(index: number, optionText: string): Promise<ActionResult> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:selectOption', { index, optionText }, tabId)
-	},
-
-	async scroll(options: ScrollOptions): Promise<ActionResult> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:scroll', options, tabId)
-	},
-
-	async scrollHorizontally(options: ScrollHorizontallyOptions): Promise<ActionResult> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:scrollHorizontally', options, tabId)
-	},
-
-	async executeJavascript(script: string): Promise<ActionResult> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:executeJavascript', script, tabId)
-	},
-
-	// Mask operations
-	async showMask(): Promise<void> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:showMask', undefined, tabId)
-	},
-
-	async hideMask(): Promise<void> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:hideMask', undefined, tabId)
-	},
-
-	// Lifecycle
-	async dispose(): Promise<void> {
-		const tabId = await getActiveTabId()
-		return pageControllerRPC.sendMessage('rpc:dispose', undefined, tabId)
-	},
+export interface RPCClient {
+	getCurrentUrl(): Promise<string>
+	getLastUpdateTime(): Promise<number>
+	getBrowserState(): Promise<BrowserState>
+	updateTree(): Promise<string>
+	cleanUpHighlights(): Promise<void>
+	clickElement(index: number): Promise<ActionResult>
+	inputText(index: number, text: string): Promise<ActionResult>
+	selectOption(index: number, optionText: string): Promise<ActionResult>
+	scroll(options: ScrollOptions): Promise<ActionResult>
+	scrollHorizontally(options: ScrollHorizontallyOptions): Promise<ActionResult>
+	executeJavascript(script: string): Promise<ActionResult>
+	showMask(): Promise<void>
+	hideMask(): Promise<void>
+	dispose(): Promise<void>
 }
-
-export type RPCClient = typeof rpcClient