feat(human-relay): add clipboard monitoring for AI responses

src/api/providers/human-relay.ts

@@ -1,4 +1,3 @@
-// filepath: e:\Project\Roo-Code\src\api\providers\human-relay.ts
 import { Anthropic } from "@anthropic-ai/sdk"
 import { ApiHandlerOptions, ModelInfo } from "../../shared/api"
 import { ApiHandler, SingleCompletionHandler } from "../index"
@@ -46,7 +45,7 @@ export class HumanRelayHandler implements ApiHandler, SingleCompletionHandler {
 		await vscode.env.clipboard.writeText(promptText)
 
 		// A dialog box pops up to request user action
-		const response = await showHumanRelayDialog(promptText)
+		const response = await showHumanRelayDialog(promptText, this.options)
 
 		if (!response) {
 			// The user canceled the operation
@@ -86,7 +85,7 @@ export class HumanRelayHandler implements ApiHandler, SingleCompletionHandler {
 		await vscode.env.clipboard.writeText(prompt)
 
 		// A dialog box pops up to request user action
-		const response = await showHumanRelayDialog(prompt)
+		const response = await showHumanRelayDialog(prompt, this.options)
 
 		if (!response) {
 			throw new Error("Human relay operation cancelled")
@@ -111,12 +110,48 @@ function getMessageContent(message: Anthropic.Messages.MessageParam): string {
 	}
 	return ""
 }
+
+// Elevate lastAIResponse variable to module level to maintain state between multiple calls
+let lastAIResponse: string | null = null
+// Add normalized cache to avoid repeatedly processing the same content
+let normalizedPrompt: string | null = null
+let normalizedLastResponse: string | null = null
+
+/**
+ * Normalize string by removing excess whitespace
+ * @param text Input string
+ * @returns Normalized string
+ */
+function normalizeText(text: string | null): string {
+	if (!text) return ""
+	// Remove all whitespace and convert to lowercase for case-insensitive comparison
+	return text.replace(/\s+/g, " ").trim()
+}
+
+/**
+ * Compare two strings, ignoring whitespace
+ * @param str1 First string
+ * @param str2 Second string
+ * @returns Whether equal
+ */
+function isTextEqual(str1: string | null, str2: string | null): boolean {
+	if (str1 === str2) return true // Fast path: same reference
+	if (!str1 || !str2) return false // One is empty
+
+	return normalizeText(str1) === normalizeText(str2)
+}
+
 /**
  * Displays the human relay dialog and waits for user response.
  * @param promptText The prompt text that needs to be copied.
  * @returns The user's input response or undefined (if canceled).
  */
-async function showHumanRelayDialog(promptText: string): Promise<string | undefined> {
+async function showHumanRelayDialog(promptText: string, options?: ApiHandlerOptions): Promise<string | undefined> {
+	// Save initial clipboard content for comparison
+	const initialClipboardContent = await vscode.env.clipboard.readText()
+	// Pre-normalize prompt text to avoid repeated processing during polling
+	normalizedPrompt = normalizeText(promptText)
+
 	return new Promise<string | undefined>((resolve) => {
 		// Create a unique request ID
 		const requestId = Date.now().toString()
@@ -126,6 +161,11 @@ async function showHumanRelayDialog(promptText: string): Promise<string | undefi
 			"roo-cline.registerHumanRelayCallback",
 			requestId,
 			(response: string | undefined) => {
+				// Clear clipboard monitoring timer
+				if (clipboardInterval) {
+					clearInterval(clipboardInterval)
+					clipboardInterval = null
+				}
 				resolve(response)
 			},
 		)
@@ -135,5 +175,69 @@ async function showHumanRelayDialog(promptText: string): Promise<string | undefi
 			requestId,
 			promptText,
 		})
+
+		// If clipboard monitoring is enabled, start polling for clipboard changes
+		let clipboardInterval: NodeJS.Timeout | null = null
+
+		if (options?.humanRelayMonitorClipboard) {
+			const monitorInterval = Math.min(Math.max(100, options?.humanRelayMonitorInterval ?? 500), 2000)
+
+			clipboardInterval = setInterval(async () => {
+				try {
+					// Check if clipboard has changed
+					const currentClipboardContent = await vscode.env.clipboard.readText()
+
+					if (!currentClipboardContent || !currentClipboardContent.trim()) {
+						return // Skip empty content
+					}
+
+					// Normalize current clipboard content to avoid repeated processing
+					const normalizedClipboard = normalizeText(currentClipboardContent)
+
+					// Validate clipboard content and check for duplicate response
+					if (
+						normalizedClipboard !== normalizeText(initialClipboardContent) &&
+						normalizedClipboard !== normalizedPrompt &&
+						normalizedClipboard !== normalizedLastResponse
+					) {
+						// Update last AI response
+						lastAIResponse = currentClipboardContent
+						normalizedLastResponse = normalizedClipboard
+
+						// Clear timer
+						if (clipboardInterval) {
+							clearInterval(clipboardInterval)
+							clipboardInterval = null
+						}
+
+						// Get current panel
+						const panel = getPanel()
+						if (panel) {
+							// Send close dialog message
+							panel.webview.postMessage({ type: "closeHumanRelayDialog" })
+						}
+
+						// Send response automatically
+						vscode.commands.executeCommand("roo-cline.handleHumanRelayResponse", {
+							requestId,
+							text: currentClipboardContent,
+						})
+					}
+
+					// New: Check if the last AI response content was copied
+					// Use improved comparison method
+					else if (
+						normalizedClipboard === normalizedLastResponse &&
+						normalizedClipboard !== normalizedPrompt
+					) {
+						// Get current panel and send warning message
+						const panel = getPanel()
+						panel?.webview.postMessage({ type: "showDuplicateResponseAlert" })
+					}
+				} catch (error) {
+					console.error("Error monitoring clipboard:", error)
+				}
+			}, monitorInterval)
+		}
 	})
 }

src/shared/WebviewMessage.ts

@@ -97,6 +97,8 @@ export interface WebviewMessage {
 		| "maxOpenTabsContext"
 		| "humanRelayResponse"
 		| "humanRelayCancel"
+		| "closeHumanRelayDialog"
+		| "showDuplicateResponseAlert"
 		| "browserToolEnabled"
 		| "telemetrySetting"
 		| "showRooIgnoredFiles"

src/shared/api.ts

@@ -73,6 +73,9 @@ export interface ApiHandlerOptions {
 	modelTemperature?: number | null
 	modelMaxTokens?: number
 	modelMaxThinkingTokens?: number
+	// Human relay specific options
+	humanRelayMonitorClipboard?: boolean // Whether to monitor clipboard for automatic content sending
+	humanRelayMonitorInterval?: number // Monitoring interval time (milliseconds)
 }
 
 export type ApiConfiguration = ApiHandlerOptions & {
@@ -126,6 +129,8 @@ export const API_CONFIG_KEYS: GlobalStateKey[] = [
 	"modelTemperature",
 	"modelMaxTokens",
 	"modelMaxThinkingTokens",
+	"humanRelayMonitorClipboard",
+	"humanRelayMonitorInterval",
 ]
 
 // Models

src/shared/globalState.ts

@@ -98,6 +98,8 @@ export const GLOBAL_STATE_KEYS = [
 	"lmStudioDraftModelId",
 	"telemetrySetting",
 	"showRooIgnoredFiles",
+	"humanRelayMonitorClipboard",
+	"humanRelayMonitorInterval",
 ] as const
 
 // Derive the type from the array - creates a union of string literals

webview-ui/src/App.tsx

@@ -25,8 +25,15 @@ const tabsByMessageAction: Partial<Record<NonNullable<ExtensionMessage["action"]
 	historyButtonClicked: "history",
 }
 const App = () => {
-	const { didHydrateState, showWelcome, shouldShowAnnouncement, telemetrySetting, telemetryKey, machineId } =
-		useExtensionState()
+	const {
+		didHydrateState,
+		showWelcome,
+		shouldShowAnnouncement,
+		telemetrySetting,
+		telemetryKey,
+		machineId,
+		apiConfiguration,
+	} = useExtensionState()
 	const [showAnnouncement, setShowAnnouncement] = useState(false)
 	const [tab, setTab] = useState<Tab>("chat")
 	const settingsRef = useRef<SettingsViewRef>(null)
@@ -139,6 +146,8 @@ const App = () => {
 				onClose={() => setHumanRelayDialogState((prev) => ({ ...prev, isOpen: false }))}
 				onSubmit={handleHumanRelaySubmit}
 				onCancel={handleHumanRelayCancel}
+				monitorClipboard={apiConfiguration?.humanRelayMonitorClipboard}
+				monitorInterval={apiConfiguration?.humanRelayMonitorInterval}
 			/>
 		</>
 	)

webview-ui/src/components/human-relay/HumanRelayDialog.tsx

@@ -3,7 +3,8 @@ import { Button } from "../ui/button"
 import { Dialog, DialogContent, DialogDescription, DialogFooter, DialogHeader, DialogTitle } from "../ui/dialog"
 import { Textarea } from "../ui/textarea"
 import { useClipboard } from "../ui/hooks"
-import { Check, Copy, X } from "lucide-react"
+import { AlertTriangle, Check, Copy, X } from "lucide-react"
+import { ProgressIndicator } from "../chat/ChatRow"
 
 interface HumanRelayDialogProps {
 	isOpen: boolean
@@ -12,6 +13,8 @@ interface HumanRelayDialogProps {
 	promptText: string
 	onSubmit: (requestId: string, text: string) => void
 	onCancel: (requestId: string) => void
+	monitorClipboard?: boolean
+	monitorInterval?: number
 }
 
 /**
@@ -25,19 +28,44 @@ export const HumanRelayDialog: React.FC<HumanRelayDialogProps> = ({
 	promptText,
 	onSubmit,
 	onCancel,
+	monitorClipboard = false,
+	monitorInterval = 500,
 }) => {
 	const [response, setResponse] = React.useState("")
 	const { copy } = useClipboard()
 	const [isCopyClicked, setIsCopyClicked] = React.useState(false)
+	const [showDuplicateWarning, setShowDuplicateWarning] = React.useState(false)
 
 	// Listen to isOpen changes, clear the input box when the dialog box is opened
 	React.useEffect(() => {
 		if (isOpen) {
 			setResponse("")
 			setIsCopyClicked(false)
 		}
+		setShowDuplicateWarning(false)
 	}, [isOpen])
 
+	React.useEffect(() => {
+		// Handle messages from extension
+		const messageHandler = (event: MessageEvent) => {
+			const message = event.data
+			if (message.type === "closeHumanRelayDialog") {
+				onClose()
+			}
+			// Handle duplicate response warning
+			else if (message.type === "showDuplicateResponseAlert") {
+				// Show warning
+				setShowDuplicateWarning(true)
+			}
+		}
+
+		window.addEventListener("message", messageHandler)
+
+		return () => {
+			window.removeEventListener("message", messageHandler)
+		}
+	}, [onClose])
+
 	// Copy to clipboard and show a success message
 	const handleCopy = () => {
 		copy(promptText)
@@ -85,6 +113,24 @@ export const HumanRelayDialog: React.FC<HumanRelayDialogProps> = ({
 					</div>
 
 					{isCopyClicked && <div className="text-sm text-emerald-500 font-medium">Copied to clipboard</div>}
+					{monitorClipboard && (
+						<>
+							{showDuplicateWarning && (
+								<div className="flex items-center gap-2 text-sm p-2 rounded-md bg-amber-100 dark:bg-amber-900 border border-amber-300 dark:border-amber-700 text-amber-800 dark:text-amber-300">
+									<AlertTriangle className="h-4 w-4 text-amber-500" />
+									<span className="font-medium">
+										It seems you copied the AI's response from the last interaction instead of the
+										current task. Please check your interaction with the web AI.
+									</span>
+								</div>
+							)}
+
+							<div className="flex items-center gap-2 text-sm text-vscode-descriptionForeground">
+								<ProgressIndicator />
+								<span>Monitoring clipboard for changes, interval: {monitorInterval}ms</span>
+							</div>
+						</>
+					)}
 
 					<div>
 						<div className="mb-2 font-medium">Please enter the AI's response:</div>

webview-ui/src/components/settings/ApiOptions.tsx

@@ -1194,6 +1194,40 @@ const ApiOptions = ({
 						automatically. You need to paste these to web versions of AI (such as ChatGPT or Claude), then
 						copy the AI's reply back to the dialog box and click the confirm button.
 					</div>
+					{/* Auto clipboard monitoring */}
+					<div className="mt-4">
+						<Checkbox
+							checked={apiConfiguration?.humanRelayMonitorClipboard ?? false}
+							onChange={handleInputChange("humanRelayMonitorClipboard", noTransform)}>
+							<span className="font-medium">Enable clipboard monitoring</span>
+						</Checkbox>
+						<div className="text-sm text-vscode-descriptionForeground ml-6">
+							Automatically detect when you copy the AI's response from the browser
+						</div>
+					</div>
+					{apiConfiguration?.humanRelayMonitorClipboard && (
+						<div className="mt-2">
+							<label className="font-medium">Monitor interval (ms)</label>
+							<input
+								type="range"
+								min="100"
+								max="2000"
+								step="100"
+								value={apiConfiguration?.humanRelayMonitorInterval || 500}
+								onChange={handleInputChange("humanRelayMonitorInterval", (e) => {
+									const target = e.target as HTMLInputElement
+									return parseInt(target.value)
+								})}
+								className="h-2 focus:outline-0 w-4/5 accent-vscode-button-background"
+							/>
+							<span style={{ minWidth: "45px", textAlign: "left" }}>
+								{apiConfiguration?.humanRelayMonitorInterval || 500} ms
+							</span>
+							<div className="text-sm text-vscode-descriptionForeground">
+								How frequently to check for clipboard changes (100-2000ms)
+							</div>
+						</div>
+					)}
 				</>
 			)}