[pull] main from Kilo-Org:main

.changeset/clean-dogs-teach.md

@@ -0,0 +1,5 @@
+---
+"kilo-code": patch
+---
+
+Add a tooltip explaining why speech-to-text may be unavailable

src/core/webview/ClineProvider.ts

@@ -2213,8 +2213,11 @@ ${prompt}
 		// kilocode_change end
 
 		// kilocode_change start - checkSpeechToTextAvailable (only when experiment enabled)
-		let speechToTextAvailable =
-			experiments?.speechToText && (await checkSpeechToTextAvailable(this.providerSettingsManager))
+		let speechToTextStatus: { available: boolean; reason?: "openaiKeyMissing" | "ffmpegNotInstalled" } | undefined =
+			undefined
+		if (experiments?.speechToText) {
+			speechToTextStatus = await checkSpeechToTextAvailable(this.providerSettingsManager)
+		}
 		// kilocode_change end - checkSpeechToTextAvailable
 
 		let cloudOrganizations: CloudOrganizationMembership[] = []
@@ -2440,7 +2443,7 @@ ${prompt}
 			featureRoomoteControlEnabled,
 			virtualQuotaActiveModel, // kilocode_change: Include virtual quota active model in state
 			debug: vscode.workspace.getConfiguration(Package.name).get<boolean>("debug", false),
-			speechToTextAvailable, // kilocode_change: Whether speech-to-text is fully configured
+			speechToTextStatus, // kilocode_change: Speech-to-text availability status with failure reason
 		}
 	}
 

src/core/webview/speechToTextCheck.ts

@@ -3,10 +3,19 @@ import type { ProviderSettingsManager } from "../config/ProviderSettingsManager"
 import { getOpenAiApiKey } from "../../services/stt/utils/getOpenAiCredentials"
 import { FFmpegCaptureService } from "../../services/stt/FFmpegCaptureService"
 
+/**
+ * Result type for speech-to-text availability check
+ */
+export type SpeechToTextAvailabilityResult = {
+	available: boolean
+	reason?: "openaiKeyMissing" | "ffmpegNotInstalled"
+}
+
 /**
  * Cached availability result with timestamp
  */
-let cachedResult: { available: boolean; timestamp: number } | null = null
+let cachedResult: { available: boolean; reason?: "openaiKeyMissing" | "ffmpegNotInstalled"; timestamp: number } | null =
+	null
 const CACHE_DURATION_MS = 30000 // 30 seconds
 
 /**
@@ -21,39 +30,39 @@ const CACHE_DURATION_MS = 30000 // 30 seconds
  *
  * @param providerSettingsManager - Provider settings manager for API configuration
  * @param forceRecheck - Force a fresh check, ignoring cache (default: false)
- * @returns Promise<boolean> - true if prerequisites are met
+ * @returns Promise<SpeechToTextAvailabilityResult> - Result with availability status and failure reason if unavailable
  */
 export async function checkSpeechToTextAvailable(
 	providerSettingsManager: ProviderSettingsManager,
 	forceRecheck = false,
-): Promise<boolean> {
+): Promise<SpeechToTextAvailabilityResult> {
 	// Return cached result if valid and not forcing recheck
 	if (cachedResult !== null && !forceRecheck) {
 		const age = Date.now() - cachedResult.timestamp
 		if (age < CACHE_DURATION_MS) {
-			return cachedResult.available
+			return { available: cachedResult.available, reason: cachedResult.reason }
 		}
 	}
 
 	try {
 		// Check 1: OpenAI API key
 		const apiKey = await getOpenAiApiKey(providerSettingsManager)
 		if (!apiKey) {
-			cachedResult = { available: false, timestamp: Date.now() }
-			return false
+			cachedResult = { available: false, reason: "openaiKeyMissing", timestamp: Date.now() }
+			return { available: false, reason: "openaiKeyMissing" }
 		}
 
 		// Check 2: FFmpeg installed
 		const ffmpegResult = FFmpegCaptureService.findFFmpeg()
 		if (!ffmpegResult.available) {
-			cachedResult = { available: false, timestamp: Date.now() }
-			return false
+			cachedResult = { available: false, reason: "ffmpegNotInstalled", timestamp: Date.now() }
+			return { available: false, reason: "ffmpegNotInstalled" }
 		}
 
 		cachedResult = { available: true, timestamp: Date.now() }
-		return true
+		return { available: true }
 	} catch (error) {
 		cachedResult = { available: false, timestamp: Date.now() }
-		return false
+		return { available: false }
 	}
 }

src/shared/ExtensionMessage.ts

@@ -538,7 +538,7 @@ export type ExtensionState = Pick<
 	virtualQuotaActiveModel?: { id: string; info: ModelInfo } // kilocode_change: Add virtual quota active model for UI display
 	showTimestamps?: boolean // kilocode_change: Show timestamps in chat messages
 	debug?: boolean
-	speechToTextAvailable?: boolean // kilocode_change: Whether speech-to-text is fully configured (FFmpeg + OpenAI key)
+	speechToTextStatus?: { available: boolean; reason?: "openaiKeyMissing" | "ffmpegNotInstalled" } // kilocode_change: Speech-to-text availability status with failure reason
 }
 
 export interface ClineSayTool {

webview-ui/src/components/chat/ChatTextArea.tsx

@@ -162,7 +162,7 @@ export const ChatTextArea = forwardRef<HTMLTextAreaElement, ChatTextAreaProps>(
 			ghostServiceSettings, // kilocode_change
 			language, // User's VSCode display language
 			experiments, // kilocode_change: For speechToText experiment flag
-			speechToTextAvailable, // kilocode_change: Whether voice transcription is configured
+			speechToTextStatus, // kilocode_change: Speech-to-text availability status with failure reason
 		} = useExtensionState()
 
 		// kilocode_change start - autocomplete profile type system
@@ -1712,10 +1712,14 @@ export const ChatTextArea = forwardRef<HTMLTextAreaElement, ChatTextAreaProps>(
 						<MicrophoneButton
 							isRecording={isRecording}
 							onClick={handleMicrophoneClick}
-							disabled={!speechToTextAvailable}
+							disabled={!speechToTextStatus?.available}
 							tooltipContent={
-								!speechToTextAvailable
-									? "Configure an OpenAI API key and install FFmpeg to use voice transcription"
+								!speechToTextStatus?.available && speechToTextStatus
+									? speechToTextStatus.reason === "openaiKeyMissing"
+										? t("kilocode:speechToText.unavailableOpenAiKeyMissing")
+										: speechToTextStatus.reason === "ffmpegNotInstalled"
+											? t("kilocode:speechToText.unavailableFfmpegNotInstalled")
+											: t("kilocode:speechToText.unavailableBoth")
 									: undefined
 							}
 						/>

webview-ui/src/components/chat/MicrophoneButton.tsx

@@ -40,10 +40,11 @@ export const MicrophoneButton: React.FC<MicrophoneButtonProps> = ({
 					"rounded-md min-w-[28px] min-h-[28px]",
 					"transition-all duration-150",
 					"focus:outline-none focus-visible:ring-1 focus-visible:ring-vscode-focusBorder",
-					"cursor-pointer",
 					isRecording
-						? "opacity-100 text-red-500 animate-pulse hover:text-red-600"
-						: "opacity-60 hover:opacity-100 text-vscode-descriptionForeground hover:text-vscode-foreground hover:bg-[rgba(255,255,255,0.03)] hover:border-[rgba(255,255,255,0.15)] active:bg-[rgba(255,255,255,0.1)]",
+						? "opacity-100 text-red-500 animate-pulse hover:text-red-600 cursor-pointer"
+						: disabled
+							? "opacity-40 cursor-not-allowed grayscale-[30%] hover:bg-transparent hover:border-[rgba(255,255,255,0.08)] active:bg-transparent text-vscode-descriptionForeground"
+							: "opacity-60 hover:opacity-100 text-vscode-descriptionForeground hover:text-vscode-foreground hover:bg-[rgba(255,255,255,0.03)] hover:border-[rgba(255,255,255,0.15)] active:bg-[rgba(255,255,255,0.1)] cursor-pointer",
 					containerWidth !== undefined && { hidden: containerWidth < 235 },
 				)}>
 				{isRecording ? <Square className="w-4 h-4 fill-current" /> : <Mic className="w-4 h-4" />}

webview-ui/src/context/ExtensionStateContext.tsx

@@ -98,7 +98,7 @@ export interface ExtensionStateContextType extends ExtensionState {
 	setCustomCondensingPrompt: (value: string) => void
 	yoloGatekeeperApiConfigId?: string // kilocode_change: AI gatekeeper for YOLO mode
 	setYoloGatekeeperApiConfigId: (value: string) => void // kilocode_change: AI gatekeeper for YOLO mode
-	speechToTextAvailable?: boolean // kilocode_change: Whether voice transcription is fully configured
+	speechToTextStatus?: { available: boolean; reason?: "openaiKeyMissing" | "ffmpegNotInstalled" } // kilocode_change: Speech-to-text availability status with failure reason
 	marketplaceItems?: any[]
 	marketplaceInstalledMetadata?: MarketplaceInstalledMetadata
 	profileThresholds: Record<string, number>

webview-ui/src/i18n/locales/en/kilocode.json

@@ -326,6 +326,9 @@
 	},
 	"speechToText": {
 		"startRecording": "Start voice input",
-		"stopRecording": "Stop voice input"
+		"stopRecording": "Stop voice input",
+		"unavailableOpenAiKeyMissing": "Speech-to-text is unavailable. It requires a valid OpenAI provider with an API key to use voice transcription.",
+		"unavailableFfmpegNotInstalled": "Speech-to-text is unavailable. Install FFmpeg to use voice transcription.",
+		"unavailableBoth": "Speech-to-text is unavailable. It requires a valid OpenAI provider and FFmpeg must be installed."
 	}
 }