feat: add thinking mode setting for OpenAI Compatible provider

- Add openAiThinkingModeEnabled setting to provider-settings.ts schema
- Add thinking parameter handling in openai.ts handler for both streaming and non-streaming modes
- Create ThinkingModeSetting.tsx UI component
- Integrate ThinkingModeSetting into OpenAICompatible.tsx settings
- Add translations for the new setting to all 18 locale files
- Add comprehensive tests for the new thinking mode feature

Fixes #10603

packages/types/src/provider-settings.ts

@@ -243,6 +243,7 @@ const openAiSchema = baseProviderSettingsSchema.extend({
 	openAiBaseUrl: z.string().optional(),
 	openAiApiKey: z.string().optional(),
 	openAiR1FormatEnabled: z.boolean().optional(),
+	openAiThinkingModeEnabled: z.boolean().optional(),
 	openAiModelId: z.string().optional(),
 	openAiCustomModelInfo: modelInfoSchema.nullish(),
 	openAiUseAzure: z.boolean().optional(),

src/api/providers/__tests__/openai.spec.ts

@@ -497,6 +497,90 @@ describe("OpenAiHandler", () => {
 			const callArgs = mockCreate.mock.calls[0][0]
 			expect(callArgs.max_completion_tokens).toBe(4096)
 		})
+
+		describe("thinking mode", () => {
+			it("should include thinking parameter when openAiThinkingModeEnabled is true (streaming)", async () => {
+				const thinkingOptions: ApiHandlerOptions = {
+					...mockOptions,
+					openAiThinkingModeEnabled: true,
+				}
+				const thinkingHandler = new OpenAiHandler(thinkingOptions)
+				const stream = thinkingHandler.createMessage(systemPrompt, messages)
+				// Consume the stream to trigger the API call
+				for await (const _chunk of stream) {
+				}
+				// Assert the mockCreate was called with thinking parameter
+				expect(mockCreate).toHaveBeenCalled()
+				const callArgs = mockCreate.mock.calls[0][0]
+				expect(callArgs.thinking).toEqual({ type: "enabled" })
+			})
+
+			it("should not include thinking parameter when openAiThinkingModeEnabled is false (streaming)", async () => {
+				const noThinkingOptions: ApiHandlerOptions = {
+					...mockOptions,
+					openAiThinkingModeEnabled: false,
+				}
+				const noThinkingHandler = new OpenAiHandler(noThinkingOptions)
+				const stream = noThinkingHandler.createMessage(systemPrompt, messages)
+				// Consume the stream to trigger the API call
+				for await (const _chunk of stream) {
+				}
+				// Assert the mockCreate was called without thinking parameter
+				expect(mockCreate).toHaveBeenCalled()
+				const callArgs = mockCreate.mock.calls[0][0]
+				expect(callArgs.thinking).toBeUndefined()
+			})
+
+			it("should not include thinking parameter when openAiThinkingModeEnabled is undefined (streaming)", async () => {
+				const defaultOptions: ApiHandlerOptions = {
+					...mockOptions,
+					// openAiThinkingModeEnabled is not set
+				}
+				const defaultHandler = new OpenAiHandler(defaultOptions)
+				const stream = defaultHandler.createMessage(systemPrompt, messages)
+				// Consume the stream to trigger the API call
+				for await (const _chunk of stream) {
+				}
+				// Assert the mockCreate was called without thinking parameter
+				expect(mockCreate).toHaveBeenCalled()
+				const callArgs = mockCreate.mock.calls[0][0]
+				expect(callArgs.thinking).toBeUndefined()
+			})
+
+			it("should include thinking parameter when openAiThinkingModeEnabled is true (non-streaming)", async () => {
+				const thinkingOptions: ApiHandlerOptions = {
+					...mockOptions,
+					openAiThinkingModeEnabled: true,
+					openAiStreamingEnabled: false,
+				}
+				const thinkingHandler = new OpenAiHandler(thinkingOptions)
+				const stream = thinkingHandler.createMessage(systemPrompt, messages)
+				// Consume the stream to trigger the API call
+				for await (const _chunk of stream) {
+				}
+				// Assert the mockCreate was called with thinking parameter
+				expect(mockCreate).toHaveBeenCalled()
+				const callArgs = mockCreate.mock.calls[0][0]
+				expect(callArgs.thinking).toEqual({ type: "enabled" })
+			})
+
+			it("should not include thinking parameter when openAiThinkingModeEnabled is false (non-streaming)", async () => {
+				const noThinkingOptions: ApiHandlerOptions = {
+					...mockOptions,
+					openAiThinkingModeEnabled: false,
+					openAiStreamingEnabled: false,
+				}
+				const noThinkingHandler = new OpenAiHandler(noThinkingOptions)
+				const stream = noThinkingHandler.createMessage(systemPrompt, messages)
+				// Consume the stream to trigger the API call
+				for await (const _chunk of stream) {
+				}
+				// Assert the mockCreate was called without thinking parameter
+				expect(mockCreate).toHaveBeenCalled()
+				const callArgs = mockCreate.mock.calls[0][0]
+				expect(callArgs.thinking).toBeUndefined()
+			})
+		})
 	})
 
 	describe("error handling", () => {

src/api/providers/openai.ts

@@ -89,6 +89,7 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 		const modelUrl = this.options.openAiBaseUrl ?? ""
 		const modelId = this.options.openAiModelId ?? ""
 		const enabledR1Format = this.options.openAiR1FormatEnabled ?? false
+		const enabledThinkingMode = this.options.openAiThinkingModeEnabled ?? false
 		const isAzureAiInference = this._isAzureAiInference(modelUrl)
 		const deepseekReasoner = modelId.includes("deepseek-reasoner") || enabledR1Format
 
@@ -153,13 +154,16 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 
 			const isGrokXAI = this._isGrokXAI(this.options.openAiBaseUrl)
 
-			const requestOptions: OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming = {
+			const requestOptions: OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming & {
+				thinking?: { type: "enabled"; budget_tokens?: number }
+			} = {
 				model: modelId,
 				temperature: this.options.modelTemperature ?? (deepseekReasoner ? DEEP_SEEK_DEFAULT_TEMPERATURE : 0),
 				messages: convertedMessages,
 				stream: true as const,
 				...(isGrokXAI ? {} : { stream_options: { include_usage: true } }),
 				...(reasoning && reasoning),
+				...(enabledThinkingMode && { thinking: { type: "enabled" } }),
 				...(metadata?.tools && { tools: this.convertToolsForOpenAI(metadata.tools) }),
 				...(metadata?.tool_choice && { tool_choice: metadata.tool_choice }),
 				...(metadata?.toolProtocol === "native" && {
@@ -224,11 +228,14 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 				yield this.processUsageMetrics(lastUsage, modelInfo)
 			}
 		} else {
-			const requestOptions: OpenAI.Chat.Completions.ChatCompletionCreateParamsNonStreaming = {
+			const requestOptions: OpenAI.Chat.Completions.ChatCompletionCreateParamsNonStreaming & {
+				thinking?: { type: "enabled"; budget_tokens?: number }
+			} = {
 				model: modelId,
 				messages: deepseekReasoner
 					? convertToR1Format([{ role: "user", content: systemPrompt }, ...messages])
 					: [systemMessage, ...convertToOpenAiMessages(messages)],
+				...(enabledThinkingMode && { thinking: { type: "enabled" } }),
 				...(metadata?.tools && { tools: this.convertToolsForOpenAI(metadata.tools) }),
 				...(metadata?.tool_choice && { tool_choice: metadata.tool_choice }),
 				...(metadata?.toolProtocol === "native" && {

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

@@ -0,0 +1,25 @@
+import { Checkbox } from "vscrui"
+
+import { useAppTranslation } from "@/i18n/TranslationContext"
+
+interface ThinkingModeSettingProps {
+	onChange: (value: boolean) => void
+	openAiThinkingModeEnabled?: boolean
+}
+
+export const ThinkingModeSetting = ({ onChange, openAiThinkingModeEnabled }: ThinkingModeSettingProps) => {
+	const { t } = useAppTranslation()
+
+	return (
+		<div>
+			<div className="flex items-center gap-2">
+				<Checkbox checked={openAiThinkingModeEnabled} onChange={onChange}>
+					<span className="font-medium">{t("settings:modelInfo.enableThinkingMode")}</span>
+				</Checkbox>
+			</div>
+			<p className="text-vscode-descriptionForeground text-sm mt-0">
+				{t("settings:modelInfo.enableThinkingModeTips")}
+			</p>
+		</div>
+	)
+}

webview-ui/src/components/settings/providers/OpenAICompatible.tsx

@@ -20,6 +20,7 @@ import { convertHeadersToObject } from "../utils/headers"
 import { inputEventTransform, noTransform } from "../transforms"
 import { ModelPicker } from "../ModelPicker"
 import { R1FormatSetting } from "../R1FormatSetting"
+import { ThinkingModeSetting } from "../ThinkingModeSetting"
 import { ThinkingBudget } from "../ThinkingBudget"
 
 type OpenAICompatibleProps = {
@@ -153,6 +154,10 @@ export const OpenAICompatible = ({
 				onChange={handleInputChange("openAiR1FormatEnabled", noTransform)}
 				openAiR1FormatEnabled={apiConfiguration?.openAiR1FormatEnabled ?? false}
 			/>
+			<ThinkingModeSetting
+				onChange={handleInputChange("openAiThinkingModeEnabled", noTransform)}
+				openAiThinkingModeEnabled={apiConfiguration?.openAiThinkingModeEnabled ?? false}
+			/>
 			<Checkbox
 				checked={apiConfiguration?.openAiStreamingEnabled ?? true}
 				onChange={handleInputChange("openAiStreamingEnabled", noTransform)}>

webview-ui/src/i18n/locales/ca/settings.json

@@ -863,6 +863,8 @@
 		"enableStreaming": "Habilitar streaming",
 		"enableR1Format": "Activar els paràmetres del model R1",
 		"enableR1FormatTips": "S'ha d'activat quan s'utilitzen models R1 com el QWQ per evitar errors 400",
+		"enableThinkingMode": "Activar mode de pensament",
+		"enableThinkingModeTips": "Activar per a models de pensament com Kimi K2, DeepSeek Reasoner, GLM-4 per retornar contingut de raonament",
 		"useAzure": "Utilitzar Azure",
 		"azureApiVersion": "Establir versió de l'API d'Azure",
 		"gemini": {

webview-ui/src/i18n/locales/de/settings.json

@@ -863,6 +863,8 @@
 		"enableStreaming": "Streaming aktivieren",
 		"enableR1Format": "R1-Modellparameter aktivieren",
 		"enableR1FormatTips": "Muss bei Verwendung von R1-Modellen wie QWQ aktiviert werden, um 400er-Fehler zu vermeiden",
+		"enableThinkingMode": "Denkmodus aktivieren",
+		"enableThinkingModeTips": "Für Denkmodelle wie Kimi K2, DeepSeek Reasoner, GLM-4 aktivieren, um Schlussfolgerungsinhalte zurückzugeben",
 		"useAzure": "Azure verwenden",
 		"azureApiVersion": "Azure API-Version festlegen",
 		"gemini": {

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

@@ -872,6 +872,8 @@
 		"enableStreaming": "Enable streaming",
 		"enableR1Format": "Enable R1 model parameters",
 		"enableR1FormatTips": "Must be enabled when using R1 models such as QWQ to prevent 400 errors",
+		"enableThinkingMode": "Enable thinking mode",
+		"enableThinkingModeTips": "Enable for thinking models like Kimi K2, DeepSeek Reasoner, GLM-4 to return reasoning content",
 		"useAzure": "Use Azure",
 		"azureApiVersion": "Set Azure API version",
 		"gemini": {

webview-ui/src/i18n/locales/es/settings.json

@@ -863,6 +863,8 @@
 		"enableStreaming": "Habilitar streaming",
 		"enableR1Format": "Habilitar parámetros del modelo R1",
 		"enableR1FormatTips": "Debe habilitarse al utilizar modelos R1 como QWQ, para evitar el error 400",
+		"enableThinkingMode": "Habilitar modo de pensamiento",
+		"enableThinkingModeTips": "Habilitar para modelos de pensamiento como Kimi K2, DeepSeek Reasoner, GLM-4 para devolver contenido de razonamiento",
 		"useAzure": "Usar Azure",
 		"azureApiVersion": "Establecer versión de API de Azure",
 		"gemini": {

webview-ui/src/i18n/locales/fr/settings.json

@@ -863,6 +863,8 @@
 		"enableStreaming": "Activer le streaming",
 		"enableR1Format": "Activer les paramètres du modèle R1",
 		"enableR1FormatTips": "Doit être activé lors de l'utilisation de modèles R1 tels que QWQ, pour éviter l'erreur 400",
+		"enableThinkingMode": "Activer le mode réflexion",
+		"enableThinkingModeTips": "Activer pour les modèles de réflexion comme Kimi K2, DeepSeek Reasoner, GLM-4 pour retourner le contenu de raisonnement",
 		"useAzure": "Utiliser Azure",
 		"azureApiVersion": "Définir la version de l'API Azure",
 		"gemini": {

webview-ui/src/i18n/locales/hi/settings.json

@@ -864,6 +864,8 @@
 		"enableStreaming": "स्ट्रीमिंग सक्षम करें",
 		"enableR1Format": "R1 मॉडल पैरामीटर सक्षम करें",
 		"enableR1FormatTips": "QWQ जैसी R1 मॉडलों का उपयोग करते समय इसे सक्षम करना आवश्यक है, ताकि 400 त्रुटि से बचा जा सके",
+		"enableThinkingMode": "थिंकिंग मोड सक्षम करें",
+		"enableThinkingModeTips": "Kimi K2, DeepSeek Reasoner, GLM-4 जैसे थिंकिंग मॉडल के लिए सक्षम करें ताकि तर्क सामग्री वापस आ सके",
 		"useAzure": "Azure का उपयोग करें",
 		"azureApiVersion": "Azure API संस्करण सेट करें",
 		"gemini": {

webview-ui/src/i18n/locales/id/settings.json

@@ -893,6 +893,8 @@
 		"enableStreaming": "Aktifkan streaming",
 		"enableR1Format": "Aktifkan parameter model R1",
 		"enableR1FormatTips": "Harus diaktifkan saat menggunakan model R1 seperti QWQ untuk mencegah error 400",
+		"enableThinkingMode": "Aktifkan mode berpikir",
+		"enableThinkingModeTips": "Aktifkan untuk model berpikir seperti Kimi K2, DeepSeek Reasoner, GLM-4 untuk mengembalikan konten penalaran",
 		"useAzure": "Gunakan Azure",
 		"azureApiVersion": "Atur versi API Azure",
 		"gemini": {

webview-ui/src/i18n/locales/it/settings.json

@@ -864,6 +864,8 @@
 		"enableStreaming": "Abilita streaming",
 		"enableR1Format": "Abilita i parametri del modello R1",
 		"enableR1FormatTips": "Deve essere abilitato quando si utilizzano modelli R1 come QWQ, per evitare l'errore 400",
+		"enableThinkingMode": "Abilita modalità di pensiero",
+		"enableThinkingModeTips": "Abilita per modelli di pensiero come Kimi K2, DeepSeek Reasoner, GLM-4 per restituire contenuti di ragionamento",
 		"useAzure": "Usa Azure",
 		"azureApiVersion": "Imposta versione API Azure",
 		"gemini": {

webview-ui/src/i18n/locales/ja/settings.json

@@ -864,6 +864,8 @@
 		"enableStreaming": "ストリーミングを有効化",
 		"enableR1Format": "R1モデルパラメータを有効にする",
 		"enableR1FormatTips": "QWQなどのR1モデルを使用する際には、有効にする必要があります。400エラーを防ぐために",
+		"enableThinkingMode": "思考モードを有効にする",
+		"enableThinkingModeTips": "Kimi K2、DeepSeek Reasoner、GLM-4などの思考モデルで推論内容を返すために有効にします",
 		"useAzure": "Azureを使用",
 		"azureApiVersion": "Azure APIバージョンを設定",
 		"gemini": {

webview-ui/src/i18n/locales/ko/settings.json

@@ -864,6 +864,8 @@
 		"enableStreaming": "스트리밍 활성화",
 		"enableR1Format": "R1 모델 매개변수 활성화",
 		"enableR1FormatTips": "QWQ와 같은 R1 모델을 사용할 때 활성화해야 하며, 400 오류를 방지합니다",
+		"enableThinkingMode": "사고 모드 활성화",
+		"enableThinkingModeTips": "Kimi K2, DeepSeek Reasoner, GLM-4와 같은 사고 모델에서 추론 내용을 반환하려면 활성화하세요",
 		"useAzure": "Azure 사용",
 		"azureApiVersion": "Azure API 버전 설정",
 		"gemini": {

webview-ui/src/i18n/locales/nl/settings.json

@@ -864,6 +864,8 @@
 		"enableStreaming": "Streaming inschakelen",
 		"enableR1Format": "R1-modelparameters inschakelen",
 		"enableR1FormatTips": "Moet ingeschakeld zijn bij gebruik van R1-modellen zoals QWQ om 400-fouten te voorkomen",
+		"enableThinkingMode": "Denkmodus inschakelen",
+		"enableThinkingModeTips": "Inschakelen voor denkmodellen zoals Kimi K2, DeepSeek Reasoner, GLM-4 om redeneerinhoud te retourneren",
 		"useAzure": "Azure gebruiken",
 		"azureApiVersion": "Azure API-versie instellen",
 		"gemini": {

webview-ui/src/i18n/locales/pl/settings.json

@@ -864,6 +864,8 @@
 		"enableStreaming": "Włącz strumieniowanie",
 		"enableR1Format": "Włącz parametry modelu R1",
 		"enableR1FormatTips": "Należy włączyć podczas korzystania z modeli R1, takich jak QWQ, aby uniknąć błędu 400",
+		"enableThinkingMode": "Włącz tryb myślenia",
+		"enableThinkingModeTips": "Włącz dla modeli myślących, takich jak Kimi K2, DeepSeek Reasoner, GLM-4, aby zwrócić treść rozumowania",
 		"useAzure": "Użyj Azure",
 		"azureApiVersion": "Ustaw wersję API Azure",
 		"gemini": {

webview-ui/src/i18n/locales/pt-BR/settings.json

@@ -864,6 +864,8 @@
 		"enableStreaming": "Ativar streaming",
 		"enableR1Format": "Ativar parâmetros do modelo R1",
 		"enableR1FormatTips": "Deve ser ativado ao usar modelos R1 como QWQ, para evitar erro 400",
+		"enableThinkingMode": "Ativar modo de pensamento",
+		"enableThinkingModeTips": "Ativar para modelos de pensamento como Kimi K2, DeepSeek Reasoner, GLM-4 para retornar conteúdo de raciocínio",
 		"useAzure": "Usar Azure",
 		"azureApiVersion": "Definir versão da API Azure",
 		"gemini": {

webview-ui/src/i18n/locales/ru/settings.json

@@ -864,6 +864,8 @@
 		"enableStreaming": "Включить потоковую передачу",
 		"enableR1Format": "Включить параметры модели R1",
 		"enableR1FormatTips": "Необходимо включить при использовании моделей R1 (например, QWQ), чтобы избежать ошибок 400",
+		"enableThinkingMode": "Включить режим мышления",
+		"enableThinkingModeTips": "Включите для моделей мышления, таких как Kimi K2, DeepSeek Reasoner, GLM-4, чтобы возвращать содержание рассуждений",
 		"useAzure": "Использовать Azure",
 		"azureApiVersion": "Установить версию API Azure",
 		"gemini": {

webview-ui/src/i18n/locales/tr/settings.json

@@ -864,6 +864,8 @@
 		"enableStreaming": "Akışı etkinleştir",
 		"enableR1Format": "R1 model parametrelerini etkinleştir",
 		"enableR1FormatTips": "QWQ gibi R1 modelleri kullanıldığında etkinleştirilmelidir, 400 hatası alınmaması için",
+		"enableThinkingMode": "Düşünme modunu etkinleştir",
+		"enableThinkingModeTips": "Kimi K2, DeepSeek Reasoner, GLM-4 gibi düşünme modelleri için akıl yürütme içeriği döndürmek üzere etkinleştirin",
 		"useAzure": "Azure kullan",
 		"azureApiVersion": "Azure API sürümünü ayarla",
 		"gemini": {

webview-ui/src/i18n/locales/vi/settings.json

@@ -864,6 +864,8 @@
 		"enableStreaming": "Bật streaming",
 		"enableR1Format": "Kích hoạt tham số mô hình R1",
 		"enableR1FormatTips": "Cần kích hoạt khi sử dụng các mô hình R1 như QWQ, để tránh lỗi 400",
+		"enableThinkingMode": "Bật chế độ suy nghĩ",
+		"enableThinkingModeTips": "Bật cho các mô hình suy nghĩ như Kimi K2, DeepSeek Reasoner, GLM-4 để trả về nội dung suy luận",
 		"useAzure": "Sử dụng Azure",
 		"azureApiVersion": "Đặt phiên bản API Azure",
 		"gemini": {

webview-ui/src/i18n/locales/zh-CN/settings.json

@@ -864,6 +864,8 @@
 		"enableStreaming": "启用流式传输",
 		"enableR1Format": "启用 R1 模型参数",
 		"enableR1FormatTips": "使用 QWQ 等 R1 系列模型时必须启用，避免出现 400 错误",
+		"enableThinkingMode": "启用思考模式",
+		"enableThinkingModeTips": "为 Kimi K2、DeepSeek Reasoner、GLM-4 等思考模型启用，以返回推理内容",
 		"useAzure": "使用 Azure 服务",
 		"azureApiVersion": "设置 Azure API 版本",
 		"gemini": {

webview-ui/src/i18n/locales/zh-TW/settings.json

@@ -864,6 +864,8 @@
 		"enableStreaming": "啟用串流輸出",
 		"enableR1Format": "啟用 R1 模型參數",
 		"enableR1FormatTips": "使用 QWQ 等 R1 模型時必須啟用，以避免發生 400 錯誤",
+		"enableThinkingMode": "啟用思考模式",
+		"enableThinkingModeTips": "為 Kimi K2、DeepSeek Reasoner、GLM-4 等思考模型啟用，以返回推理內容",
 		"useAzure": "使用 Azure",
 		"azureApiVersion": "設定 Azure API 版本",
 		"gemini": {