Add the option to use a custom Host header for openai-compatible (#2399)

.vscodeignore

@@ -26,9 +26,11 @@ demo.gif
 .prettierignore
 .clinerules*
 .roomodes
+.roo/**
 cline_docs/**
 coverage/**
 locales/**
+benchmark/**
 
 # Ignore all webview-ui files except the build directory (https://github.com/microsoft/vscode-webview-ui-toolkit-samples/blob/main/frameworks/hello-world-react-cra/.vscodeignore)
 webview-ui/src/**

src/api/providers/openai.ts

@@ -55,10 +55,20 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 				baseURL,
 				apiKey,
 				apiVersion: this.options.azureApiVersion || azureOpenAiDefaultApiVersion,
-				defaultHeaders,
+				defaultHeaders: {
+					...defaultHeaders,
+					...(this.options.openAiHostHeader ? { Host: this.options.openAiHostHeader } : {}),
+				},
 			})
 		} else {
-			this.client = new OpenAI({ baseURL, apiKey, defaultHeaders })
+			this.client = new OpenAI({
+				baseURL,
+				apiKey,
+				defaultHeaders: {
+					...defaultHeaders,
+					...(this.options.openAiHostHeader ? { Host: this.options.openAiHostHeader } : {}),
+				},
+			})
 		}
 	}
 
@@ -67,6 +77,7 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 		const modelUrl = this.options.openAiBaseUrl ?? ""
 		const modelId = this.options.openAiModelId ?? ""
 		const enabledR1Format = this.options.openAiR1FormatEnabled ?? false
+		const enabledLegacyFormat = this.options.openAiLegacyFormat ?? false
 		const isAzureAiInference = this._isAzureAiInference(modelUrl)
 		const urlHost = this._getUrlHost(modelUrl)
 		const deepseekReasoner = modelId.includes("deepseek-reasoner") || enabledR1Format
@@ -85,7 +96,7 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 			let convertedMessages
 			if (deepseekReasoner) {
 				convertedMessages = convertToR1Format([{ role: "user", content: systemPrompt }, ...messages])
-			} else if (ark) {
+			} else if (ark || enabledLegacyFormat) {
 				convertedMessages = [systemMessage, ...convertToSimpleMessages(messages)]
 			} else {
 				if (modelInfo.supportsPromptCache) {
@@ -190,7 +201,9 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 				model: modelId,
 				messages: deepseekReasoner
 					? convertToR1Format([{ role: "user", content: systemPrompt }, ...messages])
-					: [systemMessage, ...convertToOpenAiMessages(messages)],
+					: enabledLegacyFormat
+						? [systemMessage, ...convertToSimpleMessages(messages)]
+						: [systemMessage, ...convertToOpenAiMessages(messages)],
 			}
 
 			const response = await this.client.chat.completions.create(
@@ -330,7 +343,7 @@ export class OpenAiHandler extends BaseProvider implements SingleCompletionHandl
 	}
 }
 
-export async function getOpenAiModels(baseUrl?: string, apiKey?: string) {
+export async function getOpenAiModels(baseUrl?: string, apiKey?: string, hostHeader?: string) {
 	try {
 		if (!baseUrl) {
 			return []
@@ -341,9 +354,18 @@ export async function getOpenAiModels(baseUrl?: string, apiKey?: string) {
 		}
 
 		const config: Record<string, any> = {}
+		const headers: Record<string, string> = {}
 
 		if (apiKey) {
-			config["headers"] = { Authorization: `Bearer ${apiKey}` }
+			headers["Authorization"] = `Bearer ${apiKey}`
+		}
+
+		if (hostHeader) {
+			headers["Host"] = hostHeader
+		}
+
+		if (Object.keys(headers).length > 0) {
+			config["headers"] = headers
 		}
 
 		const response = await axios.get(`${baseUrl}/models`, config)

src/core/webview/webviewMessageHandler.ts

@@ -423,7 +423,11 @@ export const webviewMessageHandler = async (provider: ClineProvider, message: We
 			break
 		case "refreshOpenAiModels":
 			if (message?.values?.baseUrl && message?.values?.apiKey) {
-				const openAiModels = await getOpenAiModels(message?.values?.baseUrl, message?.values?.apiKey)
+				const openAiModels = await getOpenAiModels(
+					message?.values?.baseUrl,
+					message?.values?.apiKey,
+					message?.values?.hostHeader,
+				)
 				provider.postMessageToWebview({ type: "openAiModels", openAiModels })
 			}
 

src/exports/roo-code.d.ts

@@ -86,6 +86,8 @@ type ProviderSettings = {
 	vertexRegion?: string | undefined
 	openAiBaseUrl?: string | undefined
 	openAiApiKey?: string | undefined
+	openAiHostHeader?: string | undefined
+	openAiLegacyFormat?: boolean | undefined
 	openAiR1FormatEnabled?: boolean | undefined
 	openAiModelId?: string | undefined
 	openAiCustomModelInfo?:

src/exports/types.ts

@@ -87,6 +87,8 @@ type ProviderSettings = {
 	vertexRegion?: string | undefined
 	openAiBaseUrl?: string | undefined
 	openAiApiKey?: string | undefined
+	openAiHostHeader?: string | undefined
+	openAiLegacyFormat?: boolean | undefined
 	openAiR1FormatEnabled?: boolean | undefined
 	openAiModelId?: string | undefined
 	openAiCustomModelInfo?:

src/schemas/index.ts

@@ -338,6 +338,8 @@ export const providerSettingsSchema = z.object({
 	// OpenAI
 	openAiBaseUrl: z.string().optional(),
 	openAiApiKey: z.string().optional(),
+	openAiHostHeader: z.string().optional(),
+	openAiLegacyFormat: z.boolean().optional(),
 	openAiR1FormatEnabled: z.boolean().optional(),
 	openAiModelId: z.string().optional(),
 	openAiCustomModelInfo: modelInfoSchema.nullish(),
@@ -431,6 +433,8 @@ const providerSettingsRecord: ProviderSettingsRecord = {
 	// OpenAI
 	openAiBaseUrl: undefined,
 	openAiApiKey: undefined,
+	openAiHostHeader: undefined,
+	openAiLegacyFormat: undefined,
 	openAiR1FormatEnabled: undefined,
 	openAiModelId: undefined,
 	openAiCustomModelInfo: undefined,

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

@@ -103,6 +103,8 @@ const ApiOptions = ({
 	const [anthropicBaseUrlSelected, setAnthropicBaseUrlSelected] = useState(!!apiConfiguration?.anthropicBaseUrl)
 	const [azureApiVersionSelected, setAzureApiVersionSelected] = useState(!!apiConfiguration?.azureApiVersion)
 	const [openRouterBaseUrlSelected, setOpenRouterBaseUrlSelected] = useState(!!apiConfiguration?.openRouterBaseUrl)
+	const [openAiHostHeaderSelected, setOpenAiHostHeaderSelected] = useState(!!apiConfiguration?.openAiHostHeader)
+	const [openAiLegacyFormatSelected, setOpenAiLegacyFormatSelected] = useState(!!apiConfiguration?.openAiLegacyFormat)
 	const [googleGeminiBaseUrlSelected, setGoogleGeminiBaseUrlSelected] = useState(
 		!!apiConfiguration?.googleGeminiBaseUrl,
 	)
@@ -145,7 +147,11 @@ const ApiOptions = ({
 			} else if (selectedProvider === "openai") {
 				vscode.postMessage({
 					type: "refreshOpenAiModels",
-					values: { baseUrl: apiConfiguration?.openAiBaseUrl, apiKey: apiConfiguration?.openAiApiKey },
+					values: {
+						baseUrl: apiConfiguration?.openAiBaseUrl,
+						apiKey: apiConfiguration?.openAiApiKey,
+						hostHeader: apiConfiguration?.openAiHostHeader,
+					},
 				})
 			} else if (selectedProvider === "ollama") {
 				vscode.postMessage({ type: "requestOllamaModels", text: apiConfiguration?.ollamaBaseUrl })
@@ -779,6 +785,16 @@ const ApiOptions = ({
 						onChange={handleInputChange("openAiR1FormatEnabled", noTransform)}
 						openAiR1FormatEnabled={apiConfiguration?.openAiR1FormatEnabled ?? false}
 					/>
+					<div>
+						<Checkbox
+							checked={openAiLegacyFormatSelected}
+							onChange={(checked: boolean) => {
+								setOpenAiLegacyFormatSelected(checked)
+								setApiConfigurationField("openAiLegacyFormat", checked)
+							}}>
+							{t("settings:providers.useLegacyFormat")}
+						</Checkbox>
+					</div>
 					<Checkbox
 						checked={apiConfiguration?.openAiStreamingEnabled ?? true}
 						onChange={handleInputChange("openAiStreamingEnabled", noTransform)}>
@@ -811,6 +827,28 @@ const ApiOptions = ({
 						)}
 					</div>
 
+					<div>
+						<Checkbox
+							checked={openAiHostHeaderSelected}
+							onChange={(checked: boolean) => {
+								setOpenAiHostHeaderSelected(checked)
+
+								if (!checked) {
+									setApiConfigurationField("openAiHostHeader", "")
+								}
+							}}>
+							{t("settings:providers.useHostHeader")}
+						</Checkbox>
+						{openAiHostHeaderSelected && (
+							<VSCodeTextField
+								value={apiConfiguration?.openAiHostHeader || ""}
+								onInput={handleInputChange("openAiHostHeader")}
+								placeholder="custom-api-hostname.example.com"
+								className="w-full mt-1"
+							/>
+						)}
+					</div>
+
 					<div className="flex flex-col gap-3">
 						<div className="text-sm text-vscode-descriptionForeground">
 							{t("settings:providers.customModel.capabilities")}

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

@@ -105,6 +105,8 @@
 		"awsCustomArnDesc": "Assegureu-vos que la regió a l'ARN coincideix amb la regió d'AWS seleccionada anteriorment.",
 		"apiKeyStorageNotice": "Les claus API s'emmagatzemen de forma segura a l'Emmagatzematge Secret de VSCode",
 		"useCustomBaseUrl": "Utilitzar URL base personalitzada",
+		"useHostHeader": "Utilitzar capçalera Host personalitzada",
+		"useLegacyFormat": "Utilitzar el format d'API OpenAI antic",
 		"openRouterTransformsText": "Comprimir prompts i cadenes de missatges a la mida del context (<a>Transformacions d'OpenRouter</a>)",
 		"model": "Model",
 		"getOpenRouterApiKey": "Obtenir clau API d'OpenRouter",

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

@@ -109,6 +109,8 @@
 		"glamaApiKey": "Glama API-Schlüssel",
 		"getGlamaApiKey": "Glama API-Schlüssel erhalten",
 		"useCustomBaseUrl": "Benutzerdefinierte Basis-URL verwenden",
+		"useHostHeader": "Benutzerdefinierten Host-Header verwenden",
+		"useLegacyFormat": "Altes OpenAI API-Format verwenden",
 		"requestyApiKey": "Requesty API-Schlüssel",
 		"getRequestyApiKey": "Requesty API-Schlüssel erhalten",
 		"openRouterTransformsText": "Prompts und Nachrichtenketten auf Kontextgröße komprimieren (<a>OpenRouter Transformationen</a>)",

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

@@ -109,6 +109,8 @@
 		"glamaApiKey": "Glama API Key",
 		"getGlamaApiKey": "Get Glama API Key",
 		"useCustomBaseUrl": "Use custom base URL",
+		"useHostHeader": "Use custom Host header",
+		"useLegacyFormat": "Use legacy OpenAI API format",
 		"requestyApiKey": "Requesty API Key",
 		"getRequestyApiKey": "Get Requesty API Key",
 		"openRouterTransformsText": "Compress prompts and message chains to the context size (<a>OpenRouter Transforms</a>)",

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

@@ -109,6 +109,8 @@
 		"glamaApiKey": "Clave API de Glama",
 		"getGlamaApiKey": "Obtener clave API de Glama",
 		"useCustomBaseUrl": "Usar URL base personalizada",
+		"useHostHeader": "Usar encabezado Host personalizado",
+		"useLegacyFormat": "Usar formato API de OpenAI heredado",
 		"requestyApiKey": "Clave API de Requesty",
 		"getRequestyApiKey": "Obtener clave API de Requesty",
 		"openRouterTransformsText": "Comprimir prompts y cadenas de mensajes al tamaño del contexto (<a>Transformaciones de OpenRouter</a>)",

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

@@ -109,6 +109,8 @@
 		"glamaApiKey": "Clé API Glama",
 		"getGlamaApiKey": "Obtenir la clé API Glama",
 		"useCustomBaseUrl": "Utiliser une URL de base personnalisée",
+		"useHostHeader": "Utiliser un en-tête Host personnalisé",
+		"useLegacyFormat": "Utiliser le format API OpenAI hérité",
 		"requestyApiKey": "Clé API Requesty",
 		"getRequestyApiKey": "Obtenir la clé API Requesty",
 		"openRouterTransformsText": "Compresser les prompts et chaînes de messages à la taille du contexte (<a>Transformations OpenRouter</a>)",

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

@@ -109,6 +109,8 @@
 		"glamaApiKey": "Glama API कुंजी",
 		"getGlamaApiKey": "Glama API कुंजी प्राप्त करें",
 		"useCustomBaseUrl": "कस्टम बेस URL का उपयोग करें",
+		"useHostHeader": "कस्टम होस्ट हेडर का उपयोग करें",
+		"useLegacyFormat": "पुराने OpenAI API प्रारूप का उपयोग करें",
 		"requestyApiKey": "Requesty API कुंजी",
 		"getRequestyApiKey": "Requesty API कुंजी प्राप्त करें",
 		"openRouterTransformsText": "संदर्भ आकार के लिए प्रॉम्प्ट और संदेश श्रृंखलाओं को संपीड़ित करें (<a>OpenRouter ट्रांसफॉर्म</a>)",

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

@@ -109,6 +109,8 @@
 		"glamaApiKey": "Chiave API Glama",
 		"getGlamaApiKey": "Ottieni chiave API Glama",
 		"useCustomBaseUrl": "Usa URL base personalizzato",
+		"useHostHeader": "Usa intestazione Host personalizzata",
+		"useLegacyFormat": "Usa formato API OpenAI legacy",
 		"requestyApiKey": "Chiave API Requesty",
 		"getRequestyApiKey": "Ottieni chiave API Requesty",
 		"openRouterTransformsText": "Comprimi prompt e catene di messaggi alla dimensione del contesto (<a>Trasformazioni OpenRouter</a>)",

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

@@ -109,6 +109,8 @@
 		"glamaApiKey": "Glama APIキー",
 		"getGlamaApiKey": "Glama APIキーを取得",
 		"useCustomBaseUrl": "カスタムベースURLを使用",
+		"useHostHeader": "カスタムHostヘッダーを使用",
+		"useLegacyFormat": "レガシーOpenAI API形式を使用",
 		"requestyApiKey": "Requesty APIキー",
 		"getRequestyApiKey": "Requesty APIキーを取得",
 		"openRouterTransformsText": "プロンプトとメッセージチェーンをコンテキストサイズに圧縮 (<a>OpenRouter Transforms</a>)",

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

@@ -109,6 +109,8 @@
 		"glamaApiKey": "Glama API 키",
 		"getGlamaApiKey": "Glama API 키 받기",
 		"useCustomBaseUrl": "사용자 정의 기본 URL 사용",
+		"useHostHeader": "사용자 정의 Host 헤더 사용",
+		"useLegacyFormat": "레거시 OpenAI API 형식 사용",
 		"requestyApiKey": "Requesty API 키",
 		"getRequestyApiKey": "Requesty API 키 받기",
 		"openRouterTransformsText": "프롬프트와 메시지 체인을 컨텍스트 크기로 압축 (<a>OpenRouter Transforms</a>)",

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

@@ -109,6 +109,8 @@
 		"glamaApiKey": "Klucz API Glama",
 		"getGlamaApiKey": "Uzyskaj klucz API Glama",
 		"useCustomBaseUrl": "Użyj niestandardowego URL bazowego",
+		"useHostHeader": "Użyj niestandardowego nagłówka Host",
+		"useLegacyFormat": "Użyj starszego formatu API OpenAI",
 		"requestyApiKey": "Klucz API Requesty",
 		"getRequestyApiKey": "Uzyskaj klucz API Requesty",
 		"openRouterTransformsText": "Kompresuj podpowiedzi i łańcuchy wiadomości do rozmiaru kontekstu (<a>Transformacje OpenRouter</a>)",

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

@@ -109,6 +109,8 @@
 		"glamaApiKey": "Chave de API Glama",
 		"getGlamaApiKey": "Obter chave de API Glama",
 		"useCustomBaseUrl": "Usar URL base personalizado",
+		"useHostHeader": "Usar cabeçalho Host personalizado",
+		"useLegacyFormat": "Usar formato de API OpenAI legado",
 		"requestyApiKey": "Chave de API Requesty",
 		"getRequestyApiKey": "Obter chave de API Requesty",
 		"openRouterTransformsText": "Comprimir prompts e cadeias de mensagens para o tamanho do contexto (<a>Transformações OpenRouter</a>)",

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

@@ -109,6 +109,8 @@
 		"glamaApiKey": "Glama API Anahtarı",
 		"getGlamaApiKey": "Glama API Anahtarı Al",
 		"useCustomBaseUrl": "Özel temel URL kullan",
+		"useHostHeader": "Özel Host başlığı kullan",
+		"useLegacyFormat": "Eski OpenAI API formatını kullan",
 		"requestyApiKey": "Requesty API Anahtarı",
 		"getRequestyApiKey": "Requesty API Anahtarı Al",
 		"openRouterTransformsText": "İstem ve mesaj zincirlerini bağlam boyutuna sıkıştır (<a>OpenRouter Dönüşümleri</a>)",

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

@@ -109,6 +109,8 @@
 		"glamaApiKey": "Khóa API Glama",
 		"getGlamaApiKey": "Lấy khóa API Glama",
 		"useCustomBaseUrl": "Sử dụng URL cơ sở tùy chỉnh",
+		"useHostHeader": "Sử dụng tiêu đề Host tùy chỉnh",
+		"useLegacyFormat": "Sử dụng định dạng API OpenAI cũ",
 		"requestyApiKey": "Khóa API Requesty",
 		"getRequestyApiKey": "Lấy khóa API Requesty",
 		"anthropicApiKey": "Khóa API Anthropic",

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

@@ -107,6 +107,8 @@
 		"getOpenRouterApiKey": "获取 OpenRouter API 密钥",
 		"apiKeyStorageNotice": "API 密钥安全存储在 VSCode 的密钥存储中",
 		"useCustomBaseUrl": "使用自定义基础 URL",
+		"useHostHeader": "使用自定义 Host 标头",
+		"useLegacyFormat": "使用传统 OpenAI API 格式",
 		"glamaApiKey": "Glama API 密钥",
 		"getGlamaApiKey": "获取 Glama API 密钥",
 		"requestyApiKey": "Requesty API 密钥",

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

@@ -109,6 +109,8 @@
 		"glamaApiKey": "Glama API 金鑰",
 		"getGlamaApiKey": "取得 Glama API 金鑰",
 		"useCustomBaseUrl": "使用自訂基礎 URL",
+		"useHostHeader": "使用自訂 Host 標頭",
+		"useLegacyFormat": "使用舊版 OpenAI API 格式",
 		"requestyApiKey": "Requesty API 金鑰",
 		"getRequestyApiKey": "取得 Requesty API 金鑰",
 		"openRouterTransformsText": "將提示和訊息鏈壓縮到上下文大小 (<a>OpenRouter 轉換</a>)",