@lobehub/chat 1.106.8 → 1.107.0

package/src/app/(backend)/middleware/auth/index.ts

@@ -9,7 +9,7 @@ import {
   enableClerk,
 } from '@/const/auth';
 import { ClerkAuth } from '@/libs/clerk-auth';
-import { AgentRuntime, AgentRuntimeError, ChatCompletionErrorPayload } from '@/libs/model-runtime';
+import { AgentRuntimeError, ChatCompletionErrorPayload, ModelRuntime } from '@/libs/model-runtime';
 import { validateOIDCJWT } from '@/libs/oidc-provider/jwt';
 import { ChatErrorType } from '@/types/fetch';
 import { createErrorResponse } from '@/utils/errorResponse';
@@ -17,7 +17,7 @@ import { getXorPayload } from '@/utils/server/xor';
 
 import { checkAuthMethod } from './utils';
 
-type CreateRuntime = (jwtPayload: ClientSecretPayload) => AgentRuntime;
+type CreateRuntime = (jwtPayload: ClientSecretPayload) => ModelRuntime;
 type RequestOptions = { createRuntime?: CreateRuntime; params: Promise<{ provider: string }> };
 
 export type RequestHandler = (

package/src/app/(backend)/webapi/chat/[provider]/route.test.ts

@@ -4,7 +4,7 @@ import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
 
 import { checkAuthMethod } from '@/app/(backend)/middleware/auth/utils';
 import { LOBE_CHAT_AUTH_HEADER, OAUTH_AUTHORIZED } from '@/const/auth';
-import { AgentRuntime, LobeRuntimeAI } from '@/libs/model-runtime';
+import { LobeRuntimeAI, ModelRuntime } from '@/libs/model-runtime';
 import { ChatErrorType } from '@/types/fetch';
 import { getXorPayload } from '@/utils/server/xor';
 
@@ -57,10 +57,10 @@ afterEach(() => {
 
 describe('POST handler', () => {
   describe('init chat model', () => {
-    it('should initialize AgentRuntime correctly with valid authorization', async () => {
+    it('should initialize ModelRuntime correctly with valid authorization', async () => {
       const mockParams = Promise.resolve({ provider: 'test-provider' });
 
-      // 设置 getJWTPayload 和 initAgentRuntimeWithUserPayload 的模拟返回值
+      // 设置 getJWTPayload 和 initModelRuntimeWithUserPayload 的模拟返回值
       vi.mocked(getXorPayload).mockReturnValueOnce({
         accessCode: 'test-access-code',
         apiKey: 'test-api-key',
@@ -69,10 +69,10 @@ describe('POST handler', () => {
 
       const mockRuntime: LobeRuntimeAI = { baseURL: 'abc', chat: vi.fn() };
 
-      // migrate to new AgentRuntime init api
+      // migrate to new ModelRuntime init api
       const spy = vi
-        .spyOn(AgentRuntime, 'initializeWithProvider')
-        .mockResolvedValue(new AgentRuntime(mockRuntime));
+        .spyOn(ModelRuntime, 'initializeWithProvider')
+        .mockResolvedValue(new ModelRuntime(mockRuntime));
 
       // 调用 POST 函数
       await POST(request as unknown as Request, { params: mockParams });
@@ -111,14 +111,14 @@ describe('POST handler', () => {
       });
 
       const mockParams = Promise.resolve({ provider: 'test-provider' });
-      // 设置 initAgentRuntimeWithUserPayload 的模拟返回值
+      // 设置 initModelRuntimeWithUserPayload 的模拟返回值
       vi.mocked(getAuth).mockReturnValue({} as any);
       vi.mocked(checkAuthMethod).mockReset();
 
       const mockRuntime: LobeRuntimeAI = { baseURL: 'abc', chat: vi.fn() };
 
-      vi.spyOn(AgentRuntime, 'initializeWithProvider').mockResolvedValue(
-        new AgentRuntime(mockRuntime),
+      vi.spyOn(ModelRuntime, 'initializeWithProvider').mockResolvedValue(
+        new ModelRuntime(mockRuntime),
       );
 
       const request = new Request(new URL('https://test.com'), {
@@ -178,12 +178,12 @@ describe('POST handler', () => {
 
       const mockChatResponse: any = { success: true, message: 'Reply from agent' };
 
-      vi.spyOn(AgentRuntime.prototype, 'chat').mockResolvedValue(mockChatResponse);
+      vi.spyOn(ModelRuntime.prototype, 'chat').mockResolvedValue(mockChatResponse);
 
       const response = await POST(request as unknown as Request, { params: mockParams });
 
       expect(response).toEqual(mockChatResponse);
-      expect(AgentRuntime.prototype.chat).toHaveBeenCalledWith(mockChatPayload, {
+      expect(ModelRuntime.prototype.chat).toHaveBeenCalledWith(mockChatPayload, {
         user: 'abc',
         signal: expect.anything(),
       });
@@ -210,7 +210,7 @@ describe('POST handler', () => {
         errorMessage: 'Something went wrong',
       };
 
-      vi.spyOn(AgentRuntime.prototype, 'chat').mockRejectedValue(mockErrorResponse);
+      vi.spyOn(ModelRuntime.prototype, 'chat').mockRejectedValue(mockErrorResponse);
 
       const response = await POST(request, { params: mockParams });
 

package/src/app/(backend)/webapi/chat/[provider]/route.ts

@@ -1,10 +1,10 @@
 import { checkAuth } from '@/app/(backend)/middleware/auth';
 import {
   AGENT_RUNTIME_ERROR_SET,
-  AgentRuntime,
   ChatCompletionErrorPayload,
+  ModelRuntime,
 } from '@/libs/model-runtime';
-import { createTraceOptions, initAgentRuntimeWithUserPayload } from '@/server/modules/AgentRuntime';
+import { createTraceOptions, initModelRuntimeWithUserPayload } from '@/server/modules/ModelRuntime';
 import { ChatErrorType } from '@/types/fetch';
 import { ChatStreamPayload } from '@/types/openai/chat';
 import { createErrorResponse } from '@/utils/errorResponse';
@@ -17,11 +17,11 @@ export const POST = checkAuth(async (req: Request, { params, jwtPayload, createR
 
   try {
     // ============  1. init chat model   ============ //
-    let agentRuntime: AgentRuntime;
+    let modelRuntime: ModelRuntime;
     if (createRuntime) {
-      agentRuntime = createRuntime(jwtPayload);
+      modelRuntime = createRuntime(jwtPayload);
     } else {
-      agentRuntime = await initAgentRuntimeWithUserPayload(provider, jwtPayload);
+      modelRuntime = await initModelRuntimeWithUserPayload(provider, jwtPayload);
     }
 
     // ============  2. create chat completion   ============ //
@@ -36,7 +36,7 @@ export const POST = checkAuth(async (req: Request, { params, jwtPayload, createR
       traceOptions = createTraceOptions(data, { provider, trace: tracePayload });
     }
 
-    return await agentRuntime.chat(data, {
+    return await modelRuntime.chat(data, {
       user: jwtPayload.userId,
       ...traceOptions,
       signal: req.signal,

package/src/app/(backend)/webapi/chat/vertexai/route.ts

@@ -1,5 +1,5 @@
 import { checkAuth } from '@/app/(backend)/middleware/auth';
-import { AgentRuntime, ModelProvider } from '@/libs/model-runtime';
+import { ModelProvider, ModelRuntime } from '@/libs/model-runtime';
 import { LobeVertexAI } from '@/libs/model-runtime/vertexai';
 import { safeParseJSON } from '@/utils/safeParseJSON';
 
@@ -28,7 +28,7 @@ export const POST = checkAuth(async (req: Request, { jwtPayload }) =>
         project: !!credentials?.project_id ? credentials?.project_id : process.env.VERTEXAI_PROJECT,
       });
 
-      return new AgentRuntime(instance);
+      return new ModelRuntime(instance);
     },
     params: Promise.resolve({ provider: ModelProvider.VertexAI }),
   }),

package/src/app/(backend)/webapi/models/[provider]/pull/route.ts

@@ -1,6 +1,6 @@
 import { checkAuth } from '@/app/(backend)/middleware/auth';
 import { ChatCompletionErrorPayload, PullModelParams } from '@/libs/model-runtime';
-import { initAgentRuntimeWithUserPayload } from '@/server/modules/AgentRuntime';
+import { initModelRuntimeWithUserPayload } from '@/server/modules/ModelRuntime';
 import { ChatErrorType } from '@/types/fetch';
 import { createErrorResponse } from '@/utils/errorResponse';
 
@@ -10,7 +10,7 @@ export const POST = checkAuth(async (req, { params, jwtPayload }) => {
   const { provider } = await params;
 
   try {
-    const agentRuntime = await initAgentRuntimeWithUserPayload(provider, jwtPayload);
+    const agentRuntime = await initModelRuntimeWithUserPayload(provider, jwtPayload);
 
     const data = (await req.json()) as PullModelParams;
 

package/src/app/(backend)/webapi/models/[provider]/route.ts

@@ -2,7 +2,7 @@ import { NextResponse } from 'next/server';
 
 import { checkAuth } from '@/app/(backend)/middleware/auth';
 import { ChatCompletionErrorPayload, ModelProvider } from '@/libs/model-runtime';
-import { initAgentRuntimeWithUserPayload } from '@/server/modules/AgentRuntime';
+import { initModelRuntimeWithUserPayload } from '@/server/modules/ModelRuntime';
 import { ChatErrorType } from '@/types/fetch';
 import { createErrorResponse } from '@/utils/errorResponse';
 
@@ -16,7 +16,7 @@ export const GET = checkAuth(async (req, { params, jwtPayload }) => {
   try {
     const hasDefaultApiKey = jwtPayload.apiKey || 'dont-need-api-key-for-model-list';
 
-    const agentRuntime = await initAgentRuntimeWithUserPayload(provider, {
+    const agentRuntime = await initModelRuntimeWithUserPayload(provider, {
       ...jwtPayload,
       apiKey: noNeedAPIKey(provider) ? hasDefaultApiKey : jwtPayload.apiKey,
     });

package/src/app/(backend)/webapi/text-to-image/[provider]/route.ts

@@ -3,7 +3,7 @@ import { NextResponse } from 'next/server';
 import { checkAuth } from '@/app/(backend)/middleware/auth';
 import { ChatCompletionErrorPayload } from '@/libs/model-runtime';
 import { TextToImagePayload } from '@/libs/model-runtime/types';
-import { initAgentRuntimeWithUserPayload } from '@/server/modules/AgentRuntime';
+import { initModelRuntimeWithUserPayload } from '@/server/modules/ModelRuntime';
 import { ChatErrorType } from '@/types/fetch';
 import { createErrorResponse } from '@/utils/errorResponse';
 
@@ -52,7 +52,7 @@ export const POST = checkAuth(async (req: Request, { params, jwtPayload }) => {
 
   try {
     // ============  1. init chat model   ============ //
-    const agentRuntime = await initAgentRuntimeWithUserPayload(provider, jwtPayload);
+    const agentRuntime = await initModelRuntimeWithUserPayload(provider, jwtPayload);
 
     // ============  2. create chat completion   ============ //
 

package/src/app/[variants]/(main)/settings/provider/(detail)/github/page.tsx

@@ -48,10 +48,10 @@ const useProviderCard = (): ProviderItem => {
         ),
         desc: (
           <Markdown className={styles.markdown} fontSize={12} variant={'chat'}>
-            {t(`github.personalAccessToken.desc`)}
+            {t('github.personalAccessToken.desc')}
           </Markdown>
         ),
-        label: t(`github.personalAccessToken.title`),
+        label: t('github.personalAccessToken.title'),
         name: [KeyVaultsConfigKey, LLMProviderApiTokenKey],
       },
     ],

package/src/app/[variants]/(main)/settings/provider/features/ProviderConfig/index.tsx

@@ -97,6 +97,7 @@ const useStyles = createStyles(({ css, prefixCls, responsive, token }) => ({
 
 export interface ProviderConfigProps extends Omit<AiProviderDetailItem, 'enabled' | 'source'> {
   apiKeyItems?: FormItemProps[];
+  apiKeyUrl?: string;
   canDeactivate?: boolean;
   checkErrorRender?: CheckErrorRender;
   className?: string;
@@ -127,6 +128,7 @@ const ProviderConfig = memo<ProviderConfigProps>(
     showAceGcm = true,
     extra,
     source = AiProviderSourceEnum.Builtin,
+    apiKeyUrl,
   }) => {
     const {
       proxyUrl,
@@ -184,7 +186,7 @@ const ProviderConfig = memo<ProviderConfigProps>(
             ) : (
               <FormPassword
                 autoComplete={'new-password'}
-                placeholder={t(`providerModels.config.apiKey.placeholder`, { name })}
+                placeholder={t('providerModels.config.apiKey.placeholder', { name })}
                 suffix={
                   configUpdating && (
                     <Icon icon={Loader2Icon} spin style={{ color: theme.colorTextTertiary }} />
@@ -192,7 +194,20 @@ const ProviderConfig = memo<ProviderConfigProps>(
                 }
               />
             ),
-            desc: t(`providerModels.config.apiKey.desc`, { name }),
+            desc: apiKeyUrl ? (
+              <Trans
+                i18nKey="providerModels.config.apiKey.descWithUrl"
+                ns={'modelProvider'}
+                value={{ name }}
+              >
+                请填写你的 {{ name }} API Key,
+                <Link href={apiKeyUrl} target={'_blank'}>
+                  点此获取
+                </Link>
+              </Trans>
+            ) : (
+              t(`providerModels.config.apiKey.desc`, { name })
+            ),
             label: t(`providerModels.config.apiKey.title`),
             name: [KeyVaultsConfigKey, LLMProviderApiTokenKey],
           },

package/src/config/aiModels/aihubmix.ts

@@ -0,0 +1,164 @@
+import { AIChatModelCard } from '@/types/aiModel';
+
+const aihubmixModels: AIChatModelCard[] = [
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+    },
+    contextWindowTokens: 65_536,
+    description: 'DeepSeek R1 推理模型，具有强大的推理能力',
+    displayName: 'DeepSeek R1',
+    enabled: true,
+    id: 'DeepSeek-R1',
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+      vision: true,
+    },
+    contextWindowTokens: 200_000,
+    description:
+      'Claude Opus 4 是 Anthropic 迄今为止最强大的模型，专为处理复杂、长时间运行的任务而设计。',
+    displayName: 'Claude Opus 4',
+    enabled: true,
+    id: 'claude-opus-4-20250514',
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+      vision: true,
+    },
+    contextWindowTokens: 200_000,
+    description:
+      'Claude Sonnet 4 是一款高效且性价比高的模型，作为 Claude Sonnet 3.7 的升级版，适合日常任务和中等复杂度的应用。',
+    displayName: 'Claude Sonnet 4',
+    enabled: true,
+    id: 'claude-sonnet-4-20250514',
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+      vision: true,
+    },
+    contextWindowTokens: 200_000,
+    description: 'OpenAI o3 推理模型，具有强大的推理能力',
+    displayName: 'o3',
+    enabled: true,
+    id: 'o3',
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+      vision: true,
+    },
+    contextWindowTokens: 200_000,
+    description: 'OpenAI o4-mini 小型推理模型，高效且经济',
+    displayName: 'o4-mini',
+    enabled: true,
+    id: 'o4-mini',
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+      vision: true,
+    },
+    contextWindowTokens: 1_047_576,
+    description: 'GPT-4.1 旗舰模型，适用于复杂任务',
+    displayName: 'GPT-4.1',
+    enabled: true,
+    id: 'gpt-4.1',
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+      vision: true,
+    },
+    contextWindowTokens: 1_047_576,
+    description: 'GPT-4.1 mini 平衡智能、速度和成本',
+    displayName: 'GPT-4.1 mini',
+    enabled: true,
+    id: 'gpt-4.1-mini',
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+      search: true,
+      vision: true,
+    },
+    contextWindowTokens: 1_048_576 + 65_536,
+    description:
+      'Gemini 2.5 Pro 是 Google 最先进的思维模型，能够对代码、数学和STEM领域的复杂问题进行推理，以及使用长上下文分析大型数据集、代码库和文档。',
+    displayName: 'Gemini 2.5 Pro',
+    enabled: true,
+    id: 'gemini-2.5-pro',
+    maxOutput: 65_536,
+    pricing: {
+      input: 1.25, // prompts <= 200k tokens
+      output: 10, // prompts <= 200k tokens
+    },
+    releasedAt: '2025-06-17',
+    settings: {
+      extendParams: ['thinkingBudget'],
+      searchImpl: 'params',
+      searchProvider: 'google',
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+      search: true,
+      vision: true,
+    },
+    contextWindowTokens: 1_000_000,
+    description: 'Gemini 2.5 Flash 预览版，快速高效的多模态模型',
+    displayName: 'Gemini 2.5 Flash',
+    enabled: true,
+    id: 'gemini-2.5-flash',
+    releasedAt: '2025-06-17',
+    settings: {
+      extendParams: ['thinkingBudget'],
+      searchImpl: 'params',
+      searchProvider: 'google',
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+    },
+    contextWindowTokens: 235_000,
+    description: 'Qwen3 235B 大型语言模型',
+    displayName: 'Qwen3 235B',
+    enabled: true,
+    id: 'Qwen/Qwen3-235B-A22B',
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+    },
+    contextWindowTokens: 32_000,
+    description: 'Qwen3 32B 中型语言模型',
+    displayName: 'Qwen3 32B',
+    enabled: true,
+    id: 'Qwen/Qwen3-32B',
+    type: 'chat',
+  },
+];
+
+export default aihubmixModels;

package/src/config/aiModels/index.ts

@@ -2,6 +2,7 @@ import { AiFullModelCard, LobeDefaultAiModelListItem } from '@/types/aiModel';
 
 import { default as ai21 } from './ai21';
 import { default as ai360 } from './ai360';
+import { default as aihubmix } from './aihubmix';
 import { default as anthropic } from './anthropic';
 import { default as azure } from './azure';
 import { default as azureai } from './azureai';
@@ -78,6 +79,7 @@ const buildDefaultModelList = (map: ModelsMap): LobeDefaultAiModelListItem[] =>
 export const LOBE_DEFAULT_MODEL_LIST = buildDefaultModelList({
   ai21,
   ai360,
+  aihubmix,
   anthropic,
   azure,
   azureai,
@@ -135,6 +137,7 @@ export const LOBE_DEFAULT_MODEL_LIST = buildDefaultModelList({
 
 export { default as ai21 } from './ai21';
 export { default as ai360 } from './ai360';
+export { default as aihubmix } from './aihubmix';
 export { default as anthropic } from './anthropic';
 export { default as azure } from './azure';
 export { default as azureai } from './azureai';

package/src/config/llm.ts

@@ -171,6 +171,9 @@ export const getLLMConfig = () => {
 
       ENABLED_V0: z.boolean(),
       V0_API_KEY: z.string().optional(),
+
+      ENABLED_AIHUBMIX: z.boolean(),
+      AIHUBMIX_API_KEY: z.string().optional(),
     },
     runtimeEnv: {
       API_KEY_SELECT_MODE: process.env.API_KEY_SELECT_MODE,
@@ -340,6 +343,9 @@ export const getLLMConfig = () => {
 
       ENABLED_V0: !!process.env.V0_API_KEY,
       V0_API_KEY: process.env.V0_API_KEY,
+
+      ENABLED_AIHUBMIX: !!process.env.AIHUBMIX_API_KEY,
+      AIHUBMIX_API_KEY: process.env.AIHUBMIX_API_KEY,
     },
   });
 };

package/src/config/modelProviders/aihubmix.ts

@@ -0,0 +1,18 @@
+import { ModelProviderCard } from '@/types/llm';
+
+const AiHubMix: ModelProviderCard = {
+  apiKeyUrl: 'https://lobe.li/9mZhb4T',
+  chatModels: [],
+  checkModel: 'gpt-4.1-mini',
+  description: 'AiHubMix 通过统一的 API 接口提供对多种 AI 模型的访问。',
+  id: 'aihubmix',
+  modelsUrl: 'https://docs.aihubmix.com/cn/api/Model-List',
+  name: 'AiHubMix',
+  settings: {
+    sdkType: 'router',
+    showModelFetcher: true,
+  },
+  url: 'https://aihubmix.com?utm_source=lobehub',
+};
+
+export default AiHubMix;

package/src/config/modelProviders/huggingface.ts

@@ -1,6 +1,7 @@
 import { ModelProviderCard } from '@/types/llm';
 
 const HuggingFace: ModelProviderCard = {
+  apiKeyUrl: 'https://huggingface.co/settings/tokens',
   chatModels: [
     {
       contextWindowTokens: 32_768,

package/src/config/modelProviders/index.ts

@@ -2,6 +2,7 @@ import { ChatModelCard, ModelProviderCard } from '@/types/llm';
 
 import Ai21Provider from './ai21';
 import Ai360Provider from './ai360';
+import AiHubMixProvider from './aihubmix';
 import AnthropicProvider from './anthropic';
 import AzureProvider from './azure';
 import AzureAIProvider from './azureai';
@@ -94,6 +95,7 @@ export const LOBE_DEFAULT_MODEL_LIST: ChatModelCard[] = [
   TaichuProvider.chatModels,
   CloudflareProvider.chatModels,
   Ai360Provider.chatModels,
+  AiHubMixProvider.chatModels,
   SiliconCloudProvider.chatModels,
   GiteeAIProvider.chatModels,
   UpstageProvider.chatModels,
@@ -163,6 +165,7 @@ export const DEFAULT_MODEL_PROVIDER_LIST = [
   GiteeAIProvider,
   TaichuProvider,
   Ai360Provider,
+  AiHubMixProvider,
   Search1APIProvider,
   InfiniAIProvider,
   QiniuProvider,
@@ -179,6 +182,7 @@ export const isProviderDisableBrowserRequest = (id: string) => {
 
 export { default as Ai21ProviderCard } from './ai21';
 export { default as Ai360ProviderCard } from './ai360';
+export { default as AiHubMixProviderCard } from './aihubmix';
 export { default as AnthropicProviderCard } from './anthropic';
 export { default as AzureProviderCard } from './azure';
 export { default as AzureAIProviderCard } from './azureai';

package/src/libs/model-runtime/ModelRuntime.test.ts

@@ -6,10 +6,10 @@ import { beforeEach, describe, expect, it, vi } from 'vitest';
 import * as langfuseCfg from '@/config/langfuse';
 import { ClientSecretPayload } from '@/const/auth';
 import { TraceNameMap } from '@/const/trace';
-import { AgentRuntime, ChatStreamPayload, LobeOpenAI, ModelProvider } from '@/libs/model-runtime';
+import { ChatStreamPayload, LobeOpenAI, ModelProvider, ModelRuntime } from '@/libs/model-runtime';
 import { providerRuntimeMap } from '@/libs/model-runtime/runtimeMap';
 import { CreateImagePayload } from '@/libs/model-runtime/types/image';
-import { createTraceOptions } from '@/server/modules/AgentRuntime';
+import { createTraceOptions } from '@/server/modules/ModelRuntime';
 
 import { AgentChatOptions } from './ModelRuntime';
 
@@ -52,7 +52,7 @@ const testRuntime = (providerId: string, payload?: any) => {
   describe(`${providerId} provider runtime`, () => {
     it('should initialize correctly', async () => {
       const jwtPayload: ClientSecretPayload = { apiKey: 'user-key', ...payload };
-      const runtime = await AgentRuntime.initializeWithProvider(providerId, jwtPayload);
+      const runtime = await ModelRuntime.initializeWithProvider(providerId, jwtPayload);
 
       // @ts-ignore
       expect(runtime['_runtime']).toBeInstanceOf(providerRuntimeMap[providerId]);
@@ -64,16 +64,15 @@ const testRuntime = (providerId: string, payload?: any) => {
   });
 };
 
-let mockModelRuntime: AgentRuntime;
+let mockModelRuntime: ModelRuntime;
 beforeEach(async () => {
   const jwtPayload: ClientSecretPayload = { apiKey: 'user-openai-key', baseURL: 'user-endpoint' };
-  mockModelRuntime = await AgentRuntime.initializeWithProvider(ModelProvider.OpenAI, jwtPayload);
+  mockModelRuntime = await ModelRuntime.initializeWithProvider(ModelProvider.OpenAI, jwtPayload);
 });
 
-describe('AgentRuntime', () => {
+describe('ModelRuntime', () => {
   describe('should initialize with various providers', () => {
     const providers = Object.values(ModelProvider).filter((i) => i !== 'lobehub');
-
     const specialProviderIds = [ModelProvider.VertexAI, ...specialProviders.map((p) => p.id)];
 
     const generalTestProviders = providers.filter(
@@ -87,7 +86,7 @@ describe('AgentRuntime', () => {
     specialProviders.forEach(({ id, payload }) => testRuntime(id, payload));
   });
 
-  describe('AgentRuntime chat method', () => {
+  describe('ModelRuntime chat method', () => {
     it('should run correctly', async () => {
       const payload: ChatStreamPayload = {
         messages: [{ role: 'user', content: 'Hello, world!' }],
@@ -243,7 +242,7 @@ describe('AgentRuntime', () => {
     });
   });
 
-  describe('AgentRuntime createImage method', () => {
+  describe('ModelRuntime createImage method', () => {
     it('should run correctly', async () => {
       const payload: CreateImagePayload = {
         model: 'dall-e-3',
@@ -292,7 +291,7 @@ describe('AgentRuntime', () => {
     });
   });
 
-  describe('AgentRuntime models method', () => {
+  describe('ModelRuntime models method', () => {
     it('should run correctly', async () => {
       const mockModels = [
         { id: 'gpt-4', name: 'GPT-4' },

package/src/libs/model-runtime/ModelRuntime.ts

@@ -25,7 +25,7 @@ export interface AgentChatOptions {
   trace?: TracePayload;
 }
 
-class ModelRuntime {
+export class ModelRuntime {
   private _runtime: LobeRuntimeAI;
 
   constructor(runtime: LobeRuntimeAI) {
@@ -113,10 +113,9 @@ class ModelRuntime {
   ) {
     // @ts-expect-error runtime map not include vertex so it will be undefined
     const providerAI = providerRuntimeMap[provider] ?? LobeOpenAI;
+
     const runtimeModel: LobeRuntimeAI = new providerAI(params);
 
     return new ModelRuntime(runtimeModel);
   }
 }
-
-export default ModelRuntime;

package/src/libs/model-runtime/RouterRuntime/baseRuntimeMap.ts

@@ -0,0 +1,15 @@
+import { LobeAnthropicAI } from '../anthropic';
+import { LobeAzureAI } from '../azureai';
+import { LobeCloudflareAI } from '../cloudflare';
+import { LobeFalAI } from '../fal';
+import { LobeGoogleAI } from '../google';
+import { LobeOpenAI } from '../openai';
+
+export const baseRuntimeMap = {
+  anthropic: LobeAnthropicAI,
+  azure: LobeAzureAI,
+  cloudflare: LobeCloudflareAI,
+  fal: LobeFalAI,
+  google: LobeGoogleAI,
+  openai: LobeOpenAI,
+};