@lobehub/chat 1.124.0 → 1.124.2

package/packages/model-runtime/src/newapi/index.ts

@@ -1,3 +1,4 @@
+import { LOBE_DEFAULT_MODEL_LIST } from 'model-bank';
 import urlJoin from 'url-join';
 
 import { createRouterRuntime } from '../RouterRuntime';
@@ -54,9 +55,6 @@ const getProviderFromOwnedBy = (ownedBy: string): string => {
   return 'openai';
 };
 
-// 全局的模型路由映射，在 models 函数执行后被填充
-let globalModelRouteMap: Map<string, string> = new Map();
-
 export const LobeNewAPIAI = createRouterRuntime({
   debug: {
     chatCompletion: () => process.env.DEBUG_NEWAPI_CHAT_COMPLETION === '1',
@@ -66,180 +64,163 @@ export const LobeNewAPIAI = createRouterRuntime({
   },
   id: ModelProvider.NewAPI,
   models: async ({ client: openAIClient }) => {
-    // 每次调用 models 时清空并重建路由映射
-    globalModelRouteMap.clear();
-
-    // 获取基础 URL（移除末尾的 /v1）
-    const baseURL = openAIClient.baseURL.replace(/\/v1\/?$/, '');
-
-      const modelsPage = (await openAIClient.models.list()) as any;
-      const modelList: NewAPIModelCard[] = modelsPage.data || [];
-
-      // 尝试获取 pricing 信息以补充模型详细信息
-      let pricingMap: Map<string, NewAPIPricing> = new Map();
-      try {
-        // 使用保存的 baseURL
-        const pricingResponse = await fetch(`${baseURL}/api/pricing`, {
-          headers: {
-            Authorization: `Bearer ${openAIClient.apiKey}`,
-          },
-        });
-
-        if (pricingResponse.ok) {
-          const pricingData = await pricingResponse.json();
-          if (pricingData.success && pricingData.data) {
-            (pricingData.data as NewAPIPricing[]).forEach((pricing) => {
-              pricingMap.set(pricing.model_name, pricing);
-            });
-          }
+    // 获取基础 URL（移除末尾的 API 版本路径如 /v1、/v1beta 等）
+    const baseURL = openAIClient.baseURL.replace(/\/v\d+[a-z]*\/?$/, '');
+
+    const modelsPage = (await openAIClient.models.list()) as any;
+    const modelList: NewAPIModelCard[] = modelsPage.data || [];
+
+    // 尝试获取 pricing 信息以补充模型详细信息
+    let pricingMap: Map<string, NewAPIPricing> = new Map();
+    try {
+      // 使用保存的 baseURL
+      const pricingResponse = await fetch(`${baseURL}/api/pricing`, {
+        headers: {
+          Authorization: `Bearer ${openAIClient.apiKey}`,
+        },
+      });
+
+      if (pricingResponse.ok) {
+        const pricingData = await pricingResponse.json();
+        if (pricingData.success && pricingData.data) {
+          (pricingData.data as NewAPIPricing[]).forEach((pricing) => {
+            pricingMap.set(pricing.model_name, pricing);
+          });
         }
-      } catch (error) {
-        // If fetching pricing information fails, continue using the basic model information
-        console.debug('Failed to fetch NewAPI pricing info:', error);
       }
-
-      // Process the model list: determine the provider for each model based on priority rules
-      const enrichedModelList = modelList.map((model) => {
-        let enhancedModel: any = { ...model };
-
-        // 1. 添加 pricing 信息
-        const pricing = pricingMap.get(model.id);
-        if (pricing) {
-          // NewAPI 的价格计算逻辑：
-          // - quota_type: 0 表示按量计费（按 token），1 表示按次计费
-          // - model_ratio: 相对于基础价格的倍率（基础价格 = $0.002/1K tokens）
-          // - model_price: 直接指定的价格（优先使用）
-          // - completion_ratio: 输出价格相对于输入价格的倍率
-          //
-          // LobeChat 需要的格式：美元/百万 token
-
-          let inputPrice: number | undefined;
-          let outputPrice: number | undefined;
-
-          if (pricing.quota_type === 0) {
-            // 按量计费
-            if (pricing.model_price && pricing.model_price > 0) {
-              // model_price is a direct price value; need to confirm its unit.
-              // Assumption: model_price is the price per 1,000 tokens (i.e., $/1K tokens).
-              // To convert to price per 1,000,000 tokens ($/1M tokens), multiply by 1,000,000 / 1,000 = 1,000.
-              // Since the base price is $0.002/1K tokens, multiplying by 2 gives $2/1M tokens.
-              // Therefore, inputPrice = model_price * 2 converts the price to $/1M tokens for LobeChat.
-              inputPrice = pricing.model_price * 2;
-            } else if (pricing.model_ratio) {
-              // model_ratio × $0.002/1K = model_ratio × $2/1M
-              inputPrice = pricing.model_ratio * 2; // 转换为 $/1M tokens
-            }
-
-            if (inputPrice !== undefined) {
-              // 计算输出价格
-              outputPrice = inputPrice * (pricing.completion_ratio || 1);
-
-              enhancedModel.pricing = {
-                input: inputPrice,
-                output: outputPrice,
-              };
-            }
+    } catch (error) {
+      // If fetching pricing information fails, continue using the basic model information
+      console.debug('Failed to fetch NewAPI pricing info:', error);
+    }
+
+    // Process the model list: determine the provider for each model based on priority rules
+    const enrichedModelList = modelList.map((model) => {
+      let enhancedModel: any = { ...model };
+
+      // 1. 添加 pricing 信息
+      const pricing = pricingMap.get(model.id);
+      if (pricing) {
+        // NewAPI 的价格计算逻辑：
+        // - quota_type: 0 表示按量计费（按 token），1 表示按次计费
+        // - model_ratio: 相对于基础价格的倍率（基础价格 = $0.002/1K tokens）
+        // - model_price: 直接指定的价格（优先使用）
+        // - completion_ratio: 输出价格相对于输入价格的倍率
+        //
+        // LobeChat 需要的格式：美元/百万 token
+
+        let inputPrice: number | undefined;
+        let outputPrice: number | undefined;
+
+        if (pricing.quota_type === 0) {
+          // 按量计费
+          if (pricing.model_price && pricing.model_price > 0) {
+            // model_price is a direct price value; need to confirm its unit.
+            // Assumption: model_price is the price per 1,000 tokens (i.e., $/1K tokens).
+            // To convert to price per 1,000,000 tokens ($/1M tokens), multiply by 1,000,000 / 1,000 = 1,000.
+            // Since the base price is $0.002/1K tokens, multiplying by 2 gives $2/1M tokens.
+            // Therefore, inputPrice = model_price * 2 converts the price to $/1M tokens for LobeChat.
+            inputPrice = pricing.model_price * 2;
+          } else if (pricing.model_ratio) {
+            // model_ratio × $0.002/1K = model_ratio × $2/1M
+            inputPrice = pricing.model_ratio * 2; // 转换为 $/1M tokens
           }
-          // quota_type === 1 按次计费暂不支持
-        }
 
-        // 2. 根据优先级处理 provider 信息并缓存路由
-        let detectedProvider = 'openai'; // 默认
-
-        // 优先级1：使用 supported_endpoint_types
-        if (model.supported_endpoint_types && model.supported_endpoint_types.length > 0) {
-          if (model.supported_endpoint_types.includes('anthropic')) {
-            detectedProvider = 'anthropic';
-          } else if (model.supported_endpoint_types.includes('gemini')) {
-            detectedProvider = 'google';
-          } else if (model.supported_endpoint_types.includes('xai')) {
-            detectedProvider = 'xai';
+          if (inputPrice !== undefined) {
+            // 计算输出价格
+            outputPrice = inputPrice * (pricing.completion_ratio || 1);
+
+            enhancedModel.pricing = {
+              input: inputPrice,
+              output: outputPrice,
+            };
           }
         }
-        // 优先级2：使用 owned_by 字段
-        else if (model.owned_by) {
-          detectedProvider = getProviderFromOwnedBy(model.owned_by);
-        }
-        // 优先级3：基于模型名称检测
-        else {
-          detectedProvider = detectModelProvider(model.id);
+        // quota_type === 1 按次计费暂不支持
+      }
+
+      // 2. 根据优先级处理 provider 信息并缓存路由
+      let detectedProvider = 'openai'; // 默认
+
+      // 优先级1：使用 supported_endpoint_types
+      if (model.supported_endpoint_types && model.supported_endpoint_types.length > 0) {
+        if (model.supported_endpoint_types.includes('anthropic')) {
+          detectedProvider = 'anthropic';
+        } else if (model.supported_endpoint_types.includes('gemini')) {
+          detectedProvider = 'google';
+        } else if (model.supported_endpoint_types.includes('xai')) {
+          detectedProvider = 'xai';
         }
+      }
+      // 优先级2：使用 owned_by 字段
+      else if (model.owned_by) {
+        detectedProvider = getProviderFromOwnedBy(model.owned_by);
+      }
+      // 优先级3：基于模型名称检测
+      else {
+        detectedProvider = detectModelProvider(model.id);
+      }
 
-        // 将检测到的 provider 信息附加到模型上，供路由使用
-        enhancedModel._detectedProvider = detectedProvider;
-        // 同时更新全局路由映射表
-        globalModelRouteMap.set(model.id, detectedProvider);
+      // 将检测到的 provider 信息附加到模型上
+      enhancedModel._detectedProvider = detectedProvider;
 
-        return enhancedModel;
-      });
+      return enhancedModel;
+    });
 
-      // 使用 processMultiProviderModelList 处理模型能力
-      const processedModels = await processMultiProviderModelList(enrichedModelList, 'newapi');
+    // 使用 processMultiProviderModelList 处理模型能力
+    const processedModels = await processMultiProviderModelList(enrichedModelList, 'newapi');
 
-      // 如果我们检测到了 provider，确保它被正确应用
-      return processedModels.map((model: any) => {
-        if (model._detectedProvider) {
-          // Here you can adjust certain model properties as needed.
-          // FIXME: The current data structure does not support storing provider information, and the official NewAPI does not provide a corresponding field. Consider extending the model schema if provider tracking is required in the future.
-          delete model._detectedProvider; // Remove temporary field
-        }
-        return model;
-      });
-    },
-    // 使用动态 routers 配置，在构造时获取用户的 baseURL
-    routers: (options) => {
-      // 使用全局的模型路由映射
-      const userBaseURL = options.baseURL?.replace(/\/v1\/?$/, '') || '';
-      
-      return [
+    // 清理临时字段
+    return processedModels.map((model: any) => {
+      if (model._detectedProvider) {
+        delete model._detectedProvider;
+      }
+      return model;
+    });
+  },
+  routers: (options) => {
+    const userBaseURL = options.baseURL?.replace(/\/v\d+[a-z]*\/?$/, '') || '';
+
+    return [
       {
         apiType: 'anthropic',
-        models: () =>
-          Promise.resolve(
-            Array.from(globalModelRouteMap.entries())
-              .filter(([, provider]) => provider === 'anthropic')
-              .map(([modelId]) => modelId),
-          ),
+        models: LOBE_DEFAULT_MODEL_LIST.map((m) => m.id).filter(
+          (id) => detectModelProvider(id) === 'anthropic',
+        ),
         options: {
-          // Anthropic 在 NewAPI 中使用 /v1 路径，会自动转换为 /v1/messages
-          baseURL: urlJoin(userBaseURL, '/v1'),
+          ...options,
+          baseURL: userBaseURL,
         },
       },
       {
         apiType: 'google',
-        models: () =>
-          Promise.resolve(
-            Array.from(globalModelRouteMap.entries())
-              .filter(([, provider]) => provider === 'google')
-              .map(([modelId]) => modelId),
-          ),
+        models: LOBE_DEFAULT_MODEL_LIST.map((m) => m.id).filter(
+          (id) => detectModelProvider(id) === 'google',
+        ),
         options: {
-          // Gemini 在 NewAPI 中使用 /v1beta 路径
-          baseURL: urlJoin(userBaseURL, '/v1beta'),
+          ...options,
+          baseURL: userBaseURL,
         },
       },
       {
         apiType: 'xai',
-        models: () =>
-          Promise.resolve(
-            Array.from(globalModelRouteMap.entries())
-              .filter(([, provider]) => provider === 'xai')
-              .map(([modelId]) => modelId),
-          ),
+        models: LOBE_DEFAULT_MODEL_LIST.map((m) => m.id).filter(
+          (id) => detectModelProvider(id) === 'xai',
+        ),
         options: {
-          // xAI 使用标准 OpenAI 格式，走 /v1 路径
+          ...options,
           baseURL: urlJoin(userBaseURL, '/v1'),
         },
       },
       {
         apiType: 'openai',
         options: {
+          ...options,
           baseURL: urlJoin(userBaseURL, '/v1'),
           chatCompletion: {
             handlePayload,
           },
         },
       },
-      ];
-    },
+    ];
+  },
 });

package/packages/model-runtime/src/novita/index.test.ts

@@ -1,10 +1,9 @@
 // @vitest-environment node
+import { LobeOpenAICompatibleRuntime } from '@lobechat/model-runtime';
+import { ModelProvider } from '@lobechat/model-runtime';
 import { Mock, afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
 
-import { LobeOpenAICompatibleRuntime } from '@/libs/model-runtime';
-import { ModelProvider } from '@/libs/model-runtime';
-import { testProvider } from '@/libs/model-runtime/providerTestUtils';
-
+import { testProvider } from '../providerTestUtils';
 import models from './fixtures/models.json';
 import { LobeNovitaAI } from './index';
 

package/packages/model-runtime/src/nvidia/index.test.ts

@@ -0,0 +1,19 @@
+// @vitest-environment node
+import { ModelProvider } from '@lobechat/model-runtime';
+
+import { testProvider } from '../providerTestUtils';
+import { LobeNvidiaAI } from './index';
+
+const provider = ModelProvider.Nvidia;
+const defaultBaseURL = 'https://integrate.api.nvidia.com/v1';
+
+testProvider({
+  Runtime: LobeNvidiaAI,
+  provider,
+  defaultBaseURL,
+  chatDebugEnv: 'DEBUG_NVIDIA_CHAT_COMPLETION',
+  chatModel: 'meta/llama-3.1-8b-instruct',
+  test: {
+    skipAPICall: true,
+  },
+});

package/packages/model-runtime/src/openrouter/index.test.ts

@@ -1,9 +1,8 @@
 // @vitest-environment node
+import { LobeOpenAICompatibleRuntime } from '@lobechat/model-runtime';
 import { Mock, afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
 
-import { LobeOpenAICompatibleRuntime } from '@/libs/model-runtime';
-import { testProvider } from '@/libs/model-runtime/providerTestUtils';
-
+import { testProvider } from '../providerTestUtils';
 import models from './fixtures/models.json';
 import { LobeOpenRouterAI } from './index';
 

package/packages/model-runtime/src/perplexity/index.test.ts

@@ -1,9 +1,8 @@
 // @vitest-environment node
+import { LobeOpenAICompatibleRuntime, ModelProvider } from '@lobechat/model-runtime';
 import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
 
-import { LobeOpenAICompatibleRuntime, ModelProvider } from '@/libs/model-runtime';
-import { testProvider } from '@/libs/model-runtime/providerTestUtils';
-
+import { testProvider } from '../providerTestUtils';
 import { LobePerplexityAI } from './index';
 
 testProvider({

package/packages/model-runtime/src/ppio/index.test.ts

@@ -1,10 +1,9 @@
 // @vitest-environment node
+import { LobeOpenAICompatibleRuntime } from '@lobechat/model-runtime';
+import { ModelProvider } from '@lobechat/model-runtime';
 import { Mock, afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
 
-import { LobeOpenAICompatibleRuntime } from '@/libs/model-runtime';
-import { ModelProvider } from '@/libs/model-runtime';
-import { testProvider } from '@/libs/model-runtime/providerTestUtils';
-
+import { testProvider } from '../providerTestUtils';
 import models from './fixtures/models.json';
 import { LobePPIOAI } from './index';
 

package/packages/model-runtime/src/qwen/index.test.ts

@@ -1,7 +1,7 @@
 // @vitest-environment node
-import { ModelProvider } from '@/libs/model-runtime';
-import { testProvider } from '@/libs/model-runtime/providerTestUtils';
+import { ModelProvider } from '@lobechat/model-runtime';
 
+import { testProvider } from '../providerTestUtils';
 import { LobeQwenAI } from './index';
 
 const provider = ModelProvider.Qwen;

package/packages/model-runtime/src/sambanova/index.test.ts

@@ -0,0 +1,19 @@
+// @vitest-environment node
+import { ModelProvider } from '@lobechat/model-runtime';
+
+import { testProvider } from '../providerTestUtils';
+import { LobeSambaNovaAI } from './index';
+
+const provider = ModelProvider.SambaNova;
+const defaultBaseURL = 'https://api.sambanova.ai/v1';
+
+testProvider({
+  Runtime: LobeSambaNovaAI,
+  provider,
+  defaultBaseURL,
+  chatDebugEnv: 'DEBUG_SAMBANOVA_CHAT_COMPLETION',
+  chatModel: 'Meta-Llama-3.1-8B-Instruct',
+  test: {
+    skipAPICall: true,
+  },
+});

package/packages/model-runtime/src/search1api/index.test.ts

@@ -0,0 +1,19 @@
+// @vitest-environment node
+import { ModelProvider } from '@lobechat/model-runtime';
+
+import { testProvider } from '../providerTestUtils';
+import { LobeSearch1API } from './index';
+
+const provider = ModelProvider.Search1API;
+const defaultBaseURL = 'https://api.search1api.com/v1';
+
+testProvider({
+  Runtime: LobeSearch1API,
+  provider,
+  defaultBaseURL,
+  chatDebugEnv: 'DEBUG_SEARCH1API_CHAT_COMPLETION',
+  chatModel: 'gpt-4o-mini',
+  test: {
+    skipAPICall: true,
+  },
+});

package/packages/model-runtime/src/sensenova/index.test.ts

@@ -1,7 +1,7 @@
 // @vitest-environment node
-import { ModelProvider } from '@/libs/model-runtime';
-import { testProvider } from '@/libs/model-runtime/providerTestUtils';
+import { ModelProvider } from '@lobechat/model-runtime';
 
+import { testProvider } from '../providerTestUtils';
 import { LobeSenseNovaAI } from './index';
 
 const provider = ModelProvider.SenseNova;

package/packages/model-runtime/src/spark/index.test.ts

@@ -1,7 +1,7 @@
 // @vitest-environment node
-import { ModelProvider } from '@/libs/model-runtime';
-import { testProvider } from '@/libs/model-runtime/providerTestUtils';
+import { ModelProvider } from '@lobechat/model-runtime';
 
+import { testProvider } from '../providerTestUtils';
 import { LobeSparkAI } from './index';
 
 const provider = ModelProvider.Spark;

package/packages/model-runtime/src/stepfun/index.test.ts

@@ -1,7 +1,7 @@
 // @vitest-environment node
-import { ModelProvider } from '@/libs/model-runtime';
-import { testProvider } from '@/libs/model-runtime/providerTestUtils';
+import { ModelProvider } from '@lobechat/model-runtime';
 
+import { testProvider } from '../providerTestUtils';
 import { LobeStepfunAI } from './index';
 
 const provider = ModelProvider.Stepfun;

package/packages/model-runtime/src/taichu/index.test.ts

@@ -1,15 +1,14 @@
 // @vitest-environment node
-import OpenAI from 'openai';
-import { Mock, afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
-
 import {
   ChatStreamCallbacks,
   LobeMoonshotAI,
   LobeOpenAICompatibleRuntime,
   ModelProvider,
-} from '@/libs/model-runtime';
-import { testProvider } from '@/libs/model-runtime/providerTestUtils';
+} from '@lobechat/model-runtime';
+import OpenAI from 'openai';
+import { Mock, afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
 
+import { testProvider } from '../providerTestUtils';
 import * as debugStreamModule from '../utils/debugStream';
 import { LobeTaichuAI } from './index';
 

package/packages/model-runtime/src/tencentcloud/index.test.ts

@@ -1,5 +1,5 @@
 // @vitest-environment node
-import { ModelProvider } from '@/libs/model-runtime';
+import { ModelProvider } from '@lobechat/model-runtime';
 
 import { testProvider } from '../providerTestUtils';
 import { LobeTencentCloudAI } from './index';

package/packages/model-runtime/src/togetherai/index.test.ts

@@ -1,6 +1,5 @@
 // @vitest-environment node
-import { testProvider } from '@/libs/model-runtime/providerTestUtils';
-
+import { testProvider } from '../providerTestUtils';
 import { LobeTogetherAI } from './index';
 
 testProvider({

package/packages/model-runtime/src/upstage/index.test.ts

@@ -1,6 +1,5 @@
 // @vitest-environment node
-import { testProvider } from '@/libs/model-runtime/providerTestUtils';
-
+import { testProvider } from '../providerTestUtils';
 import { LobeUpstageAI } from './index';
 
 testProvider({

package/packages/model-runtime/src/utils/openaiCompatibleFactory/index.test.ts

@@ -1,22 +1,24 @@
 // @vitest-environment node
-import OpenAI from 'openai';
-import type { Stream } from 'openai/streaming';
-import { Mock, afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
-
 import {
   AgentRuntimeErrorType,
   ChatStreamCallbacks,
   ChatStreamPayload,
   LobeOpenAICompatibleRuntime,
   ModelProvider,
-} from '@/libs/model-runtime';
-import officalOpenAIModels from '@/libs/model-runtime/openai/fixtures/openai-models.json';
-import { sleep } from '@/utils/sleep';
+} from '@lobechat/model-runtime';
+import OpenAI from 'openai';
+import type { Stream } from 'openai/streaming';
+import { Mock, afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
 
 import * as debugStreamModule from '../debugStream';
 import * as openaiHelpers from '../openaiHelpers';
 import { createOpenAICompatibleRuntime } from './index';
 
+const sleep = async (ms: number) =>
+  await new Promise((resolve) => {
+    setTimeout(resolve, ms);
+  });
+
 const provider = 'groq';
 const defaultBaseURL = 'https://api.groq.com/openai/v1';
 const bizErrorType = 'ProviderBizError';

package/packages/model-runtime/src/utils/streams/anthropic.ts

@@ -1,7 +1,7 @@
 import Anthropic from '@anthropic-ai/sdk';
 import type { Stream } from '@anthropic-ai/sdk/streaming';
 
-import { CitationItem, ModelTokensUsage } from '@/types/message';
+import { ChatCitationItem, ModelTokensUsage } from '@/types/message';
 
 import { ChatStreamCallbacks } from '../../types';
 import {
@@ -180,7 +180,7 @@ export const transformAnthropicStream = (
             context.returnedCitationArray.push({
               title: citations.title,
               url: citations.url,
-            } as CitationItem);
+            } as ChatCitationItem);
           }
 
           return { data: null, id: context.id, type: 'text' };