@lobehub/chat 1.120.7 → 1.121.1

package/docker-compose.development.yml

@@ -0,0 +1,40 @@
+name: lobe-chat-development
+services:
+  network-service:
+    image: alpine
+    container_name: lobe-network
+    restart: always
+    ports:
+      - '${MINIO_PORT}:${MINIO_PORT}' # MinIO API
+      - '9001:9001' # MinIO Console
+      - '${CASDOOR_PORT}:${CASDOOR_PORT}' # Casdoor
+    command: tail -f /dev/null
+    networks:
+      - lobe-network
+
+  postgresql:
+    extends:
+      file: docker-compose/local/docker-compose.yml
+      service: postgresql
+  minio:
+    extends:
+      file: docker-compose/local/docker-compose.yml
+      service: minio
+  casdoor:
+    extends:
+      file: docker-compose/local/docker-compose.yml
+      service: casdoor
+  searxng:
+    extends:
+      file: docker-compose/local/docker-compose.yml
+      service: searxng
+
+volumes:
+  data:
+    driver: local
+  s3_data:
+    driver: local
+
+networks:
+  lobe-network:
+    driver: bridge

package/docs/development/basic/work-with-server-side-database.mdx

@@ -0,0 +1,77 @@
+# Work with Server-Side Database
+
+LobeChat provides a battery-included experience with its client-side database.
+While some features you really care about is only available at a server-side development.
+
+In order to work with the aspect of server-side database,
+you can setup all the prerequisites by following the [Deploying Server-Side Database](https://lobehub.com/docs/self-hosting/server-database) story.
+But here is the easier approach that can reduce your pain.
+
+## Quick Setup
+
+### Environment Configuration
+
+The project already includes a `.env.development` file with all necessary environment variables for server-side database mode. This file configures:
+
+- **Service Mode**: `NEXT_PUBLIC_SERVICE_MODE=server`
+- **Database**: PostgreSQL with connection string
+- **Authentication**: NextAuth with Casdoor SSO
+- **Storage**: MinIO S3-compatible storage
+- **Search**: SearXNG search engine
+
+### Start Docker Services
+
+Start all required services using Docker Compose:
+
+```bash
+docker-compose -f docker-compose.development.yml up -d
+```
+
+This will start the following services:
+
+- PostgreSQL database (port 5432)
+- MinIO storage (port 9000)
+- Casdoor authentication (port 8000)
+- SearXNG search (port 8080)
+
+### Run Database Migrations
+
+Execute the database migration script to create all necessary tables:
+
+```bash
+pnpm db:migrate
+```
+
+You should see: `✅ database migration pass.`
+
+### Start Development Server
+
+Launch the LobeChat development server:
+
+```bash
+pnpm dev
+```
+
+The server will start on `http://localhost:3010`
+
+And you can check all Docker services are running by running:
+
+```bash
+docker-compose -f docker-compose.development.yml ps
+```
+
+### Reset Services
+
+If you encounter issues, you can reset the entire stack:
+
+```bash
+# Stop and remove all containers
+docker-compose -f docker-compose.development.yml down
+
+# Remove volumes (this will delete all data)
+docker-compose -f docker-compose.development.yml down -v
+
+# Start fresh
+docker-compose -f docker-compose.development.yml up -d
+pnpm db:migrate
+```

package/docs/development/basic/work-with-server-side-database.zh-CN.mdx

@@ -0,0 +1,77 @@
+# 使用服务端数据库
+
+LobeChat 提供了内置的客户端数据库体验。
+但某些重要功能仅在服务端开发中可用。
+
+为了使用服务端数据库功能，
+需要参考 [部署服务端数据库](https://lobehub.com/docs/self-hosting/server-database) 的说明来配置所有前置条件。
+本文档提供了一个更简化的配置方法，能够在本地开发时快速启动简化的服务端环境。
+
+## 快速设置
+
+### 环境配置
+
+项目已经包含了一个 `.env.development` 文件，其中包含服务端数据库模式所需的所有环境变量。此文件配置了：
+
+- **服务模式**: `NEXT_PUBLIC_SERVICE_MODE=server`
+- **数据库**: 带连接字符串的 PostgreSQL
+- **身份验证**: 带 Casdoor SSO 的 NextAuth
+- **存储**: MinIO S3 兼容存储
+- **搜索**: SearXNG 搜索引擎
+
+### 启动 Docker 服务
+
+使用 Docker Compose 启动所有必需的服务：
+
+```bash
+docker-compose -f docker-compose.development.yml up -d
+```
+
+这将启动以下服务：
+
+- PostgreSQL 数据库（端口 5432）
+- MinIO 存储（端口 9000）
+- Casdoor 身份验证（端口 8000）
+- SearXNG 搜索（端口 8080）
+
+### 运行数据库迁移
+
+执行数据库迁移脚本以创建所有必要的表：
+
+```bash
+pnpm db:migrate
+```
+
+预期输出：`✅ database migration pass.`
+
+### 启动开发服务器
+
+启动 LobeChat 开发服务器：
+
+```bash
+pnpm dev
+```
+
+服务器将在 `http://localhost:3010` 上启动
+
+可以通过运行以下命令检查所有 Docker 服务运行状态：
+
+```bash
+docker-compose -f docker-compose.development.yml ps
+```
+
+### 重置服务
+
+如遇到问题，可以重置整个服务堆栈：
+
+```bash
+# 停止并删除所有容器
+docker-compose -f docker-compose.development.yml down
+
+# 删除卷（这将删除所有数据）
+docker-compose -f docker-compose.development.yml down -v
+
+# 重新启动
+docker-compose -f docker-compose.development.yml up -d
+pnpm db:migrate
+```

package/docs/self-hosting/advanced/s3/cloudflare-r2.mdx

@@ -85,7 +85,7 @@ We need to configure an S3 storage service in the server-side database to store
   <Image alt={'Configure allowed site domain'} src={'https://github.com/lobehub/lobe-chat/assets/28616219/dfcc2cb3-2958-4498-a8a4-51bec584fe7d'} />
 
   <Callout type={'info'}>
-    If you also plan to use the desktop client, add <code>http://localhost:3015</code> to <code>AllowedOrigins</code> so the desktop client (running locally) can access R2.
+    If you also plan to use the desktop client, add <code>[http://localhost:3015](http://localhost:3015)</code> to <code>AllowedOrigins</code> so the desktop client (running locally) can access R2.
   </Callout>
 
   Example configuration is as follows:

package/docs/self-hosting/advanced/s3/cloudflare-r2.zh-CN.mdx

@@ -82,11 +82,11 @@ tags:
   添加跨域规则，允许你的域名（在上文是 `https://your-project.vercel.app`）来源的请求：
 
   <Image alt={'配置允许你的站点域名'} src={'https://github.com/lobehub/lobe-chat/assets/28616219/dfcc2cb3-2958-4498-a8a4-51bec584fe7d'} />
+
   <Callout type={'info'}>
-    如果你还需要在桌面端使用，请在 <code>AllowedOrigins</code> 中额外添加 <code>http://localhost:3015</code>，以便桌面端（本地运行）能够访问 R2。
+    如果你还需要在桌面端使用，请在 <code>AllowedOrigins</code> 中额外添加 <code>[http://localhost:3015](http://localhost:3015)</code>，以便桌面端（本地运行）能够访问 R2。
   </Callout>
 
-
   示例配置如下：
 
   ```json

package/locales/zh-CN/common.json

@@ -182,6 +182,13 @@
     "title": "喜欢我们的产品？"
   },
   "fullscreen": "全屏模式",
+  "geminiImageChineseWarning": {
+    "content": "Nano Banana 使用中文有概率性生成图片失败。建议使用英文以获得更好的效果。",
+    "continueGenerate": "继续生成",
+    "continueSend": "继续发送",
+    "doNotShowAgain": "不再提示",
+    "title": "中文输入提示"
+  },
   "historyRange": "历史范围",
   "import": "导入",
   "importData": "导入数据",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lobehub/chat",
-  "version": "1.120.7",
+  "version": "1.121.1",
   "description": "Lobe Chat - an open-source, high-performance chatbot framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.",
   "keywords": [
     "framework",
@@ -327,6 +327,7 @@
     "crypto-js": "^4.2.0",
     "dbdocs": "^0.16.0",
     "dotenv": "^17.0.0",
+    "dotenv-expand": "^12.0.2",
     "dpdm-fast": "^1.0.13",
     "drizzle-dbml-generator": "^0.10.0",
     "drizzle-kit": "^0.31.4",

package/packages/database/src/repositories/aiInfra/index.ts

@@ -208,7 +208,9 @@ export class AiInfraRepos {
       const providerModels = modules[providerId];
 
       // use the serverModelLists as the defined server model list
-      const presetList = this.providerConfigs[providerId]?.serverModelLists || providerModels;
+      // fallback to empty array for custom provider
+      const presetList = this.providerConfigs[providerId]?.serverModelLists || providerModels || [];
+
       return (presetList as AIChatModelCard[]).map<AiProviderModelListItem>((m) => ({
         ...m,
         enabled: m.enabled || false,

package/packages/model-runtime/src/RouterRuntime/createRuntime.test.ts

@@ -96,7 +96,7 @@ describe('createRouterRuntime', () => {
       });
 
       const runtime = new Runtime();
-      const models = await runtime['getModels']({
+      const models = await runtime['getRouterMatchModels']({
         id: 'test',
         models: ['model-1', 'model-2'],
         runtime: mockRuntime,
@@ -105,7 +105,7 @@ describe('createRouterRuntime', () => {
       expect(models).toEqual(['model-1', 'model-2']);
     });
 
-    it('should call and cache asynchronous models function', async () => {
+    it('should call asynchronous models function', async () => {
       const mockRuntime = {
         chat: vi.fn(),
       } as unknown as LobeRuntimeAI;
@@ -131,14 +131,9 @@ describe('createRouterRuntime', () => {
         runtime: mockRuntime,
       };
 
-      // First call
-      const models1 = await runtime['getModels'](runtimeItem);
-      expect(models1).toEqual(['async-model-1', 'async-model-2']);
-      expect(mockModelsFunction).toHaveBeenCalledTimes(1);
-
-      // Second call should use cache
-      const models2 = await runtime['getModels'](runtimeItem);
-      expect(models2).toEqual(['async-model-1', 'async-model-2']);
+      // Call the function
+      const models = await runtime['getRouterMatchModels'](runtimeItem);
+      expect(models).toEqual(['async-model-1', 'async-model-2']);
       expect(mockModelsFunction).toHaveBeenCalledTimes(1);
     });
 
@@ -159,7 +154,7 @@ describe('createRouterRuntime', () => {
       });
 
       const runtime = new Runtime();
-      const models = await runtime['getModels']({
+      const models = await runtime['getRouterMatchModels']({
         id: 'test',
         runtime: mockRuntime,
       });
@@ -455,84 +450,4 @@ describe('createRouterRuntime', () => {
       expect(mockTextToSpeech).toHaveBeenCalledWith(payload, options);
     });
   });
-
-  describe('clearModelCache method', () => {
-    it('should clear specific runtime cache when runtimeId provided', async () => {
-      const mockModelsFunction = vi.fn().mockResolvedValue(['model-1']);
-
-      const Runtime = createRouterRuntime({
-        id: 'test-runtime',
-        routers: [
-          {
-            apiType: 'openai',
-            options: {},
-            runtime: vi.fn() as any,
-            models: mockModelsFunction,
-          },
-        ],
-      });
-
-      const runtime = new Runtime();
-      const runtimeItem = {
-        id: 'test-id',
-        models: mockModelsFunction,
-        runtime: {} as any,
-      };
-
-      // Build cache
-      await runtime['getModels'](runtimeItem);
-      expect(mockModelsFunction).toHaveBeenCalledTimes(1);
-
-      // Clear specific cache
-      runtime.clearModelCache('test-id');
-
-      // Should call function again
-      await runtime['getModels'](runtimeItem);
-      expect(mockModelsFunction).toHaveBeenCalledTimes(2);
-    });
-
-    it('should clear all cache when no runtimeId provided', async () => {
-      const mockModelsFunction1 = vi.fn().mockResolvedValue(['model-1']);
-      const mockModelsFunction2 = vi.fn().mockResolvedValue(['model-2']);
-
-      const Runtime = createRouterRuntime({
-        id: 'test-runtime',
-        routers: [
-          {
-            apiType: 'openai',
-            options: {},
-            runtime: vi.fn() as any,
-            models: mockModelsFunction1,
-          },
-        ],
-      });
-
-      const runtime = new Runtime();
-      const runtimeItem1 = {
-        id: 'test-id-1',
-        models: mockModelsFunction1,
-        runtime: {} as any,
-      };
-      const runtimeItem2 = {
-        id: 'test-id-2',
-        models: mockModelsFunction2,
-        runtime: {} as any,
-      };
-
-      // Build cache for both items
-      await runtime['getModels'](runtimeItem1);
-      await runtime['getModels'](runtimeItem2);
-      expect(mockModelsFunction1).toHaveBeenCalledTimes(1);
-      expect(mockModelsFunction2).toHaveBeenCalledTimes(1);
-
-      // Clear all cache
-      runtime.clearModelCache();
-
-      // Should call functions again
-      await runtime['getModels'](runtimeItem1);
-      await runtime['getModels'](runtimeItem2);
-      expect(mockModelsFunction1).toHaveBeenCalledTimes(2);
-      expect(mockModelsFunction2).toHaveBeenCalledTimes(2);
-    });
-  });
 });

package/packages/model-runtime/src/RouterRuntime/createRuntime.ts

@@ -117,7 +117,6 @@ export const createRouterRuntime = ({
   return class UniformRuntime implements LobeRuntimeAI {
     private _runtimes: RuntimeItem[];
     private _options: ClientOptions & Record<string, any>;
-    private _modelCache = new Map<string, string[]>();
 
     constructor(options: ClientOptions & Record<string, any> = {}) {
       const _options = {
@@ -143,30 +142,21 @@ export const createRouterRuntime = ({
       this._options = _options;
     }
 
-    // Get runtime's models list, supporting both synchronous arrays and asynchronous functions with caching
-    private async getModels(runtimeItem: RuntimeItem): Promise<string[]> {
-      const cacheKey = runtimeItem.id;
-
-      // If it's a synchronous array, return directly without caching
+    // Get runtime's models list, supporting both synchronous arrays and asynchronous functions
+    private async getRouterMatchModels(runtimeItem: RuntimeItem): Promise<string[]> {
+      // If it's a synchronous array, return directly
       if (typeof runtimeItem.models !== 'function') {
         return runtimeItem.models || [];
       }
 
-      // Check cache
-      if (this._modelCache.has(cacheKey)) {
-        return this._modelCache.get(cacheKey)!;
-      }
-
-      // Get model list and cache result
-      const models = await runtimeItem.models();
-      this._modelCache.set(cacheKey, models);
-      return models;
+      // Get model list
+      return await runtimeItem.models();
     }
 
     // Check if it can match a specific model, otherwise default to using the last runtime
     async getRuntimeByModel(model: string) {
       for (const runtimeItem of this._runtimes) {
-        const models = await this.getModels(runtimeItem);
+        const models = await this.getRouterMatchModels(runtimeItem);
         if (models.includes(model)) {
           return runtimeItem.runtime;
         }
@@ -226,17 +216,5 @@ export const createRouterRuntime = ({
 
       return runtime.textToSpeech!(payload, options);
     }
-
-    /**
-     * Clear model list cache, forcing reload on next access
-     * @param runtimeId - Optional, specify to clear cache for a specific runtime, omit to clear all caches
-     */
-    clearModelCache(runtimeId?: string) {
-      if (runtimeId) {
-        this._modelCache.delete(runtimeId);
-      } else {
-        this._modelCache.clear();
-      }
-    }
   };
 };

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

@@ -71,29 +71,32 @@ export const LobeOpenRouterAI = createOpenAICompatibleRuntime({
 
     // 处理前端获取的模型信息，转换为标准格式
     const formattedModels = modelList.map((model) => {
+      const { endpoint } = model;
+      const endpointModel = endpoint?.model;
+
       const displayName = model.slug?.toLowerCase().includes('deepseek')
         ? (model.name ?? model.slug)
         : (model.short_name ?? model.name ?? model.slug);
 
+      const inputModalities = endpointModel?.input_modalities || model.input_modalities;
+
       return {
-        contextWindowTokens: model.context_length,
-        description: model.description,
+        contextWindowTokens: endpoint?.context_length || model.context_length,
+        description: endpointModel?.description || model.description,
         displayName,
-        functionCall: model.endpoint?.supports_tool_parameters || false,
-        id: model.slug,
+        functionCall: endpoint?.supports_tool_parameters || false,
+        id: endpoint?.model_variant_slug || model.slug,
         maxOutput:
-          typeof model.endpoint?.max_completion_tokens === 'number'
-            ? model.endpoint.max_completion_tokens
+          typeof endpoint?.max_completion_tokens === 'number'
+            ? endpoint.max_completion_tokens
             : undefined,
         pricing: {
-          input: formatPrice(model.endpoint?.pricing?.prompt),
-          output: formatPrice(model.endpoint?.pricing?.completion),
+          input: formatPrice(endpoint?.pricing?.prompt),
+          output: formatPrice(endpoint?.pricing?.completion),
         },
-        reasoning: model.endpoint?.supports_reasoning || false,
+        reasoning: endpoint?.supports_reasoning || false,
         releasedAt: new Date(model.created_at).toISOString().split('T')[0],
-        vision:
-          (Array.isArray(model.input_modalities) && model.input_modalities.includes('image')) ||
-          false,
+        vision: Array.isArray(inputModalities) && inputModalities.includes('image'),
       };
     });
 

package/packages/model-runtime/src/openrouter/type.ts

@@ -19,11 +19,21 @@ export interface OpenRouterModelCard {
 }
 
 interface OpenRouterModelEndpoint {
+  context_length?: number;
   max_completion_tokens: number | null;
+  model?: {
+    description?: string;
+    input_modalities?: string[];
+    name?: string;
+    short_name?: string;
+    slug: string;
+  };
+  model_variant_slug?: string;
   pricing: ModelPricing;
   supported_parameters: string[];
   supports_reasoning?: boolean;
   supports_tool_parameters?: boolean;
+  variant?: 'free' | 'standard' | 'unknown';
 }
 
 interface OpenRouterOpenAIReasoning {

package/packages/model-runtime/src/utils/modelParse.test.ts

@@ -758,4 +758,70 @@ describe('modelParse', () => {
       expect(modelConfigKeys.sort()).toEqual(providerDetectionKeys.sort());
     });
   });
+
+  describe('displayName processing', () => {
+    it('should replace "Gemini 2.5 Flash Image Preview" with "Nano Banana"', async () => {
+      const modelList = [
+        {
+          id: 'gemini-2.5-flash-image-preview',
+          displayName: 'Gemini 2.5 Flash Image Preview',
+        },
+        {
+          id: 'some-other-model',
+          displayName: 'Some Other Model',
+        },
+        {
+          id: 'partial-gemini-model',
+          displayName: 'Custom Gemini 2.5 Flash Image Preview Enhanced',
+        },
+        {
+          id: 'gemini-free-model',
+          displayName: 'Gemini 2.5 Flash Image Preview (free)',
+        },
+      ];
+
+      const result = await processModelList(modelList, MODEL_LIST_CONFIGS.google);
+
+      expect(result).toHaveLength(4);
+
+      // First model should have "Nano Banana" as displayName
+      const geminiModel = result.find((m) => m.id === 'gemini-2.5-flash-image-preview');
+      expect(geminiModel?.displayName).toBe('Nano Banana');
+
+      // Second model should keep original displayName
+      const otherModel = result.find((m) => m.id === 'some-other-model');
+      expect(otherModel?.displayName).toBe('Some Other Model');
+
+      // Third model (partial match) should replace only the matching part
+      const partialModel = result.find((m) => m.id === 'partial-gemini-model');
+      expect(partialModel?.displayName).toBe('Custom Nano Banana Enhanced');
+
+      // Fourth model should preserve the (free) suffix
+      const freeModel = result.find((m) => m.id === 'gemini-free-model');
+      expect(freeModel?.displayName).toBe('Nano Banana (free)');
+    });
+
+    it('should keep original displayName when not matching Gemini 2.5 Flash Image Preview', async () => {
+      const modelList = [
+        {
+          id: 'gpt-4',
+          displayName: 'GPT-4',
+        },
+        {
+          id: 'gemini-pro',
+          displayName: 'Gemini Pro',
+        },
+      ];
+
+      const result = await processModelList(modelList, MODEL_LIST_CONFIGS.google);
+
+      expect(result).toHaveLength(2);
+
+      const gptModel = result.find((m) => m.id === 'gpt-4');
+      expect(gptModel?.displayName).toBe('GPT-4');
+
+      const geminiProModel = result.find((m) => m.id === 'gemini-pro');
+      expect(geminiProModel?.displayName).toBe('Gemini Pro');
+    });
+  });
 });

package/packages/model-runtime/src/utils/modelParse.ts

@@ -264,6 +264,20 @@ const processReleasedAt = (model: any, knownModel?: any): string | undefined =>
   return model.releasedAt ?? knownModel?.releasedAt ?? undefined;
 };
 
+/**
+ * 处理模型显示名称
+ * @param displayName 原始显示名称
+ * @returns 处理后的显示名称
+ */
+const processDisplayName = (displayName: string): string => {
+  // 如果包含 "Gemini 2.5 Flash Image Preview"，替换对应部分为 "Nano Banana"
+  if (displayName.includes('Gemini 2.5 Flash Image Preview')) {
+    return displayName.replace('Gemini 2.5 Flash Image Preview', 'Nano Banana');
+  }
+
+  return displayName;
+};
+
 /**
  * 处理模型卡片的通用逻辑
  */
@@ -331,9 +345,7 @@ const processModelCard = (
   return {
     contextWindowTokens: model.contextWindowTokens ?? knownModel?.contextWindowTokens ?? undefined,
     description: model.description ?? knownModel?.description ?? '',
-    displayName: (model.displayName ?? knownModel?.displayName ?? model.id)
-      .replaceAll(/\s*[(（][^)）]*[)）]\s*/g, '')
-      .trim(), // 去除括号内容
+    displayName: processDisplayName(model.displayName ?? knownModel?.displayName ?? model.id),
     enabled: model?.enabled || false,
     functionCall:
       model.functionCall ??

package/packages/model-runtime/src/utils/postProcessModelList.ts

@@ -5,6 +5,7 @@ import type { ChatModelCard } from '@/types/llm';
 // Whitelist for automatic image model generation
 export const IMAGE_GENERATION_MODEL_WHITELIST = [
   'gemini-2.5-flash-image-preview',
+  'gemini-2.5-flash-image-preview:free',
   // More models can be added in the future
 ] as const;
 

package/packages/utils/src/detectChinese.test.ts

@@ -0,0 +1,37 @@
+import { describe, expect, it } from 'vitest';
+
+import { containsChinese } from './detectChinese';
+
+describe('containsChinese', () => {
+  it('should return true for text containing Chinese characters', () => {
+    expect(containsChinese('你好世界')).toBe(true);
+    expect(containsChinese('Hello 世界')).toBe(true);
+    expect(containsChinese('测试 test')).toBe(true);
+    expect(containsChinese('这是一个测试')).toBe(true);
+  });
+
+  it('should return false for text without Chinese characters', () => {
+    expect(containsChinese('Hello World')).toBe(false);
+    expect(containsChinese('123456')).toBe(false);
+    expect(containsChinese('!@#$%^&*()')).toBe(false);
+    expect(containsChinese('')).toBe(false);
+    expect(containsChinese('English only text')).toBe(false);
+  });
+
+  it('should handle mixed content correctly', () => {
+    expect(containsChinese('Hello 中国')).toBe(true);
+    expect(containsChinese('English and 数字 123')).toBe(true);
+    expect(containsChinese('Japanese こんにちは and English')).toBe(false);
+    expect(containsChinese('Korean 안녕하세요 and English')).toBe(false);
+  });
+
+  it('should detect extended Chinese character ranges', () => {
+    // Test CJK Unified Ideographs Extension A (U+3400-U+4DBF)
+    expect(containsChinese('㐀㑇㒯')).toBe(true);
+    // Test CJK Compatibility Ideographs (U+F900-U+FAFF)
+    expect(containsChinese('豈更車')).toBe(true);
+    // Test traditional Chinese characters
+    expect(containsChinese('繁體中文')).toBe(true);
+    expect(containsChinese('學習語言')).toBe(true);
+  });
+});