@lobehub/chat 1.142.9 → 1.143.0

package/.env.example

@@ -13,6 +13,17 @@
 # Default is '0' (enabled)
 # ENABLED_CSP=1
 
+# SSRF Protection Settings
+# Set to '1' to allow connections to private IP addresses (disable SSRF protection)
+# WARNING: Only enable this in trusted environments
+# Default is '0' (SSRF protection enabled)
+# SSRF_ALLOW_PRIVATE_IP_ADDRESS=0
+
+# Whitelist of allowed private IP addresses (comma-separated)
+# Only takes effect when SSRF_ALLOW_PRIVATE_IP_ADDRESS is '0'
+# Example: Allow specific internal servers while keeping SSRF protection
+# SSRF_ALLOW_IP_ADDRESS_LIST=192.168.1.100,10.0.0.50
+
 ########################################
 ########## AI Provider Service #########
 ########################################

package/CHANGELOG.md

@@ -2,6 +2,31 @@
 
 # Changelog
 
+## [Version 1.143.0](https://github.com/lobehub/lobe-chat/compare/v1.142.9...v1.143.0)
+
+<sup>Released on **2025-12-01**</sup>
+
+#### ✨ Features
+
+- **misc**: Support DeepSeek Interleaved thinking.
+
+<br/>
+
+<details>
+<summary><kbd>Improvements and Fixes</kbd></summary>
+
+#### What's improved
+
+- **misc**: Support DeepSeek Interleaved thinking, closes [#10478](https://github.com/lobehub/lobe-chat/issues/10478) [#10219](https://github.com/lobehub/lobe-chat/issues/10219) [#10152](https://github.com/lobehub/lobe-chat/issues/10152) ([aee5d71](https://github.com/lobehub/lobe-chat/commit/aee5d71))
+
+</details>
+
+<div align="right">
+
+[![](https://img.shields.io/badge/-BACK_TO_TOP-151515?style=flat-square)](#readme-top)
+
+</div>
+
 ### [Version 1.142.9](https://github.com/lobehub/lobe-chat/compare/v1.142.8...v1.142.9)
 
 <sup>Released on **2025-11-02**</sup>

package/changelog/v1.json

@@ -1,4 +1,13 @@
 [
+  {
+    "children": {
+      "features": [
+        "Support DeepSeek Interleaved thinking."
+      ]
+    },
+    "date": "2025-12-01",
+    "version": "1.143.0"
+  },
   {
     "children": {
       "fixes": [

package/docs/self-hosting/environment-variables/basic.mdx

@@ -127,16 +127,62 @@ For specific content, please refer to the [Feature Flags](/docs/self-hosting/adv
 ### `SSRF_ALLOW_PRIVATE_IP_ADDRESS`
 
 - Type: Optional
-- Description: Allow to connect private IP address. In a trusted environment, it can be set to true to turn off SSRF protection.
+- Description: Controls whether to allow connections to private IP addresses. Set to `1` to disable SSRF protection and allow all private IP addresses. In a trusted environment (e.g., internal network), this can be enabled to allow access to internal resources.
 - Default: `0`
 - Example: `1` or `0`
 
+<Callout type="warning">
+  **Security Notice**: Enabling this option will disable SSRF protection and allow connections to private
+  IP addresses (127.0.0.0/8, 10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16, etc.). Only enable this in
+  trusted environments where you need to access internal network resources.
+</Callout>
+
+**Use Cases**:
+
+LobeChat performs SSRF security checks in the following scenarios:
+
+1. **Image/Video URL to Base64 Conversion**: When processing media messages (e.g., vision models, multimodal models), LobeChat converts image and video URLs to base64 format. This check prevents malicious users from accessing internal network resources.
+
+   Examples:
+
+   - Image: A user sends an image message with URL `http://192.168.1.100/admin/secrets.png`
+   - Video: A user sends a video message with URL `http://10.0.0.50/internal/meeting.mp4`
+
+   Without SSRF protection, these requests could expose internal network resources.
+
+2. **Web Crawler**: When using web crawling features to fetch external content.
+
+3. **Proxy Requests**: When proxying external API requests.
+
+**Configuration Examples**:
+
+```bash
+# Scenario 1: Public deployment (recommended)
+# Block all private IP addresses for security
+SSRF_ALLOW_PRIVATE_IP_ADDRESS=0
+
+# Scenario 2: Internal deployment
+# Allow all private IP addresses to access internal image servers
+SSRF_ALLOW_PRIVATE_IP_ADDRESS=1
+
+# Scenario 3: Hybrid deployment (most common)
+# Block private IPs by default, but allow specific trusted internal servers
+SSRF_ALLOW_PRIVATE_IP_ADDRESS=0
+SSRF_ALLOW_IP_ADDRESS_LIST=192.168.1.100,10.0.0.50
+```
+
 ### `SSRF_ALLOW_IP_ADDRESS_LIST`
 
 - Type: Optional
-- Description: Allow private IP address list, multiple IP addresses are separated by commas. Only when `SSRF_ALLOW_PRIVATE_IP_ADDRESS` is `0`, it takes effect.
+- Description: Whitelist of allowed IP addresses, separated by commas. Only takes effect when `SSRF_ALLOW_PRIVATE_IP_ADDRESS` is `0`. Use this to allow specific internal IP addresses while keeping SSRF protection enabled for other private IPs.
 - Default: -
-- Example: `198.18.1.62,224.0.0.3`
+- Example: `192.168.1.100,10.0.0.50,172.16.0.10`
+
+**Common Use Cases**:
+
+- Allow access to internal image storage server: `192.168.1.100`
+- Allow access to internal API gateway: `10.0.0.50`
+- Allow access to internal documentation server: `172.16.0.10`
 
 ### `ENABLE_AUTH_PROTECTION`
 

package/docs/self-hosting/environment-variables/basic.zh-CN.mdx

@@ -123,16 +123,61 @@ LobeChat 在部署时提供了一些额外的配置项，你可以使用环境
 ### `SSRF_ALLOW_PRIVATE_IP_ADDRESS`
 
 - 类型：可选
-- 描述：是否允许连接私有 IP 地址。在可信环境中可以设置为 true 来关闭 SSRF 防护。
+- 描述：控制是否允许连接私有 IP 地址。设置为 `1` 时将关闭 SSRF 防护并允许所有私有 IP 地址。在可信环境（如内网部署）中，可以启用此选项以访问内部资源。
 - 默认值：`0`
-- 示例：`1` or `0`
+- 示例：`1` 或 `0`
+
+<Callout type="warning">
+  **安全提示**：启用此选项将关闭 SSRF 防护，允许连接私有 IP 地址段（127.0.0.0/8、10.0.0.0/8、172.16.0.0/12、192.168.0.0/16
+  等）。仅在需要访问内网资源的可信环境中启用。
+</Callout>
+
+**应用场景**：
+
+LobeChat 会在以下场景执行 SSRF 安全检查：
+
+1. **图片 / 视频 URL 转 Base64**：在处理媒体消息时（例如视觉模型、多模态模型），LobeChat 会将图片和视频 URL 转换为 base64 格式。此检查可防止恶意用户通过媒体 URL 访问内网资源。
+
+   举例：
+
+   - 图片：用户发送图片消息，URL 为 `http://192.168.1.100/admin/secrets.png`
+   - 视频：用户发送视频消息，URL 为 `http://10.0.0.50/internal/meeting.mp4`
+
+   若无 SSRF 防护，这些请求可能导致内网资源泄露。
+
+2. **网页爬取**：使用网页爬取功能获取外部内容时。
+
+3. **代理请求**：代理外部 API 请求时。
+
+**配置示例**：
+
+```bash
+# 场景 1：公网部署（推荐）
+# 阻止所有私有 IP 访问，保证安全
+SSRF_ALLOW_PRIVATE_IP_ADDRESS=0
+
+# 场景 2：内网部署
+# 允许所有私有 IP，可访问内网图片服务器等资源
+SSRF_ALLOW_PRIVATE_IP_ADDRESS=1
+
+# 场景 3：混合部署（最常见）
+# 默认阻止私有 IP，但允许特定可信的内网服务器
+SSRF_ALLOW_PRIVATE_IP_ADDRESS=0
+SSRF_ALLOW_IP_ADDRESS_LIST=192.168.1.100,10.0.0.50
+```
 
 ### `SSRF_ALLOW_IP_ADDRESS_LIST`
 
 - 类型：可选
-- 说明：允许的私有 IP 地址列表，多个 IP 地址用逗号分隔。仅在 `SSRF_ALLOW_PRIVATE_IP_ADDRESS` 为 `0` 时生效。
+- 描述：允许访问的 IP 地址白名单，多个 IP 地址用逗号分隔。仅在 `SSRF_ALLOW_PRIVATE_IP_ADDRESS` 为 `0` 时生效。使用此选项可以在保持 SSRF 防护的同时，允许访问特定的内网 IP 地址。
 - 默认值：-
-- 示例：`198.18.1.62,224.0.0.3`
+- 示例：`192.168.1.100,10.0.0.50,172.16.0.10`
+
+**常见使用场景**：
+
+- 允许访问内网图片存储服务器：`192.168.1.100`
+- 允许访问内网 API 网关：`10.0.0.50`
+- 允许访问内网文档服务器：`172.16.0.10`
 
 ### `ENABLE_AUTH_PROTECTION`
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lobehub/chat",
-  "version": "1.142.9",
+  "version": "1.143.0",
   "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",
@@ -156,6 +156,7 @@
     "@lobechat/database": "workspace:*",
     "@lobechat/electron-client-ipc": "workspace:*",
     "@lobechat/electron-server-ipc": "workspace:*",
+    "@lobechat/fetch-sse": "workspace:*",
     "@lobechat/file-loaders": "workspace:*",
     "@lobechat/model-runtime": "workspace:*",
     "@lobechat/observability-otel": "workspace:*",

package/packages/fetch-sse/package.json

@@ -0,0 +1,29 @@
+{
+  "name": "@lobechat/fetch-sse",
+  "version": "1.0.0",
+  "private": true,
+  "description": "SSE fetch utilities with streaming support",
+  "exports": {
+    ".": {
+      "types": "./src/index.ts",
+      "default": "./src/index.ts"
+    },
+    "./parseError": {
+      "types": "./src/parseError.ts",
+      "default": "./src/parseError.ts"
+    }
+  },
+  "main": "./src/index.ts",
+  "types": "./src/index.ts",
+  "scripts": {
+    "test": "vitest",
+    "test:coverage": "vitest --coverage --silent='passed-only'"
+  },
+  "dependencies": {
+    "@lobechat/const": "workspace:*",
+    "@lobechat/model-runtime": "workspace:*",
+    "@lobechat/types": "workspace:*",
+    "@lobechat/utils": "workspace:*",
+    "i18next": "^24.2.1"
+  }
+}

package/packages/{utils/src/fetch → fetch-sse/src}/__tests__/fetchSSE.test.ts

@@ -1,10 +1,10 @@
 import { MESSAGE_CANCEL_FLAT } from '@lobechat/const';
 import { ChatMessageError } from '@lobechat/types';
+import { FetchEventSourceInit } from '@lobechat/utils/client/fetchEventSource/index';
+import { fetchEventSource } from '@lobechat/utils/client/fetchEventSource/index';
+import { sleep } from '@lobechat/utils/sleep';
 import { afterEach, describe, expect, it, vi } from 'vitest';
 
-import { FetchEventSourceInit } from '../../client/fetchEventSource';
-import { fetchEventSource } from '../../client/fetchEventSource';
-import { sleep } from '../../sleep';
 import { fetchSSE } from '../fetchSSE';
 
 // 模拟 i18next
@@ -12,7 +12,7 @@ vi.mock('i18next', () => ({
   t: vi.fn((key) => `translated_${key}`),
 }));
 
-vi.mock('../../client/fetchEventSource', () => ({
+vi.mock('@lobechat/utils/client/fetchEventSource/index', () => ({
   fetchEventSource: vi.fn(),
 }));
 

package/packages/{utils/src/fetch → fetch-sse/src}/__tests__/parseError.test.ts

@@ -1,14 +1,14 @@
 import { ErrorResponse } from '@lobechat/types';
-import { afterEach, describe, expect, it, vi } from 'vitest';
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
 
 import { getMessageError } from '../parseError';
 
-// 模拟 i18next
+// Mock i18next
 vi.mock('i18next', () => ({
   t: vi.fn((key) => `translated_${key}`),
 }));
 
-// 模拟 Response
+// Mock Response
 const createMockResponse = (body: any, ok: boolean, status: number = 200) => ({
   ok,
   status,
@@ -38,11 +38,14 @@ const createMockResponse = (body: any, ok: boolean, status: number = 200) => ({
   },
 });
 
-// 在每次测试后清理所有模拟
 afterEach(() => {
   vi.restoreAllMocks();
 });
 
+beforeEach(() => {
+  vi.clearAllMocks();
+});
+
 describe('getMessageError', () => {
   it('should handle business error correctly', async () => {
     const mockErrorResponse: ErrorResponse = {

package/packages/{utils/src/fetch → fetch-sse/src}/fetchSSE.ts

@@ -12,9 +12,9 @@ import {
   ResponseAnimation,
   ResponseAnimationStyle,
 } from '@lobechat/types';
+import { fetchEventSource } from '@lobechat/utils/client/fetchEventSource/index';
+import { nanoid } from '@lobechat/utils/uuid';
 
-import { fetchEventSource } from '../client/fetchEventSource';
-import { nanoid } from '../uuid';
 import { getMessageError } from './parseError';
 
 type SSEFinishType = 'done' | 'error' | 'abort';

package/packages/{utils/src/fetch → fetch-sse/src}/parseError.ts

@@ -1,7 +1,7 @@
 import { ChatMessageError, ErrorResponse, ErrorType } from '@lobechat/types';
 import { t } from 'i18next';
 
-export const getMessageError = async (response: Response) => {
+export const getMessageError = async (response: Response): Promise<ChatMessageError> => {
   let chatMessageError: ChatMessageError;
 
   // try to get the biz error
@@ -9,13 +9,13 @@ export const getMessageError = async (response: Response) => {
     const data = (await response.json()) as ErrorResponse;
     chatMessageError = {
       body: data.body,
-      message: t(`response.${data.errorType}` as any, { ns: 'error' }),
+      message: t(`response.${data.errorType}`, { ns: 'error' }),
       type: data.errorType,
     };
   } catch {
     // if not return, then it's a common error
     chatMessageError = {
-      message: t(`response.${response.status}` as any, { ns: 'error' }),
+      message: t(`response.${response.status}`, { ns: 'error' }),
       type: response.status as ErrorType,
     };
   }

package/packages/model-runtime/src/core/contextBuilders/anthropic.test.ts

@@ -1,8 +1,8 @@
+import { imageUrlToBase64 } from '@lobechat/utils';
 import { OpenAI } from 'openai';
-import { describe, expect, it, vi } from 'vitest';
+import { beforeEach, describe, expect, it, vi } from 'vitest';
 
 import { OpenAIChatMessage, UserMessageContentPart } from '../../types/chat';
-import { imageUrlToBase64 } from '../../utils/imageToBase64';
 import { parseDataUri } from '../../utils/uriParser';
 import {
   buildAnthropicBlock,
@@ -12,16 +12,22 @@ import {
 } from './anthropic';
 
 // Mock the parseDataUri function since it's an implementation detail
-vi.mock('../../utils/uriParser', () => ({
-  parseDataUri: vi.fn().mockReturnValue({
-    mimeType: 'image/jpeg',
-    base64: 'base64EncodedString',
-    type: 'base64',
-  }),
+vi.mock('../../utils/uriParser');
+vi.mock('@lobechat/utils', () => ({
+  imageUrlToBase64: vi.fn(),
 }));
-vi.mock('../../utils/imageToBase64');
 
 describe('anthropicHelpers', () => {
+  beforeEach(() => {
+    vi.resetAllMocks();
+    // Set default mock implementation for parseDataUri
+    vi.mocked(parseDataUri).mockReturnValue({
+      mimeType: 'image/jpeg',
+      base64: 'base64EncodedString',
+      type: 'base64',
+    });
+  });
+
   describe('buildAnthropicBlock', () => {
     it('should return the content as is for text type', async () => {
       const content: UserMessageContentPart = { type: 'text', text: 'Hello!' };
@@ -52,7 +58,7 @@ describe('anthropicHelpers', () => {
         base64: null,
         type: 'url',
       });
-      vi.mocked(imageUrlToBase64).mockResolvedValue({
+      vi.mocked(imageUrlToBase64).mockResolvedValueOnce({
         base64: 'convertedBase64String',
         mimeType: 'image/jpg',
       });
@@ -82,7 +88,7 @@ describe('anthropicHelpers', () => {
         base64: null,
         type: 'url',
       });
-      vi.mocked(imageUrlToBase64).mockResolvedValue({
+      vi.mocked(imageUrlToBase64).mockResolvedValueOnce({
         base64: 'convertedBase64String',
         mimeType: 'image/png',
       });

package/packages/model-runtime/src/core/contextBuilders/anthropic.ts

@@ -1,8 +1,8 @@
 import Anthropic from '@anthropic-ai/sdk';
+import { imageUrlToBase64 } from '@lobechat/utils';
 import OpenAI from 'openai';
 
 import { OpenAIChatMessage, UserMessageContentPart } from '../../types';
-import { imageUrlToBase64 } from '../../utils/imageToBase64';
 import { parseDataUri } from '../../utils/uriParser';
 
 export const buildAnthropicBlock = async (

package/packages/model-runtime/src/core/contextBuilders/google.test.ts

@@ -1,9 +1,9 @@
 // @vitest-environment node
 import { Type as SchemaType } from '@google/genai';
+import * as imageToBase64Module from '@lobechat/utils';
 import { describe, expect, it, vi } from 'vitest';
 
 import { ChatCompletionTool, OpenAIChatMessage, UserMessageContentPart } from '../../types';
-import * as imageToBase64Module from '../../utils/imageToBase64';
 import { parseDataUri } from '../../utils/uriParser';
 import {
   buildGoogleMessage,

package/packages/model-runtime/src/core/contextBuilders/google.ts

@@ -5,9 +5,9 @@ import {
   Part,
   Type as SchemaType,
 } from '@google/genai';
+import { imageUrlToBase64 } from '@lobechat/utils';
 
 import { ChatCompletionTool, OpenAIChatMessage, UserMessageContentPart } from '../../types';
-import { imageUrlToBase64 } from '../../utils/imageToBase64';
 import { safeParseJSON } from '../../utils/safeParseJSON';
 import { parseDataUri } from '../../utils/uriParser';
 
@@ -64,12 +64,9 @@ export const buildGooglePart = async (
       }
 
       if (type === 'url') {
-        // For video URLs, we need to fetch and convert to base64
+        // Use imageUrlToBase64 for SSRF protection (works for any binary data including videos)
         // Note: This might need size/duration limits for practical use
-        const response = await fetch(content.video_url.url);
-        const arrayBuffer = await response.arrayBuffer();
-        const base64 = Buffer.from(arrayBuffer).toString('base64');
-        const mimeType = response.headers.get('content-type') || 'video/mp4';
+        const { base64, mimeType } = await imageUrlToBase64(content.video_url.url);
 
         return {
           inlineData: { data: base64, mimeType },

package/packages/model-runtime/src/core/contextBuilders/openai.test.ts

@@ -1,7 +1,8 @@
+import { imageUrlToBase64 } from '@lobechat/utils';
 import OpenAI from 'openai';
 import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
 
-import { imageUrlToBase64 } from '../../utils/imageToBase64';
+import { OpenAIChatMessage } from '../../types';
 import { parseDataUri } from '../../utils/uriParser';
 import {
   convertImageUrlToFile,
@@ -11,7 +12,9 @@ import {
 } from './openai';
 
 // 模拟依赖
-vi.mock('../../utils/imageToBase64');
+vi.mock('@lobechat/utils', () => ({
+  imageUrlToBase64: vi.fn(),
+}));
 vi.mock('../../utils/uriParser');
 
 describe('convertMessageContent', () => {
@@ -147,11 +150,71 @@ describe('convertOpenAIMessages', () => {
 
     expect(Promise.all).toHaveBeenCalledTimes(2); // 一次用于消息数组，一次用于内容数组
   });
+
+  it('should filter out reasoning field from messages', async () => {
+    const messages = [
+      {
+        role: 'assistant',
+        content: 'Hello',
+        reasoning: { content: 'some reasoning', duration: 100 },
+      },
+      { role: 'user', content: 'Hi' },
+    ] as any;
+
+    const result = await convertOpenAIMessages(messages);
+
+    expect(result).toEqual([
+      { role: 'assistant', content: 'Hello' },
+      { role: 'user', content: 'Hi' },
+    ]);
+    // Ensure reasoning field is removed
+    expect((result[0] as any).reasoning).toBeUndefined();
+  });
+
+  it('should preserve reasoning_content field from messages (for DeepSeek compatibility)', async () => {
+    const messages = [
+      {
+        role: 'assistant',
+        content: 'Hello',
+        reasoning_content: 'some reasoning content',
+      },
+      { role: 'user', content: 'Hi' },
+    ] as any;
+
+    const result = await convertOpenAIMessages(messages);
+
+    expect(result).toEqual([
+      { role: 'assistant', content: 'Hello', reasoning_content: 'some reasoning content' },
+      { role: 'user', content: 'Hi' },
+    ]);
+    // Ensure reasoning_content field is preserved
+    expect((result[0] as any).reasoning_content).toBe('some reasoning content');
+  });
+
+  it('should filter out reasoning but preserve reasoning_content field', async () => {
+    const messages = [
+      {
+        role: 'assistant',
+        content: 'Hello',
+        reasoning: { content: 'some reasoning', duration: 100 },
+        reasoning_content: 'some reasoning content',
+      },
+    ] as any;
+
+    const result = await convertOpenAIMessages(messages);
+
+    expect(result).toEqual([
+      { role: 'assistant', content: 'Hello', reasoning_content: 'some reasoning content' },
+    ]);
+    // Ensure reasoning object is removed but reasoning_content is preserved
+    expect((result[0] as any).reasoning).toBeUndefined();
+    expect((result[0] as any).reasoning_content).toBe('some reasoning content');
+  });
 });
 
 describe('convertOpenAIResponseInputs', () => {
   it('应该正确转换普通文本消息', async () => {
-    const messages: OpenAI.ChatCompletionMessageParam[] = [
+    const messages: OpenAIChatMessage[] = [
       { role: 'user', content: 'Hello' },
       { role: 'assistant', content: 'Hi there!' },
     ];
@@ -165,7 +228,7 @@ describe('convertOpenAIResponseInputs', () => {
   });
 
   it('应该正确转换带有工具调用的消息', async () => {
-    const messages: OpenAI.ChatCompletionMessageParam[] = [
+    const messages: OpenAIChatMessage[] = [
       {
         role: 'assistant',
         content: '',
@@ -195,7 +258,7 @@ describe('convertOpenAIResponseInputs', () => {
   });
 
   it('应该正确转换工具响应消息', async () => {
-    const messages: OpenAI.ChatCompletionMessageParam[] = [
+    const messages: OpenAIChatMessage[] = [
       {
         role: 'tool',
         content: 'Function result',
@@ -215,7 +278,7 @@ describe('convertOpenAIResponseInputs', () => {
   });
 
   it('应该正确转换包含图片的消息', async () => {
-    const messages: OpenAI.ChatCompletionMessageParam[] = [
+    const messages: OpenAIChatMessage[] = [
       {
         role: 'user',
         content: [
@@ -247,7 +310,7 @@ describe('convertOpenAIResponseInputs', () => {
   });
 
   it('应该正确处理混合类型的消息序列', async () => {
-    const messages: OpenAI.ChatCompletionMessageParam[] = [
+    const messages: OpenAIChatMessage[] = [
       { role: 'user', content: 'I need help with a function' },
       {
         role: 'assistant',
@@ -287,6 +350,29 @@ describe('convertOpenAIResponseInputs', () => {
       },
     ]);
   });
+
+  it('should extract reasoning.content into a separate reasoning item', async () => {
+    const messages: OpenAIChatMessage[] = [
+      { content: 'system prompts', role: 'system' },
+      { content: '你好', role: 'user' },
+      {
+        content: 'hello',
+        role: 'assistant',
+        reasoning: { content: 'reasoning content', duration: 2706 },
+      },
+      { content: '杭州天气如何', role: 'user' },
+    ];
+
+    const result = await convertOpenAIResponseInputs(messages);
+
+    expect(result).toEqual([
+      { content: 'system prompts', role: 'developer' },
+      { content: '你好', role: 'user' },
+      { summary: [{ text: 'reasoning content', type: 'summary_text' }], type: 'reasoning' },
+      { content: 'hello', role: 'assistant' },
+      { content: '杭州天气如何', role: 'user' },
+    ]);
+  });
 });
 
 describe('convertImageUrlToFile', () => {