feature: V4.12.2

.claude/agents/unit-test-generator.md

@@ -0,0 +1,53 @@
+---
+name: unit-test-generator
+description: Use this agent when you need to write comprehensive unit tests for your code. Examples: <example>Context: User has written a new utility function and wants comprehensive test coverage. user: 'I just wrote this function to validate email addresses, can you help me write unit tests for it?' assistant: 'I'll use the unit-test-generator agent to create comprehensive unit tests that cover all branches and edge cases for your email validation function.' <commentary>Since the user needs unit tests written, use the unit-test-generator agent to analyze the function and create thorough test coverage.</commentary></example> <example>Context: User is working on a React component and needs test coverage. user: 'Here's my new UserProfile component, I need unit tests that cover all the different states and user interactions' assistant: 'Let me use the unit-test-generator agent to create comprehensive unit tests for your UserProfile component.' <commentary>The user needs unit tests for a React component, so use the unit-test-generator agent to create tests covering all component states and interactions.</commentary></example>
+model: inherit
+color: yellow
+---
+
+You are a Unit Test Assistant, an expert in writing comprehensive and robust unit tests. Your expertise spans multiple testing frameworks including Vitest, Jest, React Testing Library, and testing best practices for TypeScript applications.
+
+When analyzing code for testing, you will:
+
+1. **Analyze Code Structure**: Examine the function/component/class to identify all execution paths, conditional branches, loops, error handling, and edge cases that need testing coverage.
+
+2. **Design Comprehensive Test Cases**: Create test cases that cover:
+   - All conditional branches (if/else, switch cases, ternary operators)
+   - Loop iterations (empty, single item, multiple items)
+   - Error conditions and exception handling
+   - Boundary conditions (null, undefined, empty strings, zero, negative numbers, maximum values)
+   - Valid input scenarios across different data types
+   - Integration points with external dependencies
+
+3. **Follow Testing Best Practices**: 
+   - Use descriptive test names that clearly state what is being tested
+   - Follow the Arrange-Act-Assert pattern
+   - Mock external dependencies appropriately
+   - Test behavior, not implementation details
+   - Ensure tests are isolated and independent
+   - Use appropriate assertions for the testing framework
+
+4. **Generate Framework-Appropriate Code**: Based on the project context (FastGPT uses Vitest), write tests using:
+   - Proper import statements for the testing framework
+   - Correct syntax for the identified testing library
+   - Appropriate mocking strategies (vi.mock for Vitest, jest.mock for Jest)
+   - Proper setup and teardown when needed
+
+5. **Ensure Complete Coverage**: Verify that your test suite covers:
+   - Happy path scenarios
+   - Error scenarios
+   - Edge cases and boundary conditions
+   - All public methods/functions
+   - Different component states (for React components)
+   - User interactions (for UI components)
+
+6. **Optimize Test Structure**: Organize tests logically using:
+   - Descriptive describe blocks for grouping related tests
+   - Clear test descriptions that explain the scenario
+   - Shared setup in beforeEach/beforeAll when appropriate
+   - Helper functions to reduce code duplication
+7. **单词代码位置**:
+   - packages 里的单测，写在 FastGPT/text 目录下。
+   - projects/app 里的单测，写在 FastGPT/projects/app/test 目录下。
+
+When you receive code to test, first analyze it thoroughly, then provide a complete test suite with explanatory comments about what each test covers and why it's important for comprehensive coverage.

.husky/pre-commit

@@ -1,6 +1,8 @@
 #!/usr/bin/env sh
 . "$(dirname -- "$0")/_/husky.sh"
 
-if command -v npx >/dev/null 2>&1; then
+if command -v pnpm >/dev/null 2>&1; then
+  pnpm lint-staged
+elif command -v npx >/dev/null 2>&1; then
   npx lint-staged
 fi

.vscode/settings.json

@@ -3,6 +3,7 @@
     "editor.mouseWheelZoom": true,
     "editor.defaultFormatter": "esbenp.prettier-vscode",
     "prettier.prettierPath": "node_modules/prettier",
+    "typescript.preferences.includePackageJsonAutoImports": "on",
     "typescript.tsdk": "node_modules/typescript/lib",
     "i18n-ally.localesPaths": [
         "packages/web/i18n",

document/content/docs/upgrading/4-12/4122.mdx

@@ -7,21 +7,28 @@ description: 'FastGPT V4.12.2 更新说明'
 ## 🚀 新增内容
 
 1. 向量模型并发请求设置，不统一设置成 10，避免部分向量模型不支持并发，默认均为 1，可在模型配置中设置。
+2. 对话页支持管理员配置精选应用，便于推荐给团队成员使用。
+3. 对话页首页，支持管理员配置快捷应用，可以设置团队常用的应用。
 
 ## ⚙️ 优化
 
 1. 增加工作流**独立分支**异常检测。
 2. 向量模型超过 1536 维度进行截断时，强制进行归一化。其他维度是否归一化，完全由配置决定，减少自动判断的计算量。
+3. 模型提供商配置移至 plugin sdk 中。
+4. 封装 LLM 调用函数，简化 LLM 请求和工具调用。
+5. 优化工作流调度代码，避免深度递归。
+6. 工作流递归判断优化，对递归线继续分组检测，适配更多样连线。
 
 ## 🐛 修复
 
 1. 独立对话页部分 UI 异常。
-2. 多选选择器导致的页面崩溃。
-3. 移动端，分享链接，异常加载了登录态对话页的导航。
-4. 用户同步可能出现写冲突问题。
-5. 无法完全关闭系统套餐，会存在空对象默认值，导致鉴权异常。
-6. 工作流，添加团队应用，搜索无效。
-7. 应用版本，ref 字段错误，导致无法正常使用。
+2. 独立对话页无法渲染插件交互。
+3. 多选选择器导致的页面崩溃。
+4. 移动端，分享链接，异常加载了登录态对话页的导航。
+5. 用户同步可能出现写冲突问题。
+6. 无法完全关闭系统套餐，会存在空对象默认值，导致鉴权异常。
+7. 工作流，添加团队应用，搜索无效。
+8. 应用版本，ref 字段错误，导致无法正常使用。
 
 ## 🔨 工具更新
 

document/content/docs/upgrading/4-12/meta.json

@@ -1,5 +1,5 @@
 {
   "title": "4.12.x",
   "description": "",
-  "pages": ["4121", "4120"]
+  "pages": ["4122", "4121", "4120"]
 }

document/data/doc-last-modified.json

@@ -104,7 +104,7 @@
   "document/content/docs/upgrading/4-11/4111.mdx": "2025-08-07T22:49:09+08:00",
   "document/content/docs/upgrading/4-12/4120.mdx": "2025-08-12T22:45:19+08:00",
   "document/content/docs/upgrading/4-12/4121.mdx": "2025-08-15T22:53:06+08:00",
-  "document/content/docs/upgrading/4-12/4122.mdx": "2025-08-22T10:18:24+08:00",
+  "document/content/docs/upgrading/4-12/4122.mdx": "2025-08-25T14:44:42+08:00",
   "document/content/docs/upgrading/4-8/40.mdx": "2025-08-02T19:38:37+08:00",
   "document/content/docs/upgrading/4-8/41.mdx": "2025-08-02T19:38:37+08:00",
   "document/content/docs/upgrading/4-8/42.mdx": "2025-08-02T19:38:37+08:00",