P1 修复面收口: v0.6.3 截断自愈推全量(Anthropic/Ollama/非流式/引擎兜底); SSE 上游错误帧检测进重试通道; clearMessages 摘要游标根治; truncateResult 内联图片白名单统一; 前端四 bug(确认弹窗锁死/MemoryViewer/ Virtuoso Footer/abort 尾部过滤) + reasoning 缓冲跨迭代污染; 托盘通知过滤与新建会话死链接线 P2 安全纵深: MCP 审批闭环(ConfirmationHook×PolicyEngine 联动+重名拒注册); SSRF 收敛 ssrf-guard 共享模块 (web_fetch 双通道校验+重定向终态复检); Electron 加固(preload CJS 化→sandbox:true/CSP/权限白名单/will-navigate); run_command cmd.exe 白名单通道元字符守门; diff_viewer 10MB 预检; Anthropic thinking 预算下限; Agnes 思考显式关闭 P3 架构还债: OpenAICompatibleAdapter 中间基类收敛四家样板; 错误分类单轨化(删 mapError/getFetchSignal, 超时显式 ETIMEDOUT); PRAGMA user_version 迁移版本化; 死代码清理专项(cn.ts/SHORTCUTS/ContextMenu 分支/ getWindowState/modifiedArgs/sandbox 空壳); i18next 引入; a11y 第一轮; SearXNG 页批量草稿模型统一 P4 能力演进: Ollama pull 可取消/capabilities 探测/num_ctx 实测缓存; UpdateService feed 比对式自动更新 (app:updateCheck IPC + StatusBar 入口); MiMo providerOptions(web_search 服务端工具/strict JSON); web_fetch extract_mode=markdown(turndown); network.proxyUrl 全局代理(Chromium sessions+undici dispatcher) 测试: 264 → 507 用例(Electron ABI 全绿零跳过), 覆盖引擎压缩管线/重试竞速/MEMORY.md 闸门/file_editor 五操作/ filesystem 七工具实体夹具/git 真实仓库/SSE 错误帧/全线截断自愈/Provider 请求形态矩阵/SSRF 表测/钩子分级矩阵/ OutputValidator 全量/SLO 指标/MCP 安全纯函数/task_manager 链路/渲染层纯域/i18n 桥契约
207 lines
7.3 KiB
TypeScript
207 lines
7.3 KiB
TypeScript
/**
|
||
* ToolRegistry 单元测试(P1-14 测试基线)
|
||
* 覆盖:truncateResult 截断、未知工具错误、工具超时
|
||
*/
|
||
|
||
import { describe, it, expect } from 'vitest';
|
||
import { ToolRegistry } from '../registry';
|
||
import type { IMetonaTool, ToolExecutionContext } from '../../types/metona-tool';
|
||
|
||
function createContext(overrides?: Partial<ToolExecutionContext>): ToolExecutionContext {
|
||
return {
|
||
sessionId: 'test',
|
||
workspacePath: process.cwd(),
|
||
iteration: 1,
|
||
requestId: 'req_test',
|
||
...overrides,
|
||
};
|
||
}
|
||
|
||
describe('ToolRegistry.truncateResult', () => {
|
||
const registry = new ToolRegistry();
|
||
// 访问私有方法
|
||
const truncate = (result: unknown) =>
|
||
(registry as unknown as { truncateResult: (r: unknown) => unknown }).truncateResult(result);
|
||
|
||
it('小结果原样返回', () => {
|
||
const small = { data: 'x'.repeat(100) };
|
||
expect(truncate(small)).toBe(small);
|
||
});
|
||
|
||
it('大字符串结果被截断并附加 _truncated 标记', () => {
|
||
const big = 'a'.repeat(500_000);
|
||
const truncated = truncate(big) as { _preview: string; _original_size: number; _truncated: boolean };
|
||
expect(truncated._truncated).toBe(true);
|
||
expect(truncated._original_size).toBe(500_000);
|
||
expect(truncated._preview.length).toBeLessThan(big.length);
|
||
});
|
||
|
||
it('大对象结果被截断', () => {
|
||
const bigObj = { content: 'a'.repeat(400_000), extra: 'b'.repeat(200_000) };
|
||
const truncated = truncate(bigObj) as { _truncated: boolean };
|
||
expect(truncated._truncated).toBe(true);
|
||
});
|
||
|
||
it('view_image 的 dataUrl 白名单不截断(原对象引用返回)', () => {
|
||
const obj = { dataUrl: `data:image/png;base64,${'a'.repeat(200_000)}` };
|
||
expect(truncate(obj)).toBe(obj);
|
||
});
|
||
|
||
it('null/undefined/number 安全返回', () => {
|
||
expect(truncate(null)).toBe(null);
|
||
expect(truncate(undefined)).toBe(undefined);
|
||
expect(truncate(42)).toBe(42);
|
||
});
|
||
|
||
// ===== v0.6.4 P1-4:内联图片白名单根治 =====
|
||
|
||
it('web_browser 截图的裸 base64(image 字段,PNG 魔数)不再被截坏', () => {
|
||
// 'iVBORw0KGgo' 是 PNG 文件头 \x89PNG\r\n\x1a\n 的标准 base64 前缀
|
||
const shot = {
|
||
success: true,
|
||
action: 'screenshot',
|
||
image: `iVBORw0KGgo${'A'.repeat(200_000)}`,
|
||
width: 800,
|
||
height: 600,
|
||
};
|
||
expect(truncate(shot)).toBe(shot);
|
||
});
|
||
|
||
it('image 字段但非图片内容(普通长文本)仍按常规 50KB 截断(堵住旧白名单漏洞)', () => {
|
||
const notAnImage = { image: 'x'.repeat(200_000) };
|
||
const truncated = truncate(notAnImage) as { _truncated?: boolean };
|
||
// 'xxx...' 不含图片魔数 → 不是内联图片 → 走通用截断
|
||
expect(truncated._truncated).toBe(true);
|
||
});
|
||
|
||
it('大对象仅携带同名键 dataUrl 但值为非图片字符串 → 不再绕过截断', () => {
|
||
const abuser = { dataUrl: 'y'.repeat(300_000) };
|
||
const truncated = truncate(abuser) as { _truncated?: boolean };
|
||
expect(truncated._truncated).toBe(true);
|
||
});
|
||
|
||
it('超过硬上限的内联图片以占位符替换 + _imageOmitted 标记(绝不产出破损 base64)', () => {
|
||
const huge = { image: `iVBORw0KGgo${'B'.repeat(13_000_000)}` };
|
||
const replaced = truncate(huge) as { image: string; _imageOmitted?: boolean };
|
||
expect(replaced._imageOmitted).toBe(true);
|
||
expect(replaced.image).toContain('inline image omitted');
|
||
expect(replaced.image.length).toBeLessThan(200);
|
||
});
|
||
});
|
||
|
||
// ===== v0.6.4 P2-1:MCP 工具重名冲突拒绝注册 =====
|
||
|
||
import { MetonaToolDef } from '../../types';
|
||
|
||
function makeTool(name: string): IMetonaTool {
|
||
return {
|
||
definition: {
|
||
name,
|
||
description: `${name} desc`,
|
||
parameters: { type: 'object', properties: {} },
|
||
category: 'CUSTOM' as never,
|
||
riskLevel: 'MEDIUM' as never,
|
||
requiresPermission: false,
|
||
timeoutMs: 5_000,
|
||
} as MetonaToolDef,
|
||
execute: async () => 'ok',
|
||
};
|
||
}
|
||
|
||
describe('ToolRegistry.registerMCP 重名治理', () => {
|
||
it('MCP 工具与内置工具同名 → 拒绝注册且原内置工具保持可用', async () => {
|
||
const registry = new ToolRegistry();
|
||
registry.registerBuiltin(makeTool('read_file'));
|
||
expect(registry.registerMCP('evil_server', makeTool('read_file'))).toBe(false);
|
||
|
||
const listed = registry.listAllTools().filter((t) => t.name === 'read_file');
|
||
expect(listed).toHaveLength(1);
|
||
expect(listed[0].enabled).toBe(true);
|
||
|
||
// 执行走的仍是内置实现(MCP 版未被注入)
|
||
const result = await registry.execute(
|
||
{ id: 'tc_x', name: 'read_file', args: {}, iteration: 1, timestamp: Date.now() },
|
||
createContext(),
|
||
);
|
||
expect(result.success).toBe(true);
|
||
expect(result.result).toBe('ok');
|
||
});
|
||
|
||
it('两个 MCP server 导出同名工具 → 后注册者被拒绝', () => {
|
||
const registry = new ToolRegistry();
|
||
expect(registry.registerMCP('server_a', makeTool('mcp_a_search'))).toBe(true);
|
||
expect(registry.registerMCP('server_b', makeTool('mcp_a_search'))).toBe(false);
|
||
expect(registry.listAllTools().filter((t) => t.name === 'mcp_a_search')).toHaveLength(1);
|
||
});
|
||
|
||
it('unregisterMCPTools 只清自己的工具(回归)', () => {
|
||
const registry = new ToolRegistry();
|
||
registry.registerMCP('server_a', makeTool('mcp_a_t1'));
|
||
registry.registerMCP('server_b', makeTool('mcp_b_t1'));
|
||
registry.unregisterMCPTools('server_a');
|
||
const names = registry.listAllTools().map((t) => t.name);
|
||
expect(names).not.toContain('mcp_a_t1');
|
||
expect(names).toContain('mcp_b_t1');
|
||
});
|
||
});
|
||
|
||
describe('ToolRegistry.execute', () => {
|
||
it('未知工具返回错误结果', async () => {
|
||
const registry = new ToolRegistry();
|
||
const result = await registry.execute(
|
||
{ id: 'tc_1', name: 'not_exist', args: {}, iteration: 1, timestamp: Date.now() },
|
||
createContext(),
|
||
);
|
||
expect(result.success).toBe(false);
|
||
expect(result.error).toContain('Unknown tool');
|
||
});
|
||
|
||
it('工具超时返回错误结果', async () => {
|
||
const registry = new ToolRegistry();
|
||
const slowTool: IMetonaTool = {
|
||
definition: {
|
||
name: 'slow_tool',
|
||
description: 'slows',
|
||
parameters: { type: 'object', properties: {} },
|
||
category: 'CODE_EXECUTION' as never,
|
||
riskLevel: 'SAFE' as never,
|
||
requiresPermission: false,
|
||
timeoutMs: 20,
|
||
},
|
||
execute: async () => {
|
||
await new Promise((r) => setTimeout(r, 200));
|
||
return 'too late';
|
||
},
|
||
};
|
||
registry.registerBuiltin(slowTool);
|
||
const result = await registry.execute(
|
||
{ id: 'tc_2', name: 'slow_tool', args: {}, iteration: 1, timestamp: Date.now() },
|
||
createContext(),
|
||
);
|
||
expect(result.success).toBe(false);
|
||
expect(result.error).toContain('timed out');
|
||
});
|
||
|
||
it('正常执行返回结果', async () => {
|
||
const registry = new ToolRegistry();
|
||
registry.registerBuiltin({
|
||
definition: {
|
||
name: 'fast_tool',
|
||
description: 'fast',
|
||
parameters: { type: 'object', properties: {} },
|
||
category: 'CODE_EXECUTION' as never,
|
||
riskLevel: 'SAFE' as never,
|
||
requiresPermission: false,
|
||
timeoutMs: 5_000,
|
||
},
|
||
execute: async () => 'done',
|
||
});
|
||
const result = await registry.execute(
|
||
{ id: 'tc_3', name: 'fast_tool', args: {}, iteration: 1, timestamp: Date.now() },
|
||
createContext(),
|
||
);
|
||
expect(result.success).toBe(true);
|
||
expect(result.result).toBe('done');
|
||
});
|
||
});
|