feat: v0.7.4 时序语义修正 · 防线实效补漏 · 全量测试翻倍 — 2406 用例 + jsdom 组件测试全量回归
P1 修复面收口: - 超时三态区分(aborted→USER_INTERRUPT / ETIMEDOUT→TIMEOUT / 其余→ERROR), 根治"真实网络超时被误报为用户中断" - 流空闲超时统一(SSE/Ollama/Anthropic 读循环 60s 无数据抛 504 进重试通道) - 同会话并发 sendMessage 防重入(isRunning 守卫)+ 会话存在性预检 + 前置调用移入 try(ERROR+DONE 双事件保证,根治 isStreaming 假死) - 清空审计后 resetChainCache(根治 verifyChain 误报 TAMPERED) - DONE 不再提前清理 TRACE(TERMINATED 统一收尾,补全最终迭代录制) - IME 合成回车不发送(普通 Enter + Cmd/Ctrl+Enter 双分支)+ handleSend 闭包修复 P2 安全纵深: - preload 移除原始 electronAPI 暴露(渲染层零使用,关掉 XSS invoke 任意通道单点风险) - CORS 同源回显根治(仅当前浏览页面 Origin,did-navigate 同步) - MEMORY.md 命令保护正则扩展(括号/$/反引号/< 重定向边界 + 前导路径) - write_file append TOCTOU 统一(open 后 realpath 校验,新文件分支补漏) - 敏感键归一化(authKey 驼峰/连字符命中)+ MCP headers 鉴权值加密落库 - ReDoS 检测共享化(search_files/file_editor 统一拦截) - run_tests/lint_code 升风险 + 需确认 + npx --no-install(执行边界对齐 run_command) - MCP/SearXNG/llm.baseURL/updateFeedUrl 配置类 URL 高危目标校验(IPv6 去括号 + 十六进制映射解析 + 尾点剥离) P3 架构还债: - temperature/maxTokens 热生效(引擎/编排器/SubAgent 三处接线)+ setBatch 单事务落盘 - SessionRecorder flush 竞态根治(flushPromise 等待 + 超限内联落盘 + stopRecording async) - 内存收口(lastConsolidationBySession LRU / subTraces 清理 / 会话删除 disposeEngine) - i18n 全量收口(28 组件 + 353 key 双字典,状态标签改渲染时函数) - 死代码清理(updateTraceStep/HEADER_HEIGHT/void preA/失实注释) - 斜杠菜单 MUI 化 + 删除逻辑收敛 resetSessionState + Blob URL 统一释放 + 用户消息"仅保存"落库(saveMessage 透传前端 id 修复 id 错位) P4 能力演进: - 死循环检测拆分(驻留前置 + 乒乓后置带进度信号,合法交替不误报) - run-lock 30s 超时强制 abort(旧 run 卡死不无限排队) - RETRY 双通道 stream_reset(前端按 run 归属精确清空,根治重试文本重复) - FTS5 trigram 中文子串搜索(迁移 9 版本化 SCHEMA_VERSION=2,≤2 字符 LIKE 回退) - getContextWindow 兜底 1M→128K(未知模型防 413) 测试: - 855 → 2406 用例(+1551,2.8 倍):服务层 +325(含 MemoryManager 51 新用例)、 工具实体 +483、IPC/适配器 +390(含 OpenAI/Anthropic/Ollama 独立套件)、 纯函数表格化 +330;引入 jsdom + @testing-library(14 组件测试文件 249 用例) - 修复 R1(saveMessage id 透传)/ R2(stream_reset 精确归属)两个回归缺陷 - 遗留低危项清零:git-tools 顺序耦合 / web-fetch 真实时间退避 / slo 内存断言 / mcp-security 多余 skipIf / deepseek-balance 命名误导 / 组件 mock 注入脆弱性 版本: 0.7.4; README 同步(工具风险表/版本徽章); 依赖: 移除 @electron-toolkit/preload, 新增 jsdom/@testing-library(devDependencies 不打包) 回归: typecheck 双端 0 错误; ESLint 0/0; Electron ABI 全量 2406/2406 零跳过; 系统 Node 2110 通过 296 跳过(better-sqlite3 ABI)
This commit is contained in:
@@ -17,11 +17,12 @@ import { MetonaStreamEventType } from '../../types';
|
||||
/** 构造 Mock Adapter:sendStream 按脚本产出事件 */
|
||||
function createMockAdapter(
|
||||
scripts: MetonaStreamEvent[][],
|
||||
opts?: { failWith?: Error },
|
||||
opts?: { failWith?: Error; providerId?: string },
|
||||
): IMetonaProviderAdapter {
|
||||
let call = 0;
|
||||
const providerId = opts?.providerId ?? 'mock';
|
||||
return {
|
||||
providerId: 'mock',
|
||||
providerId,
|
||||
supportedModels: ['mock-model'],
|
||||
supportsToolCalling: true,
|
||||
supportsThinking: false,
|
||||
@@ -30,7 +31,7 @@ function createMockAdapter(
|
||||
async (): Promise<MetonaResponse> => ({
|
||||
meta: {
|
||||
requestId: 'r_test',
|
||||
provider: 'mock',
|
||||
provider: providerId,
|
||||
model: 'mock-model',
|
||||
latencyMs: 1,
|
||||
timestamp: Date.now(),
|
||||
@@ -170,9 +171,13 @@ describe('AgentLoopEngine', () => {
|
||||
// 主 adapter 每次都失败(401 不可重试)
|
||||
const primary = createMockAdapter([], {
|
||||
failWith: Object.assign(new Error('401 invalid key'), { status: 401 }),
|
||||
providerId: 'primary-mock',
|
||||
});
|
||||
// fallback 正常返回 —— v0.7.4: 使用不同 providerId,修复旧断言 `to==='mock'`
|
||||
// 无法证明"切换到了正确的 fallback"(primary 与 fallback 共用 'mock' 的假阳性)
|
||||
const fallback = createMockAdapter([textDoneEvent('fallback answer')], {
|
||||
providerId: 'fallback-mock',
|
||||
});
|
||||
// fallback 正常返回
|
||||
const fallback = createMockAdapter([textDoneEvent('fallback answer')]);
|
||||
|
||||
const engine = new AgentLoopEngine({ retryCount: 0 }, primary);
|
||||
engine.setFallbackAdapter(fallback);
|
||||
@@ -185,8 +190,8 @@ describe('AgentLoopEngine', () => {
|
||||
expect(output.terminationReason).toBe(TerminationReason.COMPLETED);
|
||||
expect(output.finalAnswer).toBe('fallback answer');
|
||||
expect(switchEvents.length).toBe(1);
|
||||
expect(switchEvents[0].from).toBe('mock');
|
||||
expect(switchEvents[0].to).toBe('mock');
|
||||
expect(switchEvents[0].from).toBe('primary-mock');
|
||||
expect(switchEvents[0].to).toBe('fallback-mock');
|
||||
// fallback 的 sendStream 被调用
|
||||
expect(fallback.sendStream).toHaveBeenCalled();
|
||||
});
|
||||
@@ -194,9 +199,11 @@ describe('AgentLoopEngine', () => {
|
||||
it('P1 故障转移仅触发一次(fallback 也失败不回切)', async () => {
|
||||
const primary = createMockAdapter([], {
|
||||
failWith: Object.assign(new Error('401'), { status: 401 }),
|
||||
providerId: 'primary-mock',
|
||||
});
|
||||
const fallback = createMockAdapter([], {
|
||||
failWith: Object.assign(new Error('500'), { status: 500 }),
|
||||
providerId: 'fallback-mock',
|
||||
});
|
||||
|
||||
const engine = new AgentLoopEngine({ retryCount: 0 }, primary);
|
||||
@@ -206,6 +213,84 @@ describe('AgentLoopEngine', () => {
|
||||
// fallback 失败 → ERROR(不回切 primary)
|
||||
expect(output.terminationReason).toBe(TerminationReason.ERROR);
|
||||
});
|
||||
|
||||
// ===== v0.7.4 P1-1: 超时误判三态 =====
|
||||
|
||||
it('P1-1 真实网络超时(ETIMEDOUT)映射为 TIMEOUT 而非 USER_INTERRUPT', async () => {
|
||||
// 模拟 fetchWithTimeout 超时:message 含 "timed out" 且 code=ETIMEDOUT
|
||||
const adapter = createMockAdapter([], {
|
||||
failWith: Object.assign(new Error('Request timed out after 30000ms'), {
|
||||
code: 'ETIMEDOUT',
|
||||
}),
|
||||
});
|
||||
const engine = new AgentLoopEngine({ retryCount: 0 }, adapter);
|
||||
const output = await engine.runStream(userMessage, 's1', [], systemPrompt);
|
||||
expect(output.terminationReason).toBe(TerminationReason.TIMEOUT);
|
||||
});
|
||||
|
||||
it('P1-1 用户主动 abort 映射为 USER_INTERRUPT(非超时)', async () => {
|
||||
// 模拟用户中断:sendStream 挂起(模拟 LLM 长响应),run 进行中调用 abort()
|
||||
// 引擎 chatStreamWithRetry 的 catch 首先检查 this.aborted → 抛错 →
|
||||
// executeRunStream catch 中 this.aborted=true → USER_INTERRUPT。
|
||||
// (真实实现中 adapter 的 fetch 会因 abortController.abort() 而 reject)
|
||||
const hangingAdapter = {
|
||||
providerId: 'mock',
|
||||
supportedModels: ['mock-model'],
|
||||
supportsToolCalling: true,
|
||||
supportsThinking: false,
|
||||
getContextWindow: () => 1_000_000,
|
||||
send: vi.fn(async () => ({
|
||||
meta: {
|
||||
requestId: 'r1',
|
||||
provider: 'mock',
|
||||
model: 'm',
|
||||
latencyMs: 1,
|
||||
timestamp: Date.now(),
|
||||
},
|
||||
content: '',
|
||||
usage: { inputTokens: 0, outputTokens: 0, totalTokens: 0 },
|
||||
finishReason: 'stop' as never,
|
||||
})),
|
||||
sendStream: vi.fn(async function* (): AsyncIterable<MetonaStreamEvent> {
|
||||
// 挂起直到 abort 信号触发(真实流读取被 abort 打断的表现)
|
||||
// 先 yield 一个占位值满足 generator 契约(require-yield),随后挂起
|
||||
await new Promise<void>((resolve, reject) => {
|
||||
const t = setTimeout(() => reject(new Error('The operation was aborted')), 30);
|
||||
// 让外层 abort() 有机会先执行;abort 后 adapter 的 setAbortSignal 收到信号
|
||||
(t as unknown as { unref?: () => void }).unref?.();
|
||||
});
|
||||
// 不可达——上方 Promise 永不 resolve(只 reject 或挂起);此处 yield 仅为满足 generator 契约
|
||||
yield {
|
||||
type: MetonaStreamEventType.TEXT_DELTA,
|
||||
requestId: 'r1',
|
||||
sessionId: 's1',
|
||||
iteration: 1,
|
||||
seq: 0,
|
||||
timestamp: Date.now(),
|
||||
delta: '',
|
||||
};
|
||||
}),
|
||||
setAbortSignal: vi.fn(),
|
||||
healthCheck: async () => true,
|
||||
} as unknown as IMetonaProviderAdapter;
|
||||
|
||||
const engine = new AgentLoopEngine({ retryCount: 0 }, hangingAdapter);
|
||||
// 启动 run,稍后 abort(模拟用户在响应期间点击停止)
|
||||
const runPromise = engine.runStream(userMessage, 's1', [], systemPrompt);
|
||||
await new Promise((r) => setTimeout(r, 10));
|
||||
engine.abort();
|
||||
const output = await runPromise;
|
||||
expect(output.terminationReason).toBe(TerminationReason.USER_INTERRUPT);
|
||||
});
|
||||
|
||||
it('P1-1 其他错误(401)映射为 ERROR', async () => {
|
||||
const adapter = createMockAdapter([], {
|
||||
failWith: Object.assign(new Error('401 unauthorized'), { status: 401 }),
|
||||
});
|
||||
const engine = new AgentLoopEngine({ retryCount: 0 }, adapter);
|
||||
const output = await engine.runStream(userMessage, 's1', [], systemPrompt);
|
||||
expect(output.terminationReason).toBe(TerminationReason.ERROR);
|
||||
});
|
||||
});
|
||||
|
||||
// ===== v0.7.3 P4-4 / P3-1: 死循环乒乓检测 + REFLECTING 状态接线 =====
|
||||
@@ -246,6 +331,64 @@ describe('AgentLoopEngine — 死循环乒乓检测(ABAB,P4-4)', () => {
|
||||
const output = await engine.runStream(userMessage, 's1', [], systemPrompt);
|
||||
expect(output.terminationReason).toBe(TerminationReason.DEAD_LOOP);
|
||||
});
|
||||
|
||||
// ===== v0.7.4 P4-1: 乒乓检测进度信号二次确认 =====
|
||||
|
||||
it('P4-1 ABAB 但工具结果有差异(进度推进)不误报死循环', async () => {
|
||||
// A=read_file B=write_file,参数相同构成 ABAB;但 write 的**结果**每轮变化
|
||||
// (如写入内容随文件内容推进而不同)→ 进度信号存在 → 不应判定死循环
|
||||
const readScript = toolCallEvent('read_file', { file_path: 'x.ts' });
|
||||
const writeScript = toolCallEvent('write_file', { file_path: 'x.ts' });
|
||||
// 工具执行结果由 ToolRegistry 提供;此处用"伪造 registry 返回变化结果"的方式:
|
||||
// 引擎的 detectPingPong 在 EXECUTING 之后用 step.toolResults 做进度比对,
|
||||
// 而 toolResults 来自 registry.execute —— 注入一个结果变化的 registry。
|
||||
let execCounter = 0;
|
||||
const registry = {
|
||||
execute: vi.fn(async (tc: { id: string; name: string; args: Record<string, unknown> }) => {
|
||||
// 每次执行返回递增 revision(模拟结果随推进变化)
|
||||
execCounter++;
|
||||
return {
|
||||
toolCallId: tc.id,
|
||||
toolName: tc.name,
|
||||
result: { ok: true, revision: execCounter },
|
||||
success: true,
|
||||
durationMs: 1,
|
||||
timestamp: Date.now(),
|
||||
};
|
||||
}),
|
||||
get: () => ({ definition: { timeoutMs: 5000 } }),
|
||||
listTools: () => [],
|
||||
} as never;
|
||||
|
||||
const adapter = createMockAdapter([readScript, writeScript, readScript, writeScript]);
|
||||
const engine = new AgentLoopEngine({ maxIterations: 6 }, adapter, registry);
|
||||
const output = await engine.runStream(userMessage, 's1', [], systemPrompt);
|
||||
// 结果在推进 → 不判定死循环 → 因 maxIterations=6 且持续调用工具,最终 MAX_ITERATIONS
|
||||
expect(output.terminationReason).toBe(TerminationReason.MAX_ITERATIONS);
|
||||
});
|
||||
|
||||
it('P4-1 ABAB 且工具结果完全相同(空转)仍判定死循环', async () => {
|
||||
const readScript = toolCallEvent('read_file', { file_path: 'x.ts' });
|
||||
const writeScript = toolCallEvent('write_file', { file_path: 'x.ts' });
|
||||
// 结果恒相同:模拟 read 返回同一内容、write 返回同一字节数(空转无推进)
|
||||
const registry = {
|
||||
execute: vi.fn(async (tc: { id: string; name: string; args: Record<string, unknown> }) => ({
|
||||
toolCallId: tc.id,
|
||||
toolName: tc.name,
|
||||
result: { ok: true, hash: 'same' },
|
||||
success: true,
|
||||
durationMs: 1,
|
||||
timestamp: Date.now(),
|
||||
})),
|
||||
get: () => ({ definition: { timeoutMs: 5000 } }),
|
||||
listTools: () => [],
|
||||
} as never;
|
||||
|
||||
const adapter = createMockAdapter([readScript, writeScript, readScript, writeScript]);
|
||||
const engine = new AgentLoopEngine({ maxIterations: 6 }, adapter, registry);
|
||||
const output = await engine.runStream(userMessage, 's1', [], systemPrompt);
|
||||
expect(output.terminationReason).toBe(TerminationReason.DEAD_LOOP);
|
||||
});
|
||||
});
|
||||
|
||||
describe('AgentLoopEngine — REFLECTING 状态接线(P3-1 enableReflection)', () => {
|
||||
|
||||
Reference in New Issue
Block a user