P0 会话可靠性收口(根治"模型思考着会话就停止"): - P0-1 finish_reason 全链路贯通:DONE 事件与 IterationStep 新增 finishReason,OpenAI 共享 SSE / Anthropic message_delta.stop_reason / Ollama done_reason 三路采集,TRACE 层弃用硬编码 'stop' 记录真值 - P0-2 空响应守卫 + 降级重试:零产出流→可重试错误走退避;思考耗尽输出预算(reasoning-only + length)→自动关闭思考降级重试一次;仍失败→OUTPUT_LENGTH_EXCEEDED 结构化错误 + 故障转移;附带根治 abort 恰逢零工具调用轮被 COMPLETED 抢占的真实缺陷 - P0-3 思考×能力×预算三对齐:DeepSeek/MiMo/Agnes/Ollama 四家 supportsThinking=false 强制不发思考参数;小输出预算告警;设置页联动提示 - P0-4 渲染层可见性:截断/空完成/友好错误三类提示,i18n 全部出层 - P0-5 回归四件套:reasoning-only 终止判定、集成级空闲超时、504 引擎重试归类、思考中 abort→USER_INTERRUPT、P4-2 强制收尾路径 FEAT-1:LLM 设置新增「最大输出上限」——Provider 支持矩阵显隐 + 模型上限钳制提示 + 超限保存警告 + llm.maxTokens 热生效 P1 修复面收口: - 渲染层三缺陷根治:后台会话回放缓冲(2000 条/4MB 有界 + agent:getReplayState + 事件总线)+ abort 双层自愈 + sendMessage 收尾兜底 + 中断卡片清扫 - 工具 abort 信号全覆盖:web_search/web_fetch/http_request/code_search/git 系列/delegate_task 全部接入引擎中断;web_search 时间预算收敛(720s→≤240s);移除伪造 ToolExecutionContext 与死代码 - 安全:本地 Pinned CONNECT 代理根治浏览器通道 DNS rebinding(校验期 IP pinning,可注入 resolver 表测);配置 URL 域名解析深校验(DeepCheckSoftFailure 软失败);SSE 空 error 帧防御修复;Ollama generate/embed AbortSignal.any 合并 - 缺陷清单:UTF-16 BOM 读取、tmp 同毫秒碰撞(nanoid 后缀)、code_search JS 回退参数对称(case_sensitive/前后文独立)、list_directory include_node_modules、崩溃自愈退避(60s 窗 ≥3 次停 reload)、MemoryViewer/Sidebar i18n 收口 P2 能力演进: - 会话回收站:SCHEMA_VERSION 3 + 迁移 10(deleted_at,存在性守卫),软删除/恢复/彻底删除/30 天自动清理(启动+24h),searchMessages 聚合剔除,Sidebar 回收站面板 - 会话回放播放器:sessions:listRecordings/readRecording(白名单+目录边界+20MB 上限),SessionReplayPlayer 时间轴/步进/变速,Trace 面板入口 - electron-updater 自动更新:双轨(手动 feed 比对保留),生产环境启动静默检查 + update:status 广播 + app:updateInstall + LogsSettings UpdatePanel + builder publish 配置 - @ 文件提及:workspace.listFiles/readFileClip(边界/512KB/NUL 拒绝/MEMORY.md 保护),ChatInput Fuse 联想+键盘导航+附件管线注入 - MCP Resources/Prompts 发现:可选能力 try/catch 降级,mcp:listServerContents,MCPSettings 展开视图 - 文档对齐:内部 API 标准 HTML(Adapter 清单补 MiMo/已实现注记/STREAM_RESET/DONE.finishReason/ repetition_truncation 映射);README v0.8.0 亮点表 P3 测试基建: - 新增 4 个测试文件:engine-stream-contract(6)、engine-stream-reliability(4:集成空闲超时/504 重试/思考中 abort/P4-2 强制收尾)、thinking-capability-gate(7)、pinned-proxy(9,含深校验 5)、session-trash(5,DB 域)、use-agent-stream hook 级(5)、agent.test 回放缓冲(2) - 契约更新:orchestrator 被中断 SubAgent success=false(abort 优先级修复语义)、SSE 空 error 帧、UTF-16 正常读取、DeepSeek 未配置思考显式 disabled、迁移矩阵 v2→3 - 弱断言根治:registry WEBP 单向断言、hooks-contracts 自比恒真、memory 空 token 补强 全量验证:typecheck 0 错误 / lint 0 问题 / 系统 Node 2144 通过(301 DB 用例按 ABI 跳过)/ Electron ABI 2445/2445 全量通过 0 跳过
231 lines
7.8 KiB
TypeScript
231 lines
7.8 KiB
TypeScript
/**
|
||
* v0.8.0 P0-5: 引擎流结束契约测试 —— "模型思考着会话停止"根治回归防线。
|
||
*
|
||
* 钉住 chatStreamWithRetry 的流结束契约(P0-2)与 finish_reason 贯通(P0-1):
|
||
* 1. 零产出流 → EmptyResponseError → 退避重试 → 重试耗尽 ERROR;
|
||
* 2. reasoning-only + finish_reason=length → 一次性降级重试(thinkingEnabled=false)
|
||
* → 降级后成功 COMPLETED;
|
||
* 3. 降级重试仍 length → OutputTruncatedError → ERROR 事件携带
|
||
* MetonaErrorCode.OUTPUT_LENGTH_EXCEEDED;
|
||
* 4. 有正文 + length → COMPLETED 且最终 DONE 事件携带 finishReason='length';
|
||
* 5. 引擎最终 DONE 贯通最后一轮 step.finishReason。
|
||
*/
|
||
|
||
import { describe, it, expect } from 'vitest';
|
||
import { AgentLoopEngine } from '../engine';
|
||
import type { AgentLoopOutput } from '../types';
|
||
import type {
|
||
MetonaRequest,
|
||
MetonaStreamEvent,
|
||
IMetonaProviderAdapter,
|
||
MetonaResponse,
|
||
} from '../../types';
|
||
import { MetonaStreamEventType, MetonaErrorCode } from '../../types';
|
||
|
||
const SYSTEM_PROMPT = {
|
||
roleDefinition: 'test',
|
||
outputConstraints: '',
|
||
safetyGuidelines: '',
|
||
};
|
||
|
||
function makeAdapter(
|
||
script: (request: MetonaRequest) => MetonaStreamEvent[],
|
||
): IMetonaProviderAdapter & { calls: MetonaRequest[] } {
|
||
const calls: MetonaRequest[] = [];
|
||
const adapter = {
|
||
providerId: 'mock' as const,
|
||
supportedModels: ['m'],
|
||
supportsToolCalling: true,
|
||
supportsThinking: true,
|
||
calls,
|
||
getContextWindow(): number {
|
||
return 128_000;
|
||
},
|
||
async send(_request: MetonaRequest): Promise<MetonaResponse> {
|
||
throw new Error('send() not expected in this test');
|
||
},
|
||
async *sendStream(request: MetonaRequest): AsyncIterable<MetonaStreamEvent> {
|
||
calls.push(request);
|
||
for (const ev of script(request)) {
|
||
yield ev;
|
||
}
|
||
},
|
||
};
|
||
return adapter as unknown as IMetonaProviderAdapter & { calls: MetonaRequest[] };
|
||
}
|
||
|
||
function makeEngine(adapter: IMetonaProviderAdapter, retryCount = 1): AgentLoopEngine {
|
||
return new AgentLoopEngine(
|
||
{
|
||
maxIterations: 3,
|
||
totalTimeoutMs: 10_000,
|
||
retryCount,
|
||
thinkingEnabled: true,
|
||
thinkingEffort: 'high',
|
||
contextWindow: 128_000,
|
||
},
|
||
adapter,
|
||
);
|
||
}
|
||
|
||
function textDelta(t: string): MetonaStreamEvent {
|
||
return {
|
||
type: MetonaStreamEventType.TEXT_DELTA,
|
||
requestId: 'r',
|
||
sessionId: 's',
|
||
iteration: 1,
|
||
seq: 0,
|
||
timestamp: Date.now(),
|
||
delta: t,
|
||
};
|
||
}
|
||
|
||
function reasoningDelta(t: string): MetonaStreamEvent {
|
||
return {
|
||
type: MetonaStreamEventType.REASONING_DELTA,
|
||
requestId: 'r',
|
||
sessionId: 's',
|
||
iteration: 1,
|
||
seq: 0,
|
||
timestamp: Date.now(),
|
||
delta: t,
|
||
};
|
||
}
|
||
|
||
function doneEvent(finishReason?: string): MetonaStreamEvent {
|
||
return {
|
||
type: MetonaStreamEventType.DONE,
|
||
requestId: 'r',
|
||
sessionId: 's',
|
||
iteration: 1,
|
||
seq: 0,
|
||
timestamp: Date.now(),
|
||
...(finishReason ? { finishReason } : {}),
|
||
};
|
||
}
|
||
|
||
describe('engine stream-end contract (v0.8.0 P0-2)', () => {
|
||
it('scenario 2: reasoning-only + length → one degraded retry with thinking disabled → COMPLETED', async () => {
|
||
let call = 0;
|
||
const adapter = makeAdapter((req) => {
|
||
call++;
|
||
if (call === 1) {
|
||
// 第一次:思考烧满输出预算(reasoning-only + length 截断)
|
||
expect(req.params.thinkingEnabled).toBe(true);
|
||
return [reasoningDelta('thinking...'), doneEvent('length')];
|
||
}
|
||
// 降级重试:thinkingEnabled 必须已被引擎置 false
|
||
expect(req.params.thinkingEnabled).toBe(false);
|
||
return [textDelta('answer'), doneEvent('stop')];
|
||
});
|
||
const engine = makeEngine(adapter);
|
||
const events: MetonaStreamEvent[] = [];
|
||
engine.on('streamEvent', (e: MetonaStreamEvent) => events.push(e));
|
||
|
||
const output = await engine.runStream(
|
||
{ role: 'user', content: 'hi', timestamp: Date.now() },
|
||
's1',
|
||
[],
|
||
SYSTEM_PROMPT,
|
||
);
|
||
|
||
expect(output.terminationReason).toBe('completed');
|
||
expect(output.finalAnswer).toBe('answer');
|
||
expect(adapter.calls.length).toBe(2);
|
||
// 降级重试经过 RETRY 事件(引擎消费后向渲染层转 STREAM_RESET)
|
||
expect(events.some((e) => e.type === MetonaStreamEventType.STREAM_RESET)).toBe(true);
|
||
});
|
||
|
||
it('scenario 2b: degraded retry still length → OUTPUT_LENGTH_EXCEEDED error (no infinite retry)', async () => {
|
||
const adapter = makeAdapter(() => [reasoningDelta('thinking...'), doneEvent('length')]);
|
||
const engine = makeEngine(adapter);
|
||
const events: MetonaStreamEvent[] = [];
|
||
engine.on('streamEvent', (e: MetonaStreamEvent) => events.push(e));
|
||
|
||
const output: AgentLoopOutput = await engine.runStream(
|
||
{ role: 'user', content: 'hi', timestamp: Date.now() },
|
||
's1',
|
||
[],
|
||
SYSTEM_PROMPT,
|
||
);
|
||
|
||
// 恰好两次调用(原始 + 一次性降级重试),随后终止
|
||
expect(adapter.calls.length).toBe(2);
|
||
expect(output.terminationReason).toBe('error');
|
||
const errorEvent = events.find((e) => e.type === MetonaStreamEventType.ERROR);
|
||
expect(errorEvent).toBeDefined();
|
||
expect(errorEvent?.error?.code).toBe(MetonaErrorCode.OUTPUT_LENGTH_EXCEEDED);
|
||
});
|
||
|
||
it('scenario 1: zero-content stream → retryable → retry exhausted → ERROR', async () => {
|
||
const adapter = makeAdapter(() => []);
|
||
const engine = makeEngine(adapter, 1);
|
||
const events: MetonaStreamEvent[] = [];
|
||
engine.on('streamEvent', (e: MetonaStreamEvent) => events.push(e));
|
||
|
||
const output = await engine.runStream(
|
||
{ role: 'user', content: 'hi', timestamp: Date.now() },
|
||
's1',
|
||
[],
|
||
SYSTEM_PROMPT,
|
||
);
|
||
|
||
expect(output.terminationReason).toBe('error');
|
||
expect(adapter.calls.length).toBe(2); // 原始 1 次 + 重试 1 次(retryCount=1)
|
||
const errorEvent = events.find((e) => e.type === MetonaStreamEventType.ERROR);
|
||
expect(errorEvent?.error?.message).toContain('empty stream');
|
||
});
|
||
|
||
it('scenario 4: text + length → COMPLETED with DONE.finishReason=length (no retry)', async () => {
|
||
const adapter = makeAdapter(() => [textDelta('partial answer'), doneEvent('length')]);
|
||
const engine = makeEngine(adapter);
|
||
const events: MetonaStreamEvent[] = [];
|
||
engine.on('streamEvent', (e: MetonaStreamEvent) => events.push(e));
|
||
|
||
const output = await engine.runStream(
|
||
{ role: 'user', content: 'hi', timestamp: Date.now() },
|
||
's1',
|
||
[],
|
||
SYSTEM_PROMPT,
|
||
);
|
||
|
||
expect(output.terminationReason).toBe('completed');
|
||
expect(output.finalAnswer).toBe('partial answer');
|
||
expect(adapter.calls.length).toBe(1); // 有产出不重试
|
||
const done = events.find((e) => e.type === MetonaStreamEventType.DONE);
|
||
expect(done?.finishReason).toBe('length');
|
||
});
|
||
|
||
it('P0-1: normal stop → DONE carries finishReason=stop from adapter', async () => {
|
||
const adapter = makeAdapter(() => [textDelta('ok'), doneEvent('stop')]);
|
||
const engine = makeEngine(adapter);
|
||
const events: MetonaStreamEvent[] = [];
|
||
engine.on('streamEvent', (e: MetonaStreamEvent) => events.push(e));
|
||
|
||
const output = await engine.runStream(
|
||
{ role: 'user', content: 'hi', timestamp: Date.now() },
|
||
's1',
|
||
[],
|
||
SYSTEM_PROMPT,
|
||
);
|
||
|
||
expect(output.terminationReason).toBe('completed');
|
||
const done = events.find((e) => e.type === MetonaStreamEventType.DONE);
|
||
expect(done?.finishReason).toBe('stop');
|
||
});
|
||
|
||
it('reasoning-only + natural stop (no length) → retryable empty response path', async () => {
|
||
const adapter = makeAdapter(() => [reasoningDelta('hmm'), doneEvent('stop')]);
|
||
const engine = makeEngine(adapter, 1);
|
||
const output = await engine.runStream(
|
||
{ role: 'user', content: 'hi', timestamp: Date.now() },
|
||
's1',
|
||
[],
|
||
SYSTEM_PROMPT,
|
||
);
|
||
// 无产出且非 length 截断 → EmptyResponseError(可重试)→ 重试耗尽 ERROR
|
||
expect(output.terminationReason).toBe('error');
|
||
expect(adapter.calls.length).toBe(2);
|
||
});
|
||
});
|