fix(consult): five-minute answer clock as a hang guard; general mode runs on it from the start (BUG-1142)
The 70s answer clock was sized for a non-reasoning writer; thinking tokens come out of the same clock and deepseek-v4-pro took 77s on a parents answer. Product 2026-10-01 chose five minutes. The general / no-birth-minute loop has no tools, so it now starts on the answer clock instead of the 110s tool clock. Tool phase and domain budget unchanged; maxDuration 240 -> 480. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01N4f2nya58RoRu4yEmJgRGE
This commit is contained in:
Jesse_Chen
co-authored by
Claude Opus 5.5
Claude Opus 5.5
parent
8912ba685a
commit
04c09cbf4d
@@ -1,5 +1,11 @@
|
||||
# 印度占星 Skill 更新日志
|
||||
|
||||
## 2026-10-01 — 普通对话写回答最多可以等 5 分钟（原来 70 秒）
|
||||
|
||||
- 推理较重的模型（例如 deepseek-v4-pro）写长回答不再在最后一节被截断；5 分钟只用来防模型卡死，不限制回答长度（BUG-1142）。
|
||||
- 没填出生分钟的档案，回答也按 5 分钟算，不再被 110 秒的计算时限一起截断。
|
||||
- 计算阶段仍是 110 秒，算几个领域不变；生时校正不受影响。Skill 不 bump，不改数据库。
|
||||
|
||||
## 2026-10-01 — 生时校正：每道题只出现一次，复述变短，回复不再挂「已完成 N 步」（待验收）
|
||||
|
||||
- 同一道题被重新出题时，以前会出现「题干、题干、选项」，或者只剩一行没选项的题干加「没有拿到下一个问题」；现在屏幕上只有当前这道题出一次题干，旧题什么都不显示（BUG-1135）。
|
||||
|
||||
@@ -15332,3 +15332,18 @@
|
||||
- 复发自：无
|
||||
- 修复版本：研究分支 `codex/rectification-angle-timing-research-20261001`。
|
||||
|
||||
|
||||
## BUG-1142 | 普通对话答题时钟 70 s 容不下推理模型；无出生分钟路线只有 110 s 一道闸
|
||||
|
||||
- 状态：resolved（分支已验收；部署后以 `/api/health` 版本为准）
|
||||
- 首次发现 / 最近更新：2026-10-01 / 2026-10-01
|
||||
- 影响面：普通对话（`/api/consult`，agentic 运行时）写回答的阶段。生时校正有自己的预算（`RECTIFICATION_RUN_BUDGET_MS`），不受影响。
|
||||
- 用户现象：选推理较重的模型时，回答写到最后一节被截断，显示「回答未完成，已保留现有内容；本次不会扣点」。
|
||||
- 触发条件：10-01 真实模型对比（`docs/testing/consult-plain-answer-20261001-model-runs.md`）中，`deepseek-v4-pro` 父母题从发出到写完 77 s，超过 70 s 答题时钟；`deepseek-flash` 30–41 s，其中约 85% 是推理 token。同日取消了回答字数上限（TASK-consult-plain-answer-20261001 D8），回答平均变长约 50%。另一条：无出生分钟 / 公开日历路线（`usesPublicDailyGeneralAgent`）没有工具，`answerReady` 永远为 false，答题时钟只在续写和重试时才启动，整段模型循环被 110 s 工具时钟管着。
|
||||
- 根因：70 s 按非推理写作模型的吞吐（约 90–100 tok/s）定，而回答步骤开着 provider thinking，推理 token 也从同一只钟里扣；无工具的路线没有「拿到计算结果」这个交接点。
|
||||
- 修复：`CONSULTATION_ANSWER_TIMEOUT_MS` 70_000 → 300_000（产品 10-01 定 5 分钟，只防卡死、不限长度）；`createConsultationRunClock` 加 `answerFromStart`，route 对 `usesPublicDailyGeneralAgent` 的路线从第一步起走答题时钟；`maxDuration` 240 → 480（110 + 300 + 准备与结算）。工具阶段 110 s、领域预算（65 s / 2 个领域，由工具时钟与 45 s 预留推出，不读答题时钟）不变。截断语义不变：答题时钟到点仍是 `answer_truncated`、不扣点（BUG-1051）。
|
||||
- 验证：`frontend/tests/consult-answer-clock-20261001.test.ts`（答题时钟 300 s / 工具 110 s / maxDuration 留余量；领域预算仍 65 s、2 个且公式不读答题时钟；`answerFromStart` 时工具钟到点不掐循环、答题钟到点仍掐；route 对无分钟路线接线）；两条既有断言按三栏改；全量测试失败名单与基线逐条一致。
|
||||
- 防复发：答题时钟的注释写明它只防卡死；下游各层（Caddy 无响应超时、Node 无自定义超时、undici bodyTimeout 300 s 为「两块数据之间」的空闲时间、推理 token 也是流式下发、预留扣点租约 15 分钟）均在 410 s 以上或不按总时长计。
|
||||
- 相关记录：BUG-1051、BUG-1053、BUG-944。
|
||||
- 复发自：无
|
||||
- 修复版本：`codex/consult-answer-clock-20261001`
|
||||
|
||||
@@ -0,0 +1,35 @@
|
||||
# PROGRESS · 普通对话答题时钟 5 分钟（2026-10-01）
|
||||
|
||||
## 改动
|
||||
|
||||
- `frontend/src/mastra/consultation-tools.ts`：`CONSULTATION_ANSWER_TIMEOUT_MS` 70_000 → 300_000，注释改写；`createConsultationRunClock` 新增 `answerFromStart`（创建时即启动答题时钟，工具钟到点因 `if (answer) return` 不再掐循环）。
|
||||
- `frontend/src/app/api/consult/route.ts`：`maxDuration` 240 → 480，注释改 410 s；run clock 传 `answerFromStart: usesPublicDailyGeneralAgent(consultationMode, generalDailyContext)`。
|
||||
- 申报时段路线：有窗口工具，拿到窗口计算结果时经 `onAnswerPhase` 切答题时钟（原有行为不变）；预计算在 warmup 里用工具钟。
|
||||
- 新测试 `frontend/tests/consult-answer-clock-20261001.test.ts`（4 条）。
|
||||
|
||||
## 改动的既有断言
|
||||
|
||||
| 文件 | 原值 | 新值 | 原因 |
|
||||
|---|---|---|---|
|
||||
| `tests/consult-answer-truncation-20260926.test.ts` | `CONSULTATION_ANSWER_TIMEOUT_MS = 70_000`；总和 ≤ 180_000 | `= 300_000`；总和 ≤ 410_000 | 产品 10-01 定 5 分钟防卡死 |
|
||||
| `tests/consult-single-pass-answer-20260927.test.ts` | 总和 ≤ 180_000 | 总和 ≤ 410_000 | 同上 |
|
||||
|
||||
`tests/consult-evidence-lookup-20260927.test.ts` 的测试名与注释里仍写「70 s」，测试本身用自己的短时钟、不读常量，未改（描述性文字，记在此处）。
|
||||
|
||||
## 门禁（Node 22.14）
|
||||
|
||||
| 项 | 基线 `8912ba68` | 本分支 |
|
||||
|---|---|---|
|
||||
| tsc | — | 0 |
|
||||
| lint | — | 0 error / 126 warning |
|
||||
| npm test | 4,842 条，fail 24 | 4,846 条，fail 24，与基线逐条同名 |
|
||||
| next build | — | `/` Static |
|
||||
|
||||
## 下游时限核对（T5，只报告）
|
||||
|
||||
- Caddy（`deploy/Caddyfile*`）：反代未设置读写超时，默认不限响应时长。
|
||||
- Next standalone / Node：无自定义 server；Node `requestTimeout` 300 s 管的是接收请求体，不管流式响应。
|
||||
- Provider（undici 默认）：`bodyTimeout` 300 s 是两块数据之间的空闲时间；DeepSeek 推理内容也是流式下发（10-01 实测推理 5,200 token 在 30 s 内持续到达），正常不会空闲到 300 s。
|
||||
- 浏览器：无停滞检测；断线后服务端继续生成，客户端每 1.75 s 轮询状态。
|
||||
- 预留扣点：`/api/consult/status` 15 分钟租约后才懒取消，> 410 s。
|
||||
- 输出 token 上限 16,384 + 思考 8,192（`agent-generation-settings.ts`，与校正共用，未动）。
|
||||
@@ -398,3 +398,4 @@
|
||||
- 任务书里的行号会随代码漂移，定位以符号名为准。
|
||||
- **实现合入 `staging` 的同一次推送里，必须同时把状态板那一行改掉。** 2026-09-16 的对账发现 10 份早已合入的单仍写着「待领取 / 待验收」，会导致重复派活。
|
||||
| `TASK-consult-plain-answer-20261001.md` | `PROGRESS-consult-plain-answer-20261001.md` | 普通对话「还是废话」：09-17 四步开场形状（格局名 → 谁推谁修 → 扮演哪个象）逼出谜语，问父母时爸妈被揉成一段；产品授权推翻该形状，改成先答 + 按问题里的对象分段 + 人话自检 + 空宫不单独下结论 + 父母卡标 mother/father（BUG-1132～1134） | **已实现，待 Claude 验收**（fork 子代理直接执行）；未推 staging、未部署；模型对比为环境缺口，真机清单 `docs/testing/consult-plain-answer-20261001.md` | 分支 `codex/consult-plain-answer-20261001` |
|
||||
| `TASK-consult-answer-clock-20261001.md` | `PROGRESS-consult-answer-clock-20261001.md` | 普通对话答题时钟 70 s 容不下推理模型（v4-pro 77 s），无出生分钟路线只受 110 s 工具钟管；产品定 5 分钟防卡死（BUG-1142） | **已实现，待 Claude 验收**（直接执行） | 分支 `codex/consult-answer-clock-20261001` |
|
||||
|
||||
@@ -0,0 +1,31 @@
|
||||
# TASK · 普通对话答题时钟 70 s → 5 分钟（2026-10-01）
|
||||
|
||||
- 基线：`origin/staging` `8912ba68`
|
||||
- 分支 / worktree：`codex/consult-answer-clock-20261001` / `.worktrees/consult-answer-clock-20261001`
|
||||
- 模式：直接执行（产品 10-01 选择「直接执行」），Claude 派子代理实现、独立验收
|
||||
- BUG：BUG-1142
|
||||
|
||||
## 事故实证
|
||||
|
||||
10-01 真实模型对比（`docs/testing/consult-plain-answer-20261001-model-runs.md`）：`deepseek-v4-pro` 父母题 77 s 写完，超过 `CONSULTATION_ANSWER_TIMEOUT_MS`（`frontend/src/mastra/consultation-tools.ts`）的 70 s，线上会以 `answer_truncated` 截断。时限全链路排查：从浏览器、Caddy、Node 到 provider，能在写作中截断回答的只有本应用的两只钟（工具 110 s、答题 70 s），其余层都 ≥ 240 s 或不按总时长计。另发现无出生分钟路线（`usesPublicDailyGeneralAgent`）不切答题时钟，整段循环受 110 s 工具钟管。
|
||||
|
||||
## 决策记录
|
||||
|
||||
- 产品 2026-10-01：答题时长「可以延长、别定 70 秒这个上限」；在「3 / 5 / 10 分钟」中选 **5 分钟**，定位为防卡死，不是长度限制。推翻 BUG-1051 / 1053 记录中「110 + 70 = 180 s，产品接受三分钟」的口径，新上限 110 + 300 = 410 s。
|
||||
- 无出生分钟路线一并走答题时钟。
|
||||
- 工具阶段 110 s 与领域预算（2 个领域）不变。
|
||||
|
||||
## 硬红线
|
||||
|
||||
不动生时校正（`rectification-*`、`RECTIFICATION_RUN_BUDGET_MS`）、`agent-generation-settings.ts`、Caddy / deploy / workflow；截断仍记 `answer_truncated` 且不扣点。
|
||||
|
||||
## 任务与验收
|
||||
|
||||
| # | 内容 | 验收 |
|
||||
|---|---|---|
|
||||
| T1 | 答题时钟 300_000，注释写清原因 | 单测锁 300 s，工具钟仍 110 s |
|
||||
| T2 | 领域预算不随之变化 | 单测锁 65 s / 2 个领域，公式不读答题时钟 |
|
||||
| T3 | 无出生分钟路线从第一步起走答题时钟（`answerFromStart`） | 单测：工具钟到点不掐循环，答题钟到点仍掐；route 接线断言 |
|
||||
| T4 | `maxDuration` 240 → 480 | 单测：110 + 300 + 60 s 以内 |
|
||||
| T5 | 下游时限核对（只报告） | 见 PROGRESS |
|
||||
| T6 | 记录：BUG-1142、CHANGELOG、README、真机清单加一步 | — |
|
||||
@@ -11,3 +11,5 @@
|
||||
| 5 | 新开对话，打「你好」 | 只回一句寒暄 |
|
||||
| 6 | 用没填出生分钟的人物档案问事业 | 不点个人大运，不写月份 |
|
||||
| 7 | 第 1 步的回答如果明显被截断（最后一节没写完就停了） | 记下时间和模型名报给 Claude——这是取消字数上限后要盯的风险（见进度记录「长度副作用核对」） |
|
||||
| 8 | （BUG-1142，答题时钟改 5 分钟后）选 `deepseek-v4-pro` 新开对话，问「我和父母关系如何，他们怎么对待我」 | 回答完整写到「这周可以做的一件事」，不出现「回答未完成」；记下从发出到写完大约多少秒 |
|
||||
| 9 | 用没填出生分钟的人物档案、选 `deepseek-v4-pro` 问一个长问题（例如「完整讲讲我今年的事业和感情」） | 回答写完，不出现「回答未完成」 |
|
||||
|
||||
@@ -115,14 +115,14 @@ import { generateSessionTitle, shouldGenerateSessionTitle } from "@/lib/session-
|
||||
import { z } from "zod";
|
||||
|
||||
export const runtime = "nodejs";
|
||||
export const maxDuration = 240;
|
||||
export const maxDuration = 480;
|
||||
|
||||
// The step budget, the wall-clock budget and the domain cap all bound this same
|
||||
// run, so they are declared as one group in @/mastra/consultation-tools with the
|
||||
// reasoning that ties them together. maxDuration above is the ceiling they must
|
||||
// stay under: the tool phase (AGENT_TIMEOUT_MS) plus the answer phase's own
|
||||
// clock (CONSULTATION_ANSWER_TIMEOUT_MS), 180s, plus setup and settlement
|
||||
// (BUG-1051, BUG-1053). Self-hosted `node server.js` does not enforce it; it
|
||||
// clock (CONSULTATION_ANSWER_TIMEOUT_MS), 410s, plus setup and settlement
|
||||
// (BUG-1051, BUG-1053, TASK-consult-answer-clock-20261001). Self-hosted `node server.js` does not enforce it; it
|
||||
// documents the ceiling.
|
||||
|
||||
const chatRequestMetadataSchema = z.object({
|
||||
@@ -1103,10 +1103,13 @@ export async function POST(request: Request) {
|
||||
// that deadline until the calculation result is in hand, then on the answer
|
||||
// clock (CONSULTATION_ANSWER_TIMEOUT_MS), because its next step writes the
|
||||
// answer. Continuation and answer retries share the same answer clock.
|
||||
// The general / no-birth-minute agent has no tools, so its loop is the
|
||||
// answer from the first step and runs on the answer clock from the start.
|
||||
const runClock = createConsultationRunClock({
|
||||
toolPhaseMs: AGENT_TIMEOUT_MS,
|
||||
answerMs: CONSULTATION_ANSWER_TIMEOUT_MS,
|
||||
answerReady: () => state.consultationToolCompleted,
|
||||
answerFromStart: usesPublicDailyGeneralAgent(consultationMode, generalDailyContext),
|
||||
});
|
||||
const agentAbortSignal = runClock.toolSignal;
|
||||
const streamOptions = {
|
||||
|
||||
@@ -78,19 +78,24 @@ export const AGENT_TIMEOUT_MS = 110_000;
|
||||
* covers the rest of the loop plus any length continuation or empty-answer
|
||||
* retry.
|
||||
*
|
||||
* 70s is measured, not guessed: staging writer calls on the default model
|
||||
* produced 560-1240 output tokens in 5.7-13.6s end to end (about 90-100 tok/s
|
||||
* including first-token latency, PROGRESS-report-writer-failure-20260902). 70s
|
||||
* therefore holds about 6,300-7,000 output tokens, three to four times a
|
||||
* typical four-heading answer; at half that throughput it still holds about
|
||||
* 3,000 tokens. The answer step keeps provider thinking on (it is the loop's
|
||||
* step), so thinking tokens now come out of the same 70s; the loop's step
|
||||
* after the tool result used to think and write inside the 110s as well, and
|
||||
* its text was thrown away. The calculation must finish inside the 110s tool
|
||||
* phase, so the worst case is 110 + 70 = 180s, the three minutes the product
|
||||
* accepted. The route's maxDuration must stay above the sum.
|
||||
* 300s is a hang guard, not a length limit (product decision 2026-10-01,
|
||||
* TASK-consult-answer-clock-20261001). The old 70s was sized for a
|
||||
* non-reasoning writer (about 90-100 tok/s, PROGRESS-report-writer-failure-20260902).
|
||||
* The answer step keeps provider thinking on, so thinking tokens come out of
|
||||
* this clock too, and a reasoning model spends most of its time there: in the
|
||||
* 10-01 real-model run (docs/testing/consult-plain-answer-20261001-model-runs.md)
|
||||
* a parents answer on deepseek-v4-pro took 77s end to end and would have been
|
||||
* cut, while deepseek-flash took 30-41s with about 85% of its tokens in
|
||||
* thinking. The answer shape no longer has a word cap either, so the clock
|
||||
* only has to stop a provider that hangs. A run that does hit it still ends
|
||||
* as answer_truncated and is not charged (BUG-1051).
|
||||
*
|
||||
* The calculation must finish inside the 110s tool phase, so the worst case is
|
||||
* 110 + 300 = 410s; the route's maxDuration must stay above the sum. The
|
||||
* domain budget below is derived from the tool phase and its own reserve, not
|
||||
* from this clock, so raising it does not change how many domains run.
|
||||
*/
|
||||
export const CONSULTATION_ANSWER_TIMEOUT_MS = 70_000;
|
||||
export const CONSULTATION_ANSWER_TIMEOUT_MS = 300_000;
|
||||
|
||||
export type ConsultationRunClock = Readonly<{
|
||||
/** The tool phase's deadline. Tools and precompute run under it, unchanged. */
|
||||
@@ -114,11 +119,17 @@ export type ConsultationRunClock = Readonly<{
|
||||
* where the calculation settles just before the tool-phase timer fires but
|
||||
* the stream consumer has not seen the result yet: the loop is then handed to
|
||||
* the answer clock instead of being cut.
|
||||
*
|
||||
* `answerFromStart` is for a loop with no calculation to wait for (the
|
||||
* general / no-birth-minute agent has no tools): its first step already writes
|
||||
* the answer, so the loop runs on the answer clock from the start instead of
|
||||
* being cut by the tool phase's timer (TASK-consult-answer-clock-20261001).
|
||||
*/
|
||||
export function createConsultationRunClock(options: {
|
||||
toolPhaseMs?: number;
|
||||
answerMs?: number;
|
||||
answerReady?: () => boolean;
|
||||
answerFromStart?: boolean;
|
||||
} = {}): ConsultationRunClock {
|
||||
const toolPhaseMs = options.toolPhaseMs ?? AGENT_TIMEOUT_MS;
|
||||
const answerMs = options.answerMs ?? CONSULTATION_ANSWER_TIMEOUT_MS;
|
||||
@@ -141,6 +152,7 @@ export function createConsultationRunClock(options: {
|
||||
}
|
||||
loop.abort(toolSignal.reason);
|
||||
}, { once: true });
|
||||
if (options.answerFromStart) answerSignal();
|
||||
return Object.freeze({ toolSignal, loopSignal: loop.signal, answerSignal });
|
||||
}
|
||||
export const CONSULTATION_NATAL_CALC_TOOL_ID = "run-jyotish-consultation";
|
||||
|
||||
@@ -0,0 +1,55 @@
|
||||
// TASK-consult-answer-clock-20261001: the answer clock is a five-minute hang
|
||||
// guard, the general / no-birth-minute loop runs on it from the start, and the
|
||||
// domain budget does not move with it.
|
||||
import assert from "node:assert/strict";
|
||||
import { readFileSync } from "node:fs";
|
||||
import test from "node:test";
|
||||
|
||||
import {
|
||||
AGENT_TIMEOUT_MS,
|
||||
CONSULTATION_ANSWER_TIMEOUT_MS,
|
||||
CONSULTATION_DOMAIN_WALL_CLOCK_MS,
|
||||
MAX_CONSULTATION_DOMAINS,
|
||||
createConsultationRunClock,
|
||||
} from "../src/mastra/consultation-tools.ts";
|
||||
|
||||
const route = readFileSync(new URL("../src/app/api/consult/route.ts", import.meta.url), "utf8");
|
||||
const tools = readFileSync(new URL("../src/mastra/consultation-tools.ts", import.meta.url), "utf8");
|
||||
|
||||
test("the answer clock is five minutes and the tool phase stays 110s", () => {
|
||||
assert.equal(CONSULTATION_ANSWER_TIMEOUT_MS, 300_000);
|
||||
assert.equal(AGENT_TIMEOUT_MS, 110_000);
|
||||
const maxDuration = Number(route.match(/export const maxDuration = (\d+);/)?.[1]);
|
||||
assert.ok(AGENT_TIMEOUT_MS + CONSULTATION_ANSWER_TIMEOUT_MS + 60_000 <= maxDuration * 1000, "setup and settlement still fit");
|
||||
});
|
||||
|
||||
test("the domain budget is unchanged by the answer clock", () => {
|
||||
// Same numbers as before the change: 110s tool phase minus a 45s reserve,
|
||||
// 31s per domain, so two domains run.
|
||||
assert.equal(CONSULTATION_DOMAIN_WALL_CLOCK_MS, 65_000);
|
||||
assert.equal(MAX_CONSULTATION_DOMAINS, 2);
|
||||
const formula = tools.match(/export const CONSULTATION_DOMAIN_WALL_CLOCK_MS = [^;]+;/)?.[0] ?? "";
|
||||
assert.match(formula, /AGENT_TIMEOUT_MS - CONSULTATION_ANSWER_RESERVE_MS/);
|
||||
assert.doesNotMatch(formula, /CONSULTATION_ANSWER_TIMEOUT_MS/);
|
||||
});
|
||||
|
||||
test("a loop with no calculation runs on the answer clock from the start", async () => {
|
||||
const clock = createConsultationRunClock({ toolPhaseMs: 30, answerMs: 200, answerFromStart: true });
|
||||
await new Promise((resolve) => setTimeout(resolve, 60));
|
||||
assert.equal(clock.toolSignal.aborted, true, "the tool phase's timer still fires");
|
||||
assert.equal(clock.loopSignal.aborted, false, "but it does not cut a loop that is already writing the answer");
|
||||
await new Promise((resolve) => setTimeout(resolve, 220));
|
||||
assert.equal(clock.loopSignal.aborted, true, "the answer clock still bounds it");
|
||||
assert.equal(clock.answerSignal(), clock.answerSignal(), "retries and continuations share the same clock");
|
||||
});
|
||||
|
||||
test("the consult route starts the general / no-birth-minute loop on the answer clock", () => {
|
||||
assert.match(
|
||||
route,
|
||||
/const runClock = createConsultationRunClock\(\{[\s\S]*?answerFromStart: usesPublicDailyGeneralAgent\(consultationMode, generalDailyContext\),\s*\}\);/,
|
||||
);
|
||||
// That branch still streams on the shared loop signal and keeps its traces.
|
||||
const general = route.match(/if \(usesPublicDailyGeneralAgent\(consultationMode, generalDailyContext\)\) \{[\s\S]*?\n {4}\}\n/)?.[0] ?? "";
|
||||
assert.match(general, /await streamWithOverflowRetry\(agent\)/);
|
||||
assert.match(general, /onError: \(error\) => settleRun\(\s*cancel,/);
|
||||
});
|
||||
@@ -356,7 +356,11 @@ test("the consult route gives the answer phase its own clock inside maxDuration"
|
||||
// 原因: BUG-1053 删除 compose 流；「写回答有自己的时钟、不与工具阶段共用闸刀」这一性质不变
|
||||
const route = readFileSync(new URL("../src/app/api/consult/route.ts", import.meta.url), "utf8");
|
||||
const tools = readFileSync(new URL("../src/mastra/consultation-tools.ts", import.meta.url), "utf8");
|
||||
assert.match(tools, /export const CONSULTATION_ANSWER_TIMEOUT_MS = 70_000;/);
|
||||
// 原值: CONSULTATION_ANSWER_TIMEOUT_MS = 70_000；总上限 AGENT_TIMEOUT_MS + 答题时钟 <= 180_000
|
||||
// 新值: CONSULTATION_ANSWER_TIMEOUT_MS = 300_000；总上限 <= 410_000（110s 工具 + 300s 答题）
|
||||
// 原因: 产品 2026-10-01 决定答题时钟只防卡死、不限长度（TASK-consult-answer-clock-20261001）；
|
||||
// 推理模型 deepseek-v4-pro 父母题 77s 写完，70s 会截断
|
||||
assert.match(tools, /export const CONSULTATION_ANSWER_TIMEOUT_MS = 300_000;/);
|
||||
assert.match(route, /const runClock = createConsultationRunClock\(\{\s+toolPhaseMs: AGENT_TIMEOUT_MS,\s+answerMs: CONSULTATION_ANSWER_TIMEOUT_MS,/);
|
||||
assert.match(route, /const answerPhaseSignal = runClock\.answerSignal;/);
|
||||
assert.equal(route.match(/onAnswerPhase: startAnswerPhase,/g)?.length, 2, "natal and window hand the loop over");
|
||||
@@ -371,5 +375,5 @@ test("the consult route gives the answer phase its own clock inside maxDuration"
|
||||
assert.match(route, /const agentAbortSignal = runClock\.toolSignal;/);
|
||||
const maxDuration = Number(route.match(/export const maxDuration = (\d+);/)?.[1]);
|
||||
assert.ok(AGENT_TIMEOUT_MS + CONSULTATION_ANSWER_TIMEOUT_MS < maxDuration * 1000);
|
||||
assert.ok(AGENT_TIMEOUT_MS + CONSULTATION_ANSWER_TIMEOUT_MS <= 180_000, "product accepted about three minutes");
|
||||
assert.ok(AGENT_TIMEOUT_MS + CONSULTATION_ANSWER_TIMEOUT_MS <= 410_000, "product accepted a five-minute answer guard (2026-10-01)");
|
||||
});
|
||||
|
||||
@@ -446,5 +446,8 @@ test("the consult route has no separate compose stream and wires the single-pass
|
||||
assert.match(route, /const answerPhaseSignal = runClock\.answerSignal;/);
|
||||
const maxDuration = Number(route.match(/export const maxDuration = (\d+);/)?.[1]);
|
||||
assert.ok(AGENT_TIMEOUT_MS + CONSULTATION_ANSWER_TIMEOUT_MS < maxDuration * 1000);
|
||||
assert.ok(AGENT_TIMEOUT_MS + CONSULTATION_ANSWER_TIMEOUT_MS <= 180_000, "product accepted about three minutes");
|
||||
// 原值: AGENT_TIMEOUT_MS + CONSULTATION_ANSWER_TIMEOUT_MS <= 180_000（110s + 70s）
|
||||
// 新值: <= 410_000（110s + 300s）
|
||||
// 原因: 产品 2026-10-01 把答题时钟改成 5 分钟防卡死（TASK-consult-answer-clock-20261001）
|
||||
assert.ok(AGENT_TIMEOUT_MS + CONSULTATION_ANSWER_TIMEOUT_MS <= 410_000, "product accepted a five-minute answer guard (2026-10-01)");
|
||||
});
|
||||
|
||||
Reference in New Issue
Block a user