fix(quota): parse absolute ISO datetime reset timestamps in weekly quota fallback (#11353)

Merged via consolidated batch validation. Fixes GLM/Z.AI weekly quota fallback: parseDayGranularityResetMs only recognized 'reset in N days', dropping the real multi-day cooldown when upstream returns a full absolute ISO datetime. Own repro test passes.

open-sse/services/quotaResetParsing.ts

@@ -26,19 +26,121 @@ export function shouldPreserveQuotaSignals(
 }
 
 /**
- * Parse a day-granularity quota reset countdown ("Your quota will reset in
- * 3 days.", "Resets in 13 days") out of an upstream 429 body.
+ * Parse a day-granularity quota reset countdown (\"Your quota will reset in
+ * 3 days.\", \"Resets in 13 days\") out of an upstream 429 body.
  *
  * Companion to the Xh/Ym/Zs countdown parsing already handled inline by
  * `parseRetryFromErrorText` — none of those patterns match when the upstream
  * expresses the reset window in whole days rather than hours/minutes/seconds,
  * so a multi-day quota reset previously parsed to `null` and fell back to the
  * engine's ~seconds-scale default cooldown.
+ *
+ * Delegates to `parseIsoDateTimeResetMs` (absolute \"reset at YYYY-MM-DD HH:MM:SS\")
+ * and then `parseMonthDayResetMs` (year-less \"reset at MM-DD HH:MM:SS UTC\") so
+ * every absolute-reset shape an upstream uses resolves to the real wait.
  */
-export function parseDayGranularityResetMs(msg: string, maxMs: number): number | null {
+export function parseDayGranularityResetMs(
+  msg: string,
+  maxMs: number,
+  nowMs: number = Date.now()
+): number | null {
   const dayMatch = /reset(?:s)?\s+in\s+(\d+)\s*day(?:s)?/i.exec(msg);
-  if (!dayMatch) return null;
-  const days = Number.parseInt(dayMatch[1], 10);
-  if (!Number.isFinite(days) || days <= 0) return null;
-  return Math.min(days * 24 * 3600 * 1000, maxMs);
+  if (dayMatch) {
+    const days = Number.parseInt(dayMatch[1], 10);
+    if (Number.isFinite(days) && days > 0) {
+      return Math.min(days * 24 * 3600 * 1000, maxMs);
+    }
+  }
+  const isoMs = parseIsoDateTimeResetMs(msg, maxMs, nowMs);
+  if (isoMs !== null) return isoMs;
+  return parseMonthDayResetMs(msg, maxMs, nowMs);
+}
+
+/**
+ * Z.AI (GLM) reports an exhausted weekly/monthly cap with a FULL absolute
+ * datetime rather than a countdown:
+ *
+ *   \"[1310][Weekly/Monthly Limit Exhausted. … Your limit will reset at
+ *    2026-08-29 21:01:21]\"
+ *
+ * `parseRetryFromErrorText` (accountFallback.ts) has an equivalent ISO matcher,
+ * but `buildWeeklyQuotaFallback` never reaches it: it calls
+ * `parseDayGranularityResetMs` directly, and neither the \"reset in N days\" nor
+ * the year-less MM-DD parser matched this shape. The weekly fallback therefore
+ * fell back to WEEKLY_QUOTA_COOLDOWN_MS (24h) and the connection was dispatched
+ * again — into a real upstream 429 — every day until the true reset ~6 days out.
+ *
+ * The datetime may use a `T` or a space separator, and may carry `Z` or a
+ * `±HH:MM` offset. A NAIVE datetime (no zone) is interpreted as UTC: Z.AI
+ * reports in UTC, and treating it as local time would shift the cooldown by the
+ * host offset. Returns null when the instant is not in the future.
+ */
+export function parseIsoDateTimeResetMs(
+  msg: string,
+  maxMs: number,
+  nowMs: number = Date.now()
+): number | null {
+  const match =
+    /\b(?:try again at|wait until|reset(?:s)?\s+at|available at|retry after)\s+(\d{4}-\d{2}-\d{2}[Tt ]\d{2}:\d{2}(?::\d{2})?(?:\.\d+)?)\s*(Z|[+-]\d{2}:?\d{2})?/i.exec(
+      msg
+    );
+  if (!match) return null;
+  const stamp = match[1].replace(/[Tt ]/, "T");
+  // No zone in the body → UTC (see doc comment). Normalize \"+0200\" to \"+02:00\":
+  // the bare-offset form is not part of the ES Date.parse grammar.
+  const rawZone = match[2] ? match[2].toUpperCase() : "Z";
+  const zone = /^[+-]\d{4}$/.test(rawZone)
+    ? `${rawZone.slice(0, 3)}:${rawZone.slice(3)}`
+    : rawZone;
+  const resetMs = Date.parse(`${stamp}${zone}`);
+  if (!Number.isFinite(resetMs)) return null;
+  const waitMs = resetMs - nowMs;
+  if (waitMs <= 0) return null;
+  return Math.min(waitMs, maxMs);
+}
+
+/**
+ * Qwen token-plan (and similar apikey providers) report the weekly reset as
+ * \"The quota will reset at 08-29 15:29:00 UTC\" without a year. Treat that as
+ * the next occurrence of MM-DD HH:MM[:SS] UTC; if the date already passed this
+ * year, roll to next year. Returns null when the parsed instant is not in the
+ * future or the wait would exceed maxMs.
+ */
+export function parseMonthDayResetMs(
+  msg: string,
+  maxMs: number,
+  nowMs: number = Date.now()
+): number | null {
+  const match =
+    /reset(?:s)?\s+at\s+(\d{2})-(\d{2})\s+(\d{2}):(\d{2})(?::(\d{2}))?\s*(?:UTC|Z)?/i.exec(
+      msg
+    );
+  if (!match) return null;
+  const month = Number.parseInt(match[1], 10);
+  const day = Number.parseInt(match[2], 10);
+  const hour = Number.parseInt(match[3], 10);
+  const minute = Number.parseInt(match[4], 10);
+  const second = match[5] ? Number.parseInt(match[5], 10) : 0;
+  if (
+    month < 1 ||
+    month > 12 ||
+    day < 1 ||
+    day > 31 ||
+    hour > 23 ||
+    minute > 59 ||
+    second > 59
+  ) {
+    return null;
+  }
+  const now = new Date(nowMs);
+  let year = now.getUTCFullYear();
+  let resetMs = Date.UTC(year, month - 1, day, hour, minute, second);
+  if (!Number.isFinite(resetMs)) return null;
+  if (resetMs <= nowMs) {
+    year += 1;
+    resetMs = Date.UTC(year, month - 1, day, hour, minute, second);
+  }
+  const waitMs = resetMs - nowMs;
+  if (!Number.isFinite(waitMs) || waitMs <= 0) return null;
+  return Math.min(waitMs, maxMs);
 }

open-sse/services/quotaTextCooldowns.ts

@@ -11,6 +11,7 @@
  */
 
 import { RateLimitReason } from "../config/constants.ts";
+import { parseDayGranularityResetMs } from "./quotaResetParsing.ts";
 
 type RateLimitReasonValue = (typeof RateLimitReason)[keyof typeof RateLimitReason];
 
@@ -97,16 +98,29 @@ export function isWeeklyUsageLimitText(lower: string): boolean {
   return (
     lower.includes("weekly usage limit") ||
     lower.includes("weekly limit reached") ||
-    lower.includes("reached your weekly")
+    lower.includes("reached your weekly") ||
+    lower.includes("1-week quota") ||
+    lower.includes("week quota") ||
+    lower.includes("weekly/monthly limit") ||
+    (lower.includes("weekly") && lower.includes("quota") && lower.includes("exhaust"))
   );
 }
 
+const MAX_WEEKLY_QUOTA_COOLDOWN_MS = 30 * 24 * 60 * 60 * 1000;
+
 export function buildWeeklyQuotaFallback(errorStr: string): QuotaTextFallback | null {
   if (!isWeeklyUsageLimitText(errorStr.toLowerCase())) return null;
+  const parsedResetMs = parseDayGranularityResetMs(errorStr, MAX_WEEKLY_QUOTA_COOLDOWN_MS);
+  const cooldownMs =
+    typeof parsedResetMs === "number" && parsedResetMs > 0
+      ? parsedResetMs
+      : WEEKLY_QUOTA_COOLDOWN_MS;
   return {
     shouldFallback: true,
-    cooldownMs: WEEKLY_QUOTA_COOLDOWN_MS,
+    cooldownMs,
     reason: RateLimitReason.QUOTA_EXHAUSTED,
+    usedUpstreamRetryHint: typeof parsedResetMs === "number" && parsedResetMs > 0,
+    quotaResetHintMs: typeof parsedResetMs === "number" && parsedResetMs > 0 ? parsedResetMs : undefined,
   };
 }
 

tests/unit/repro-glm-iso-reset-24h-cap.test.ts

@@ -0,0 +1,135 @@
+/**
+ * Regression: Z.AI (GLM) weekly quota was capped at a 24h cooldown instead of
+ * the real ~6-day reset the upstream reported.
+ *
+ * Body from production (connection zai/glm-5.3):
+ *   "[1310][Weekly/Monthly Limit Exhausted. Your limit will reset at 2026-08-29 21:01:21]"
+ *
+ * looksLikeQuotaExhausted() and isWeeklyUsageLimitText() both matched, so the
+ * weekly branch was taken — but buildWeeklyQuotaFallback() calls
+ * parseDayGranularityResetMs() FIRST and that only knew "reset in N days" and
+ * the year-less "reset at MM-DD HH:MM:SS UTC" shape (#qwen). A full ISO
+ * datetime parsed to null, so the weekly fallback used its
+ * WEEKLY_QUOTA_COOLDOWN_MS default of 24h. The ISO matcher that DOES handle
+ * this shape lives in parseRetryFromErrorText() and is never reached from the
+ * weekly branch.
+ *
+ * Result: rate_limited_until was written 24h out instead of the true reset,
+ * and the connection was dispatched into a real upstream 429 every day for
+ * the rest of the week.
+ */
+
+import { describe, it } from "node:test";
+import assert from "node:assert/strict";
+
+import { looksLikeQuotaExhausted } from "../../src/shared/utils/classify429.ts";
+import {
+  isWeeklyUsageLimitText,
+  buildWeeklyQuotaFallback,
+} from "../../open-sse/services/quotaTextCooldowns.ts";
+import {
+  parseDayGranularityResetMs,
+  parseIsoDateTimeResetMs,
+  parseMonthDayResetMs,
+  shouldPreserveQuotaSignals,
+} from "../../open-sse/services/quotaResetParsing.ts";
+import { RateLimitReason } from "../../open-sse/config/constants.ts";
+
+const GLM_BODY =
+  "[1310][Weekly/Monthly Limit Exhausted. Your current plan has run out of its weekly/monthly quota. " +
+  "Your limit will reset at 2026-08-29 21:01:21]";
+const MAX_MS = 30 * 24 * 60 * 60 * 1000; // MAX_WEEKLY_QUOTA_COOLDOWN_MS
+const DAY_MS = 24 * 60 * 60 * 1000;
+const NOW = Date.UTC(2026, 7, 23, 20, 30, 56); // 2026-08-23 20:30:56 UTC
+const RESET = Date.UTC(2026, 7, 29, 21, 1, 21); // 2026-08-29 21:01:21 UTC
+
+describe("Z.AI GLM weekly quota — absolute ISO reset", () => {
+  it("looksLikeQuotaExhausted matches the [1310] weekly/monthly body", () => {
+    assert.equal(looksLikeQuotaExhausted(GLM_BODY), true);
+  });
+
+  it("shouldPreserveQuotaSignals is true for zai with this body", () => {
+    assert.equal(shouldPreserveQuotaSignals("zai", GLM_BODY), true);
+  });
+
+  it("isWeeklyUsageLimitText matches weekly/monthly limit wording", () => {
+    assert.equal(isWeeklyUsageLimitText(GLM_BODY.toLowerCase()), true);
+  });
+
+  it("parseIsoDateTimeResetMs reads a space-separated naive datetime as UTC", () => {
+    assert.equal(parseIsoDateTimeResetMs(GLM_BODY, MAX_MS, NOW), RESET - NOW);
+  });
+
+  it("parseIsoDateTimeResetMs accepts the T separator and an explicit Z", () => {
+    assert.equal(
+      parseIsoDateTimeResetMs("reset at 2026-08-29T21:01:21Z", MAX_MS, NOW),
+      RESET - NOW
+    );
+  });
+
+  it("parseIsoDateTimeResetMs honours an explicit UTC offset", () => {
+    // 23:01:21+02:00 is the same instant as 21:01:21Z.
+    assert.equal(
+      parseIsoDateTimeResetMs("reset at 2026-08-29 23:01:21+02:00", MAX_MS, NOW),
+      RESET - NOW
+    );
+    assert.equal(
+      parseIsoDateTimeResetMs("reset at 2026-08-29 23:01:21+0200", MAX_MS, NOW),
+      RESET - NOW
+    );
+  });
+
+  it("parseIsoDateTimeResetMs returns null for a past reset and caps at maxMs", () => {
+    assert.equal(parseIsoDateTimeResetMs("reset at 2026-08-22 10:00:00", MAX_MS, NOW), null);
+    assert.equal(parseIsoDateTimeResetMs("reset at 2027-08-29 21:01:21", MAX_MS, NOW), MAX_MS);
+  });
+
+  it("parseDayGranularityResetMs returns the real reset, not the 24h cap", () => {
+    const waitMs = parseDayGranularityResetMs(GLM_BODY, MAX_MS, NOW);
+    assert.equal(waitMs, RESET - NOW);
+    assert.ok(waitMs! > DAY_MS, `expected more than 24h, got ${waitMs}`);
+  });
+
+  it("keeps the Qwen year-less MM-DD parser working", () => {
+    const qwenBody =
+      "Your token-plan 1-week quota has been exhausted. The quota will reset at 08-29 15:29:00 UTC.";
+    const expected = Date.UTC(2026, 7, 29, 15, 29, 0) - NOW;
+    assert.equal(parseMonthDayResetMs(qwenBody, MAX_MS, NOW), expected);
+    assert.equal(parseDayGranularityResetMs(qwenBody, MAX_MS, NOW), expected);
+  });
+
+  it("keeps the 'reset in N days' parser winning over the ISO branch", () => {
+    assert.equal(parseDayGranularityResetMs("quota will reset in 3 days", MAX_MS, NOW), 3 * DAY_MS);
+  });
+
+  it("buildWeeklyQuotaFallback uses the parsed ISO reset, not the 24h default", () => {
+    const result = buildWeeklyQuotaFallback(GLM_BODY);
+    assert.ok(result);
+    assert.equal(result!.reason, RateLimitReason.QUOTA_EXHAUSTED);
+    assert.equal(result!.usedUpstreamRetryHint, true);
+    assert.ok(
+      result!.cooldownMs > 5 * DAY_MS,
+      `expected a multi-day cooldown, got ${result!.cooldownMs}`
+    );
+    assert.ok(result!.cooldownMs <= MAX_MS);
+    assert.ok(result!.cooldownMs !== DAY_MS, "must not fall back to WEEKLY_QUOTA_COOLDOWN_MS (24h)");
+  });
+
+  it("checkFallbackError classifies the GLM 429 as QUOTA_EXHAUSTED with the real wait", async () => {
+    const { checkFallbackError, parseRetryFromErrorText } = await import(
+      "../../open-sse/services/accountFallback.ts"
+    );
+
+    const parsed = parseRetryFromErrorText(GLM_BODY);
+    assert.ok(parsed && parsed > 5 * DAY_MS, `parsed reset was ${parsed}`);
+
+    const out = checkFallbackError(429, GLM_BODY, 0, "glm-5.3", "zai", null, null, null);
+    assert.equal(out.shouldFallback, true);
+    assert.equal(out.reason, RateLimitReason.QUOTA_EXHAUSTED);
+    assert.ok(
+      (out.cooldownMs ?? 0) > 5 * DAY_MS,
+      `expected a multi-day cooldown, got ${out.cooldownMs}`
+    );
+    assert.ok((out.cooldownMs ?? 0) !== DAY_MS, "must not land on the 24h weekly default");
+  });
+});