[Bug] 2.77.0: "Use credits" never applies after weekly limit — credits.allowed copied from rate_limit.allowed #6571

Client or integration

Codex App

Area

Authentication and account pool

Summary

Regression in 2.77.0 (introduced by 6ef255f "fix(codex): require consent for spendable-credit headroom (carry #6466)").

After the main Codex account hits its weekly included-usage limit, enabling "Use credits" for the main account in Codex Auth has no effect. Every request is rejected locally with HTTP 429 CodexMainAccountCreditsOffError even though upstream reports spendable credits and the model as available.

Root cause: credits.allowed is populated from data.rate_limit.allowed, and hasSpendableCodexCredits() returns false whenever credits.allowed === false.

// src/codex/quota.ts:929 (v2.77.0, credits snapshot)
...typeof data.rate_limit?.allowed === "boolean" ? { allowed: data.rate_limit.allowed } : {},

// src/codex/quota-types.ts:45 (v2.77.0, hasSpendableCodexCredits)
if (age < 0 || age > CODEX_CREDITS_FRESHNESS_MS || credits.allowed === false || ...) return false;

rate_limit.allowed describes the included plan window only. It is always false once limit_reached: true, which is exactly the situation where credits are supposed to take over. So hasSpendableCodexCredits() can never return true at the moment it matters, mainCreditsHoldResetAt() always returns a reset time, and assertMainAccountPolicy() throws before the request reaches upstream (~10 ms 429s).

Expected: with creditCodexAccountIds containing __main__, and upstream reporting credits.has_credits: true, balance > 0, overage_limit_reached: false, spend_control.reached: false and model_usage[model].available: true, requests should be forwarded and billed to credits (as they were on 2.76.0).

Suggested fix: do not derive credits.allowed from rate_limit.allowed. If a refusal signal is needed, use model_usage[<model>].available / spend_control.reached / credits.overage_limit_reached instead.

Reproduction

1. Have the main ChatGPT (Pro) account at 100% of its weekly Codex limit, with a positive ChatGPT credits balance.
2. Codex Auth → main account → enable "Use credits" (creditCodexAccountIds now contains __main__).
3. Send any request from Codex App through the proxy (model gpt-6-astra).
4. Observe: 429 rate_limit_exceeded, failureStage: "headers-only", failureCause: "rate-limit", with the message below. Toggling the setting or waiting for a quota refresh doesn't help.

Timeline from dpkg.log / usage.jsonl on my machine:

• 12:06:03 upgrade open-codex 2.76.0 → 2.77.0
• before that: main-account openai requests returned 200
• 12:06:18 first CodexMainAccountCreditsOffError, and every main-account request since then is a 429

Version

2.77.0 (deb open-codex). Last known good: 2.76.0

Operating system

Ubuntu 26.04.1 LTS (x86_64)

Provider and model

openai (native Codex / ChatGPT main account), gpt-6-astra

Logs or error output

Proxy usage record (trimmed):

{
  "provider": "openai",
  "model": "gpt-6-astra",
  "admissionKind": "loopback",
  "status": 429,
  "durationMs": 10,
  "errorCode": "rate_limit_exceeded",
  "upstreamError": "Codex main account reached its usage limit, and spending ChatGPT credits is off or no fresh spendable balance is available. Choose another account, wait for the limit to reset, or allow the main account under \"Use credits\" in Codex Auth.",
  "routeDecision": { "routeKind": "native", "selected": { "provider": "openai", "model": "gpt-6-astra", "reason": "native-family" } },
  "failureStage": "headers-only",
  "failureCause": "rate-limit"
}

Upstream GET /backend-api/wham/usage at the same time (identifiers removed, balance rounded):

{
  "plan_type": "pro",
  "rate_limit": { "allowed": false, "limit_reached": true,
    "primary_window": { "used_percent": 100, "limit_window_seconds": 604800 } },
  "model_usage": { "gpt-6-astra": { "available": true, "available_at": null, "credits_would_enable": false } },
  "credits": { "has_credits": true, "unlimited": false, "overage_limit_reached": false, "balance": "~52000" },
  "spend_control": { "reached": false, "individual_limit": null },
  "rate_limit_reached_type": { "type": "rate_limit_reached", "details": "default" }
}

Resulting cached main quota in codex-quota-cache.json:

"credits": { "hasCredits": true, "unlimited": false, "overageLimitReached": false, "allowed": false, "balance": 52015.75 },
"weeklyPercent": 100

Redacted configuration

{
  "codexMainAccountHardLock": false,
  "creditCodexAccountIds": ["__main__", "<account-2>", "<account-3>"]
}

Checks

• I searched existing issues and documentation.
• I removed secrets, tokens, account details, request credentials, and personal data.