```密码本已下线```

```密码本已下线```
Codex Is Down

原始链接: https://github.com/openai/codex/issues/28756

有用户报告在使用 GPT 5.4 版本的 Codex App（版本号 26.611.61753）时，会出现持续性的“404 Not Found”错误。该问题通常发生在会话进行期间，具体表现为编排代理（orchestrator agent）在尝试生成子代理或处理约 5 万个 token 后出现中断。虽然点击“继续”后偶尔能恢复会话，但系统往往会耗尽重试次数（5/5）并导致崩溃。用户建议系统应具备更稳健的错误处理机制，特别是建议采用指数退避策略，即在 10 分钟内最多重试 20 次，以防止会话中断。值得注意的是，使用 GPT 5.5 的会话似乎未受影响，这表明该问题可能仅限于 5.4 模型环境。

抱歉。

EDIT: To everyone here, the solution was actually reasonably simple, just follow these instructions.

What version of the Codex App are you using (From “About Codex” dialog)?

26.611.61753

What subscription do you have?

Pro x20

What platform is your computer?

Darwin 25.3.0 arm64 arm

What issue are you seeing?

Repeatedly encountering in my GPT 5.4 xhigh session:

unexpected status 404 Not Found: Unknown error, url: https://chatgpt.com/backend-api/codex/responses, cf-ray: ...

I can send a message and get an update back, but then when the agent would think and continue I encountered this error 3 times in a row.

I have another session running with GPT 5.5 xhigh and it seems to be unaffected.

What steps can reproduce the bug?

I have a long running session going, it's had many compactions and plenty of steering. I've also been spinning up sub-agents for smaller tasks where this session is acting as an orchestrator. It failed shortly after steering it to spin up a sub-agent, not sure if that's related.

Session stopped with this error, it retried 5/5 times and gave up. This did not occur immediately after compaction, but it was shortly there after, maybe around 50k tokens.

Sending Continue did work, but it failed again after taking a turn. I tried this twice more and it does seem that for now the session is continuing on.

What is the expected behavior?

Don't die mid session.

Ideally we'd also have more error resistant retries with exponential back-off to something like 20 retries over 10 minutes.

Additional information

No response