Revert "fix: raise default LLM idle timeout to 240 seconds to reduce false timeouts on slow providers"#1172
Conversation
…false ti…" This reverts commit 502d9d5.
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d4539ba583
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| /** OpenClaw LLM idle timeout before an internal retry. */ | ||
| const LLM_IDLE_HINT_MS = 120_000; | ||
| /** Wait past one LLM idle window before declaring a hard no-response failure. */ | ||
| const NO_RESPONSE_SAFETY_TIMEOUT_MS = 130_000; |
There was a problem hiding this comment.
When a provider stays silent until OpenClaw's 120s LLM idle watchdog, the code above says OpenClaw starts an internal retry at that boundary, but this hard failure fires at 130s. That gives the retry only ~10 seconds to produce streamed output, so slow providers or first-token delays after the retry will be marked as “No response received”, clear sending, and stop the fallback history poll even though the runtime may still complete shortly afterward. Please keep the no-response timeout at least one full retry window past the idle hint.
Useful? React with 👍 / 👎.
1 participant
Reverts #1170