fix(session): retry HTTP 408 request timeouts - #47524
Open
Hotragn wants to merge 1 commit into
Open
Conversation
retryable() only bypasses the provider SDK's isRetryable flag for status >= 500, so a 408 the SDK did not mark retryable ended the turn and the user had to resend the prompt. OpenAI-compatible proxies that normalize an aborted upstream stream into 408 request_timeout hit this. 408 is the one 4xx that is transient in the same way a 5xx is: the request timed out before completing, so it is safe to send again. Treat it exactly like the existing 5xx case, same backoff and retry-after handling. Other 4xx statuses are unchanged. Closes anomalyco#39221
Contributor
|
Thanks for your contribution! This PR doesn't have a linked issue. All PRs must reference an existing issue. Please:
See CONTRIBUTING.md for details. |
3 tasks
6 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Issue for this PR
Closes #47525
Type of change
What does this PR do?
Replaces #39413, which was closed by the automated cleanup rather than by a fix. Rebased onto the reworked
retry.ts(message patterns, jitter, max retries) and re-verified against currentdev.retryable()only bypasses the provider SDK'sisRetryableflag forstatus >= 500, so an HTTP 408 that the SDK didn't mark retryable ends the turn and the user has to resend the prompt by hand. This shows up with OpenAI-compatible proxies that normalise an aborted upstream stream into408 request_timeout.Note on the newer
RETRYABLE_MESSAGE_PATTERNS: they don't cover this case. The timeout pattern requires a literal space (request timeout), while the payload in #39221 carriesrequest_timeoutwith an underscore. I checked the reportedmessageandresponseBodyagainst the current pattern list and both returnfalse, so a 408 still falls through today.408 Request Timeout is the one 4xx that is transient in the same way a 5xx is — the request never completed, so sending it again is the defined behaviour for that status. It takes exactly the same path as 5xx: same classification, same backoff, same
retry-afterhandling. Other 4xx statuses are unchanged.#39221 was closed as addressed by #39391, which fixed 408/409 in
packages/ai(the v2 path). This targets the v1packages/opencode/src/sessionpath, which is still wired in throughsession/processor.tsand still classifies these as non-retryable. Happy to close this if that path is being retired.How did you verify your code works?
retries 408 request timeout errorsintest/session/retry.test.ts, built from the payload in the issue. Ran the whole suite — 61/61 pass, includingdoes not retry 4xx errors when isRetryable is false(status 400), so 4xx isn't loosened generally.bun typecheckclean.Checklist