Skip to content

Stop labeling short xAI rate limits as quota exhaustion - #551

Open
TheGreatAxios wants to merge 2 commits into
mainfrom
cl-6935-stop-labeling-short-xai-rate-limits-as-quota-exhaustion
Open

Stop labeling short xAI rate limits as quota exhaustion#551
TheGreatAxios wants to merge 2 commits into
mainfrom
cl-6935-stop-labeling-short-xai-rate-limits-as-quota-exhaustion

Conversation

@TheGreatAxios

Copy link
Copy Markdown
Collaborator

Summary

  • Known xAI/Grok bare HTTP 429s remapped to retryable rate limits instead of quota_exhausted, so moderate Retry-After no longer aborts as a long-window quota
  • Transcript and retry policy stamp the live catalog provider id (including after /model) so remapping stays correct mid-session
  • Clear usage/quota body markers still surface as quota exhaustion; non-xAI bare 429 behavior unchanged

Verification

  • bun run check passes (lint, typecheck, build, 5278 tests)
  • Focused suites: gateway-error, error-message, retry-policy, stream-event-map (63 pass)

Fixes CL-6935

Bare HTTP 429s from known xAI/Grok providers remapped to retryable rate
limits so moderate Retry-After no longer aborts as a long-window quota.
The TUI stamps the live provider onto transcript formatting and the
retry policy follows /model switches via a providerId getter.
@linear-code

linear-code Bot commented Aug 23, 2026

Copy link
Copy Markdown

CL-6935

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant