Repository navigation
Mapper: a rate-limit 429 is labelled "Waiting for capacity"; tell rate limits and capacity apart #2865
Description
Since oclmap#86, Auto Match waits out every HTTP 429 and labels the wait "Waiting for capacity: results may take longer due to demand". That's right for the API's capacity limit (body error_code: "capacity_exceeded"), but not for an ordinary rate limit (the API's per-user request throttle), where the reason is the user's own request rate, not demand. This showed up in the overnight set test of oclapi2#922 + oclmap#86 (see the readout on OpenConceptLab/ocl_online#275), with the match rate limit lowered locally.
Acceptance criteria
- The Mapper tells the two 429s apart by the response body's
error_code:capacity_exceededkeeps today's wording; any other 429 says the user is sending requests too quickly and when it will retry. - Both still wait out
Retry-After, as now. Only the label and the run log's reason change. - Tests cover a capacity 429, a rate-limit 429 with
Retry-After, and one without. - Ships before oclmap#86 reaches production (that deploy is held until the capacity limit has run in shadow mode for a few days).
Notes
- The server side, so that a capacity refusal doesn't count against the user's rate limit, is part of enforcement: OpenConceptLab/ocl_online#340.
Activity
ocl-automation commentedon Oct 1, 2026
Automated Triage — 2026-10-01
Decision:
decision/ai-readyType:
type/bugSignals:
signal/quick-win,signal/low-risk,signal/small-scope,signal/well-specifiedThe Mapper already waits out every 429, so the fix is to branch on the body's
error_codeand change the label and run-log reason. That is a small change in the retry/label logic plus tests. The issue lists explicit acceptance criteria, including the three test cases. The only loose dependency is that it must ship before oclmap#86 reaches production, and the server-side change (ocl_online#340) is stated as separate. Risk is low because only wording changes and the wait behavior stays the same. The existing type/bug label fits because the user-facing message is wrong. The linked cross-references (ocl_issues#86 and #922) are unrelated to this change.Generated by ticket-triage pipeline v3. Model: anthropic/claude-sonnet-5.5 (openrouter).
snyaggarwal commentedon Oct 6, 2026
@paynejd raised PR OpenConceptLab/oclmap#93
OpenConceptLab/ocl_issues#2865 | review feedbacks
OpenConceptLab/ocl_issues#2865 | pt-BR: add the missing translations …
OpenConceptLab/ocl_issues#2865 | es, zh: add the missing translations…
Merge pull request #93 from OpenConceptLab/issues#2865
OpenConceptLab/ocl_issues#2865 | rate limit 429 and capacity limit 42…