mirror of
https://github.com/BerriAI/litellm.git
synced 2026-11-14 09:03:16 +00:00
The io-token check reserves, reconciles and refunds its global_router itpm and otpm counters through the native router_usage store, so ModelRateLimitingCheck keeps no DualCache. A reservation stashes its counter's fields in the request metadata instead of a Redis key, so nothing parses a key back into ids. The Redis keys and values are unchanged, pinned by contract tests on both sides. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| stores | ||
| __init__.py | ||