What's New
🔄 Empty Response Retry Mechanism
When API returns empty responses (common with high thinking_budget values), the proxy now automatically retries up to 2 times with exponential backoff before falling back.
Production tested results:
- Recovery rate: 88% (49 errors → 2)
- Backoff: 500ms → 1s → 2s
Thanks to @BrunoMarc for this contribution! 🙏
📦 Other Changes
- Updated
npxandnpm installcommands to use@latestfor better compatibility - Fixed flaky tests by making thinking requirements more lenient (models may skip thinking on obvious steps)
Full Changelog: v1.2.11...v1.2.12