Bug Fixes
- Bottleneck 429 infinite wait (PR #495 by @xandr0s): On 429,
limiter.stop({ dropWaitingJobs: true })immediately drops all queued requests. Prevents infinite hangs for long-window rate limits (e.g., Codex). Limiter is deleted and recreated on next request. - Custom embedding models (#496):
POST /v1/embeddingsnow resolves custom embedding models from all provider_nodes (not just localhost). FixesUnknown embedding providererror for models likegoogle/gemini-embedding-001.
Issues Responded
| # | Title | Status |
|---|---|---|
| #496 | Custom embedding provider resolution | ✅ Fixed |
| #452 | Per-API-key request-count limits | 📋 Roadmap |
| #464 | Auto-issue API keys | ❓ Needs detail |
| #488 | Auto-update model lists | 📋 Roadmap |
What's Changed
Full Changelog: v2.8.6...v2.8.7