v2.13.2 v3.8.36
cheaper-inference/OmniRouteJun 24, 2026 → Jun 25, 20261 day between releases
What's New
Adds Quota-Share combo strategy and Fusion routing strategy.
New Features
- Quota-Share combo strategy with Deficit Round Robin scheduling.
- Fusion routing strategy (fan-out + judge model).
- Headroom strategy for proactive saturation.
- Multi-window usage buckets with per-(key,model) caps.
- Proxy-pool support for Cloudflare Workers and Deno Deploy.
Improvements
- Extensive refactoring of chatCore and pricing modules.
- God-file decomposition.
- Universal proactive saturation via upstream token-usage response headers.
- Task-aware routing.
- Session stickiness preserving prompt-cache integrity.
Upgrade Notes
Automatic DB migration promotes existing `qtSd/*` combos, and new quota strategies require configuration updates for per-key/model caps and session stickiness.
The Two Releases
2.13.2 Desktop performance fix, and security fixes
Jun 24, 2026· 4 months ago
This release focuses on desktop performance improvements and security fixes, including UI enhancements for desktop users and fixes for Mac compatibility. It also removes unused dependencies from Docker images and includes upgrades to TypeScript and Gradle. Various bug fixes and dependency updates are also included.
Full release notesv3.8.36
Jun 25, 2026· 4 months ago
Introduces a major Quota-Share combo strategy and the Fusion routing strategy, alongside significant refactoring of the codebase.
Full release notes