v1.9.0

1.9.0 (2026-04-02)

Features

• add a "API" page to see details of the API keys (#269) (938c734)
• add stickysession selection box to select multiple sessions too be deleted (#286) (c64b860)
• api-keys: add per-key enforced service tier (#288) (cc851a5)
• api-keys: add self-service /v1/usage endpoint (#295) (652f600)
• balancer: add capacity-weighted routing for tier-aware load distribution (#297) (fa8eab4)

Bug Fixes

• balancer: trust usage data over stale runtime_reset for early quota resets (#289) (a269b37)
• chat: prevent duplicated tool-call arguments in chat completions (#287) (41ceb4f)
• deploy: restore Docker auto-migration, cache/rate-limiter fixes, Helm/K8s CI/CD (#274) (16391ae)
• docker: resolve distroless ARM64 build by detecting arch-specific lib paths (b21d4bd)
• prevent sticky session thrashing when all accounts exceed budget threshold (#279) (502db37)
• proxy: resolve k8s-era TC regressions (#290) (020784a)
• tests: stabilize proxy retry logging assertions (0f86737)

Performance Improvements

• usage: replace DISTINCT ON with lateral join in latest_by_account (#277) (8be87a6)

Documentation

• add Daltonganger as a contributor for code, and test (#298) (7f17d72)