Skip to content

v0.4.0

Choose a tag to compare

View all tags
@github-actions github-actions released this 17 Jul 16:57
· 818 commits to main since this release
v0.4.0
6559eb4

Changes

  • Added an experimental, default-off Responses Lite mode for GPT-5.6 Sol, Luna, and Terra. Enable it with responsesLite or CORTEXKIT_OPENAI_AUTH_RESPONSES_LITE.
  • Added per-session routing state so concurrent sessions can show and retain different active accounts without overwriting each other.
  • Made quota windows length-driven, including dynamic labels, reset credits, malformed-frame protection, and policy-aware Retry-After calculation.
  • Restored canonical models.dev pricing when costZeroing.enabled is false. The default subscription-backed zero-cost behavior is unchanged.
  • Removed legacy -pro model handling in favor of OpenCode's native max reasoning variant.

Reliability and compatibility

  • Preserved full Request method, body, signal, and header semantics through the auth fetch override.
  • Sanitized internal session and Responses Lite markers across HTTP, WebSocket fallback, and cache keep-warm replays.
  • Added fenced cross-process sidebar updates and made write failures observable without breaking the write queue.
  • Updated AI SDK, OpenCode plugin types, Pi 0.80, OpenTUI, Solid, Biome, Lefthook, and pinned workflow actions.

Thanks to @chrisolszewski for the initial Responses Lite implementation, and @iceteaSA for dynamic quota windows, per-session routing, and models.dev cost restoration.

Full changelog: v0.3.4...v0.4.0

Contributors

chrisolszewski and iceteaSA
Assets 2
Loading