You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
🐛 Patch Changes
2cdaee6: Fix auto routing resolving to a provider the agent never connected. A stale legacy auto-assigned (or promoted fallback) route now reuses the gateway's provider-key lookup before it becomes primary, so an unconfigured provider is skipped without adding a separate model-discovery query. When nothing routable remains, the request returns the neutral M101 "no providers configured" error. The proxy also treats the resolver's fallback chain as definitive so a fallback promoted to primary is not retried as its own fallback.
9cf2571: Bump modelparams to 0.0.13 to pick up the new xAI Grok model parameter specs, including grok-4.5 and subscription entries for the Grok subscription models.
6f1345c: Scope Auto-fix parameter repairs by provider authentication type.
d3d44e3: Add Kimi K3 (cline-pass/kimi-k3) to the ClinePass subscription's known models list.
d8fb0b2: Fix Auto-fix attempt recording so no-patch consultations remain plain provider failures and failed patched retries keep their own outcome.
84b2112: The password reset page now detects when no email provider is configured and shows a clear notice (pointing to the authenticated Change Password flow) instead of silently pretending a reset link was sent. Self-hosted installs without an EMAIL_PROVIDER no longer dead-end here.
6519ea5: Add opt-in Sentry error monitoring. It stays disabled without a SENTRY_DSN; request contents, user data, tracing, and profiling remain off.
189a6eb: Sync the model parameter catalog to modelparams@0.0.15. Adds 24 new model routes (including a new xiaomi provider) with no schema changes — the param types, groups, and auth types are unchanged, so the bump is purely additive.
89c1757: Preserve xAI reasoning effort when forwarding Chat Completions requests.