v0.4.71

What's Changed

New Features

• Codex now defaults to GPT-5.6 — gpt-5.6-sol by default, with gpt-5.6-terra and gpt-5.6-luna selectable (372K-token context), matching the latest Codex CLI.
• Native max / ultra reasoning efforts for GPT-5.6. The effort picker is now per-model — sol/terra expose max and ultra, luna exposes max — and the chosen effort is sent to Codex verbatim instead of being silently clamped down to xhigh.

Internal

• Reworked effort dispatch so Codex effort tokens flow through as-is, while Claude keeps ultra as its display alias for max + workflow orchestration. Removed the stale legacy-driver max→xhigh clamp so the app-server receives the real reasoning level.
• Bundled @pikiloom/kernel bumped to 0.3.13.

Full Changelog: v0.4.70...v0.4.71