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
Installer: add --personality and wizard prompt to set personality = "none"|"friendly"|"pragmatic" in ~/.codex/config.toml.
Config: migrate legacy model_personality to personality.
Changed
Installer: align --experimental / wizard “/experimental” toggles to Codex v0.102 (apps, multi-agents, plus platform-specific bubblewrap-sandbox and prevent-idle-sleep), while accepting sub-agents as a legacy alias.
Config: migrate legacy collab feature flags to canonical multi_agent.
Defaults: update the Safe profile to approval_policy = "untrusted" (Codex marks on-failure as deprecated).
Defaults: switch installer-written profiles and template to model = "gpt-5.3-codex" (Balanced uses model_reasoning_effort = "high").
Template: align comments and defaults to Codex CLI (codex-rs) v0.102 config keys.
Template/installer docs: refresh Codex schema/version references to v0.104.
Fixed
CLI: accept --profiles-scope selected.
Config: avoid reasoning.summary API errors on *-codex-spark models by forcing model_reasoning_summary = "none".
Installer: avoid silently ignoring --experimental feature toggles when no profiles are being written (fallback to root [features]).
Installer: validate generated config.toml as TOML before writing (avoid producing unparsable configs).
Config: remove legacy/removed sandbox flags (enable_experimental_windows_sandbox, experimental_windows_sandbox) during patching.