umputun / ralphex / 26263058647

83%

master: 83%

Build Type

Pull #350

github

Committed by umputun

Commit Message

fix(config): allow clearing codex_model and codex_reasoning_effort

embedded defaults ship codex_model=gpt-5.5 and codex_reasoning_effort=xhigh
as active values, but mergeFrom propagated only non-empty values, so users
could not opt out. docs claimed "comment the line out to inherit from
~/.codex/config.toml" — which never worked: a commented-out line is
indistinguishable from an absent one, so the embedded default survived.

add CodexModelSet/CodexReasoningEffortSet merge sentinels (mirroring
CodexSandboxSet) so an explicit empty value clears the embedded default and
codex inherits from ~/.codex/config.toml; commenting the line out keeps the
default. docs corrected to describe the explicit-empty escape hatch.

also fix the idle_timeout comment in defaults/config: it claimed the timeout
applies to external codex review, which the code intentionally excludes in
default-claude mode.

Pull Request Pull Request #350: Add first-class codex executor mode

Coverage Stats

611 of 679 new or added lines in 9 files covered. (89.99%)

17 existing lines in 4 files now uncovered.

7393 of 8893 relevant lines covered (83.13%)

231.57 hits per line