#557

fix: drop hardcoded gpt-5.4-mini from codex health check

Openim-whatTheSigmawants to merge intofrom

codex_live_ok() pinned gpt-5.4-mini, which ChatGPT-account auth cannot invoke:

ERROR: {"type":"error","status":400,"error":{"type":"invalid_request_error","message":"The 'gpt-5.4-mini' model is not supported when using Codex with a ChatGPT account."}}

This made 'make docker-login' always report 'codex: NOT logged in' and exit 1, even immediately after a successful login with a valid token in the volume. Omitting -m lets Codex use the account default.

Repro

  1. make docker-build
  2. make docker-login — complete the Codex OAuth flow in the browser (succeeds)
  3. Script reports codex: NOT logged in, exits 1

auth.json in the pentestgpt-codex volume is fully populated (auth_mode: chatgpt, valid tokens, fresh last_refresh) — the login worked. Only the health check fails.

Note

codex_live_ok() sends stderr to /dev/null, so the 400 is invisible and users see "NOT logged in" — pointing at auth/socat/volumes rather than the model. Surfacing the error on failure would make this class of issue much easier to diagnose. Happy to add that in this PR if wanted thank you!

Environment

Codex CLI v0.155.1, ChatGPT Plus account, Claude Pro, Docker on Kali.