chore(deps): bump Claude Code to 2.1.204, Codex to 0.143.0#48
Conversation
Installer sha256 unchanged (verified against live claude.ai/install.sh). Image rebuilt and both CLI versions verified in berth:latest. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 19fccb64c2
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| "description": "Pinned AI CLIs for berth", | ||
| "dependencies": { | ||
| "@openai/codex": "0.142.5" | ||
| "@openai/codex": "0.143.0" |
There was a problem hiding this comment.
When Berth spawns Codex agents in its Ubuntu/Linux container, they rely on Codex tool calls to run shell commands, inspect repos, and apply patches. Upstream reports for Codex CLI 0.143.0 on Linux show those tool calls failing with unsupported call: exec_command / duplicated names such as exec_commandexec_command, and rolling back to 0.142.5 restores the same workflows; pinning this exact version means berth spawn codex can start but then cannot perform normal agent work. Please keep 0.142.5 or move to a release that fixes the tool-routing regression.
Useful? React with 👍 / 👎.
1 participant
Bumps the pinned AI CLIs to latest:
2.1.201→2.1.204(Dockerfile ARG; installer sha256 unchanged, re-verified against liveclaude.ai/install.sh)@openai/codex0.142.5→0.143.0(package.json + lockfile)Image rebuilt with
berth update --quick;claude --versionandcodex --versioninsideberth:latestconfirm 2.1.204 / 0.143.0.🤖 Generated with Claude Code