Intel / repo

codex rust-v0.167.0 — 0.167.0

Source: github.com
Author: github-actions[bot]Top Viber
Date: Published Nov 6, 2026

Context

Codex 0.167.0, a release of the open-source command-line version of OpenAI's coding , stops saved chats from disappearing when a user changes which model provider Codex sends its requests to. According to pull request #58202, switching providers could hide saved chats from the resume list, which showed only chats from the current provider, or stop them from opening when the server rejected the resume over a provider mismatch, missing configuration or a policy conflict. The resume picker, lookup by name and resume --last now include chats from every provider. A chat that still cannot resume opens as read-only history with an explanation, and pressing c continues it in a new chat by preparing an unsent prompt that refers to the saved conversation, for the user to edit and send.

Switching providers is becoming a more ordinary step as Codex adds ways to run on models outside OpenAI's own service. Version 0.165.0 added a built-in provider for Google Cloud's Vertex AI, and the 0.167.0 release notes add an optional setup wizard for it, turned on with the gcp_setup_wizard feature flag, which pull request #57857 leaves off by default. The wizard asks for a Google Cloud project, an optional location and an optional credential file, and falls back to Application Default Credentials, the credentials Google Cloud tools find on the machine, when no file is given. The fix keeps earlier work visible and readable across such a switch, but a chat that cannot resume is not resumed in place on the new provider. Pressing c starts a new conversation instead, and neither the pull request nor the release notes say how much of the old transcript the new chat's model receives.

Key takeaways · AI-distilled

If a saved chat cannot resume after a model provider switch, Codex 0.167.0 lets you read its transcript and prepare an editable prompt to continue in a new chat.
Network approval prompts now close when their requests end, and concurrent requests to the same endpoint prompt one at a time.
Automatic now accounts for retained output from failed responses and from responses with missing usage data.
The app-server README now documents prompt-user command rules that require fresh human approval, alongside managed network constraints under [permissions.network].
Linux musl builds use fewer allocator arenas to reduce idle memory after conversations unload.

Terms in this piece · Glossary

AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
context compaction — Summarizing an agent's earlier conversation to free room in the context window so a long session can keep going.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.

Read the source github.com