fix(runtime/codex): ship sandbox/network/review defaults in config.toml#629
Merged
Conversation
Codex runs through the local Responses shim 404'd on the auto-review reviewer (default model "codex-auto-review" is not exposed by the chat backend), and the default read-only sandbox blocked network fetches (e.g. arXiv), surfacing as approval-review rejections. _prepare_codex_home now writes these defaults into the generated config.toml: - review_model pinned to the configured model (fixes the 404) - approval_policy=never + sandbox_mode=danger-full-access (no approval round-trip; full fs/command/network access) - network_access=true under [sandbox_workspace_write] - disable_response_storage (chat-backed shim has no response store) - model_reasoning_effort=medium, personality=pragmatic
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
Running the Codex runtime through veadk's local Responses shim hit two issues:
InvalidEndpointOrModel.NotFound: codex-auto-review does not exist— the auto-review reviewer defaults to a model (codex-auto-review) the chat backend doesn't expose; the shim forwards it verbatim → 404.The automatic approval review rejected the request to fetch from arXiv— the default read-only sandbox blocks network access, escalating to an approval round-trip.Change
_prepare_codex_homenow writes these defaults into the generatedconfig.toml:review_modelapproval_policyneversandbox_modedanger-full-access[sandbox_workspace_write] network_accesstruedisable_response_storagetruemodel_reasoning_effortmediumpersonalitypragmaticNotes
review_modelis pinned to the configured model (not a hard-coded name), so it always matches what the backend accepts.network_accessis a boolean under[sandbox_workspace_write](redundant under full-access, correct for narrower modes).approval_policy=never+danger-full-accessis now the default for every Codex run — broad, but required to unblock network fetches. Can be made env-overridable if preferred.Verification
config.tomlparses withtomllib(valid TOML; scalar keys precede tables).