The Codex readiness check and the second-model review pick the same model for a custom Codex provider (GST-3089) (#4491)

With a custom Codex provider, `gstack-codex-probe role-ready` and
`gstack-outside-voice run --runner codex-cli` now resolve the
plan-review model the same way, so a session no longer skips a review
that would have run.

Why: the check refused a `config.toml` naming `gpt-6-astra` while the
runner ran it, and a day-one eval session dropped the second-model
review in every phase.

Nothing to do: a custom provider naming the frontier catalog ID just
works; any other name now says how to opt in.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
This commit is contained in:
Garry Tanandcapy-ai[bot] authored and GitHub committed 2026-11-18 16:44:07 +00:00
1 parent b6ca54ec09
commit 45f879b534
8 files changed
+215 -20

No files matched your search

+10 -1
View File
@@ -106,7 +106,16 @@ an unknown or multi-provider host can show both choices.
Custom endpoints and partner platforms need an explicitly compatible model or
`host` mode. Gstack does not guess that public API IDs work with Bedrock, Vertex,
Foundry or a custom Codex provider. A model identifier passing syntax validation
Foundry or a custom Codex provider. One case counts as explicit: when a custom
Codex provider's own `config.toml` sets `model` to exactly the selected tier's
catalog ID (the catalog's OpenAI frontier model, by default), plan reviews use it,
because you already route that ID to your endpoint. Any other name there, such
as a smart-tier model or a deployment alias, is refused
(`custom_provider_model_not_tier`): it could be any model, so name it with the
tier setting or the environment override if you want it as your plan reviewer.
`gstack-codex-probe role-ready` and `gstack-outside-voice run --runner codex-cli`
(for the `autoplan` and `spec` gates) share this selection, so the readiness
check and the review never pick different models. A model identifier passing syntax validation
is not proof of account availability. See [provider recovery](troubleshooting.md#model-policy-provider).
Detection is conservative, not a replacement for the CLI's settings engine.
+6
View File
@@ -68,6 +68,12 @@ custom endpoint or partner platform. A public API catalog identifier is not
necessarily a valid deployment identifier there. Malformed native provider
configuration also prevents gstack from assuming the public endpoint.
A custom Codex provider whose `config.toml` `model` is exactly the selected
tier's catalog ID is used as is. The refusal prints
`CODEX_MODEL_POLICY: custom_provider_requires_model` when `config.toml` names
no model, and `CODEX_MODEL_POLICY: custom_provider_model_not_tier` when it names
a different one.
**Fix.** Supply a compatible model for that request, set the provider's gstack
environment override or tier-model setting, or choose
`gstack-config set plan_review_tier host`. Keep credentials in the provider's