Summary

After the Nous live-model-fetch fix (#1174, v0.50.228), the live endpoint now returns a full catalog of Nous models. However, the openai-codex provider group — which has its own static model list in _PROVIDER_MODELS — has no automatic detection path via environment variables. It only appears if the user explicitly lists openai-codex under providers: in config.yaml.

Confirmed post-fix: a user with the Codex provider configured reported the group disappeared from the model picker entirely after pulling v0.50.228+.

Steps to reproduce

1. Set Nous as active provider (model.provider: nous in config.yaml)
2. Have OPENAI_API_KEY set in the environment (for Codex access via OpenAI)
3. Open the model picker — the "OpenAI Codex" group does not appear

Contrast: OPENAI_API_KEY triggers detection of the openai provider group but not openai-codex.

Root cause

get_available_models() in api/config.py maps specific env vars to providers:

if all_env.get("OPENAI_API_KEY"):
    detected_providers.add("openai")   # ← only "openai", not "openai-codex"

openai-codex has a full static model list in _PROVIDER_MODELS["openai-codex"] (9 models) and a display name in _PROVIDER_DISPLAY, but zero detection logic. It is entirely absent from the env-var detection block (lines ~1415–1448). The only path that adds it is the config.yaml providers: section reader (line ~1453), which requires the user to manually configure it.

Since both openai-codex and openai share OPENAI_API_KEY for authentication, there is no functional reason the Codex group should require separate manual config to appear.

Expected behavior

When OPENAI_API_KEY is set, the "OpenAI Codex" model group should appear in the picker automatically alongside the "OpenAI" group — the same key authenticates both.

Proposed fix

Add detection in the env-var block in api/config.py:

if all_env.get("OPENAI_API_KEY"):
    detected_providers.add("openai")
    detected_providers.add("openai-codex")   # ← add this

Single-line change, no logic side-effects. Users who don't want the Codex group can remove it via providers: section overrides (existing mechanism).

Files

• api/config.py line ~1424 (env-var detection block for OPENAI_API_KEY)
• api/config.py line ~625 (_PROVIDER_MODELS["openai-codex"] static list — already populated)

Impact

All users who authenticate via OPENAI_API_KEY and have Codex models available. The group never appears unless they manually add an openai-codex: section to config.yaml, which no onboarding flow or documentation currently suggests.