fix: add GPT-5.6 Sol/Terra/Luna to openai-codex models - #58
Open
droopy-snoot wants to merge 1 commit into
Open
Conversation
Adds the three new OpenAI Codex model ids (gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna) to the hardcoded codex model list in generate-models.ts and to the generated catalog. The reasoning-capability helpers hardcode a gpt-5.2/5.3/5.4/5.5 prefix whitelist and did not recognize gpt-5.6: - supportsXhigh() in packages/ai/src/models.ts would not report xhigh thinking support for the new models. - clampReasoningEffort() in openai-codex-responses.ts would not clamp minimal effort to low for the new models, which the Codex API requires for this model family. Extended both to include gpt-5.6, and refreshed a stale doc comment in packages/agent/src/types.ts that already listed an outdated set of xhigh-capable models. Fixes JuliusBrussee#54
droopy-snoot
changed the title
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes #54.
Root cause
The
openai-codexmodel list inpackages/ai/scripts/generate-models.tsis a hardcoded static list (not sourced from models.dev), so adding new ids there is safe on its own — I confirmed the generator runs cleanly with the 3 new entries added.The actual bug is in two reasoning-capability helpers that hardcode a
gpt-5.2/gpt-5.3/gpt-5.4/gpt-5.5prefix whitelist and never got updated forgpt-5.6:supportsXhigh()inpackages/ai/src/models.ts— would silently report that the new gpt-5.6 models don't supportxhighthinking.clampReasoningEffort()inpackages/ai/src/providers/openai-codex-responses.ts— would fail to clamp"minimal"reasoning effort to"low"for the new models, which the Codex API requires for this model family (same requirement as gpt-5.2/5.3/5.4).Changes
gpt-5.6-sol,gpt-5.6-terra,gpt-5.6-lunato the codex model list ingenerate-models.tsand regenerated the corresponding entries inmodels.generated.ts.supportsXhigh()andclampReasoningEffort()to recognize thegpt-5.6prefix.packages/agent/src/types.tsthat already listed an outdated set of xhigh-capable models (didn't even include gpt-5.4/5.5), pointing it atsupportsXhigh()instead so it can't drift again.Testing
npx tsgo -p tsconfig.build.json --noEmitinpackages/ai— clean.npx vitest --runinpackages/ai— 173 passed, 609 skipped; the only failure (lazy-module-load.test.ts) is a pre-existing Windows/Node ESM URL-scheme issue in an environment probe unrelated to models, reproducible onmainbefore this change.