Skip to content

fix(opencode): allow integer GPT versions in Codex model filter - #47384

Merged
rekram1-node merged 1 commit into
devfrom
codex-major-versions
Sep 4, 2026
Merged

fix(opencode): allow integer GPT versions in Codex model filter#47384
rekram1-node merged 1 commit into
devfrom
codex-major-versions

Conversation

@rekram1-node

Copy link
Copy Markdown
Collaborator

Issue for this PR

No linked issue. gpt-6-astra is incorrectly hidden when using ChatGPT OAuth.

Type of change

  • Bug fix
  • New feature
  • Refactor / code improvement
  • Documentation

What does this PR do?

Makes the decimal portion of the GPT version optional in the Codex model filter. Previously, gpt-6-astra failed the regex and was removed from the model list. Integer versions now reach the existing version check; explicit exclusions and pro-mode filtering are unchanged. V2 is untouched.

Adds regression coverage for integer and decimal versions, older models, and existing exclusions.

How did you verify your code works?

  • Confirmed the new integer-version cases fail before the fix.
  • bun test test/plugin/codex.test.ts --timeout 30000: 33 passed.
  • bun typecheck: passed in packages/opencode.
  • Prettier and git diff --check: passed.

Screenshots / recordings

Not applicable.

Checklist

  • I have tested my changes locally
  • I have not included unrelated changes in this PR

@rekram1-node
rekram1-node merged commit 500c46e into dev Sep 4, 2026
10 checks passed
@rekram1-node
rekram1-node deleted the codex-major-versions branch September 4, 2026 23:14
fersilva16 added a commit to fersilva16/nix-config that referenced this pull request Sep 5, 2026
gpt-6-astra shipped on the ChatGPT/Codex backend, but opencode's oauth
model filter matched ids against /^gpt-(\d+\.\d+)/ and compared them with
parseFloat, so a version with no minor component never matched and the
model was dropped before it ever reached the picker.

1.18.29 fixes it (anomalyco/opencode#47384, #47385): the minor component
becomes optional and major/minor are compared separately. That also
un-breaks gpt-5.10 and later, where parseFloat read "5.10" as 5.1 and
sorted it under the 5.4 cutoff.

models.dev added the entry at the same time, so the model arrives with
its real limits — 1,050,000 context — and a -fast variant alongside.
filipeforattini added a commit to reddb-io/redcode that referenced this pull request Sep 5, 2026
filipeforattini added a commit to reddb-io/redcode that referenced this pull request Sep 5, 2026
1056674754 added a commit to 1056674754/opencode that referenced this pull request Sep 7, 2026
Upstream v1.18.27 (4b7e19e, merged at 26dff3f) -> v1.18.29 (1674747),
29 commits / 116 files. Merge base is b046973 (default header timeout to
five minutes). Staged delta vs pre-merge fork HEAD is identical to the
upstream tag-to-tag diff, confirming a clean adoption.

Functional changes:
- fix(opencode): compare Codex GPT versions by major and minor (anomalyco#47385) and
  allow integer GPT versions in the Codex model filter (anomalyco#47384) — codex.ts;
  disjoint from fork snapshot architecture, auto-merged.
- feat(copilot): send X-Interaction-Id header with session id (anomalyco#47215) —
  github-copilot/copilot.ts, 1 line.
- chore: bump gitlab-ai-provider to 6.13.0 (anomalyco#46914).
- test(core): disable npm audits in the test preload (anomalyco#47222).
- console: limits graph, i18n for ar/br/da/de/es/fr/it/ja + others;
  app: dialog-connect-provider and open-in-app-v2 fixes; go.mdx crawl rules
  gain the x-opencode-session header bullet (took upstream whole).

Conflicts resolved: 29 package.json version bumps (rebranded 1.18.29-sscity;
HEAD had moved to the local v1.18.27 merge meanwhile), bun.lock (took
upstream, rebranded workspace entries via bun install), go.mdx (took
upstream — trivial <br/> placement in the crawl-rules list). Fork deltas
verified intact vs v1.18.29: provider ProviderSnapshot architecture,
deferred snapshot init, GLM/ZhiPu normalization in transform.ts (+77 lines
fork-only), config getFresh/commitFresh. Tests: provider, transform, config,
v2, azure plugin, session retry/llm — 1036 pass / 0 fail across 14 files;
typecheck clean.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
arbocdi pushed a commit to arbocdi/opencode that referenced this pull request Sep 9, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant