feat(codex): allow gpt-6-astra - #129
Merged
Merged
Conversation
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Owner
|
Thanks for adding this! The lane and context metadata line up with Codex's model catalog. Merged. |
Harshkamdar67
pushed a commit
to Harshkamdar67/claude-code-proxy
that referenced
this pull request
Sep 6, 2026
Allow Codex users to select the GPT-6 Astra model directly and through its priority-tier fast alias. Register the model with Responses Lite routing and model discovery. Keep the existing Claude aliases unchanged so opting into Astra remains explicit. (cherry picked from commit 55bf0b5)
tak848
added a commit
to tak848/dotfiles
that referenced
this pull request
Sep 7, 2026
## Why Codex の live カタログ（`codex debug models`）に GPT-6 Astra（`gpt-6-astra`、description は "Our most capable model for complex, demanding work."、priority 1）が入った。`claudex` は Claude 側の fable / opus スロットをどちらも `gpt-5.6-sol` に向けていたが、fable > opus の重みに合わせて fable スロットだけ astra に向けたい。 ## What - `dot_zsh/functions/claudex.zsh` - `CLAUDEX_FABLE_MODEL`（既定 `gpt-6-astra`）を追加し、`ANTHROPIC_DEFAULT_FABLE_MODEL` に渡す。`/model fable` と `model: fable` の subagent が astra で動く - primary（`--model`）と opus スロットは `gpt-5.6-sol` のまま。素の Claude の既定が Opus であるのに合わせている。primary も astra にしたいときは `CLAUDEX_MODEL=gpt-6-astra claudex` - `claudexf` は `gpt-6-astra-fast` を渡す - claude-code-proxy は Codex モデルを allowlist で弾き、`gpt-6-astra` の登録は [raine/claude-code-proxy#129](https://github.com/raine/claude-code-proxy/commit/55bf0b5818b461e1860964809726f99d2fd52c10)（2026-09-04、main）で、mise で入れている v0.1.35 には含まれない。実際に v0.1.35 へ `gpt-6-astra` を投げると `Unknown model` で即エラーになることを確認した。対応版が出るまで壊れないよう、起動時に `claude-code-proxy models` を見て未対応なら fable スロットを primary に落として警告を出す。Renovate で対応版に上がったら起動済みの旧プロキシプロセスを止めて再起動すると astra が有効になる - `CLAUDEX_CONTEXT_TOKENS` の既定 872000 は据え置き。astra の `max_context_window` も sol と同じ 872000 - `CLAUDE.md`: スロットの割り当てと allowlist フォールバックの説明を更新 動作確認は stub の `claude` を PATH 先頭に置いて環境変数を出力させた。v0.1.35 では警告付きで fable=sol にフォールバック、astra 対応の stub プロキシでは fable=`gpt-6-astra`（claudexf は `gpt-6-astra-fast`）、`CLAUDEX_FABLE_MODEL` 明示時はそれが優先されることを確認。
|
@raine Could you please create a new release that includes this PR? Thank you! |
11 tasks
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
3 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.
Adds
gpt-6-astra(released 3 September) to the Codex provider.ALLOWED_MODELS(translate/model_allowlist.rs): selectable by name and via the-fastpriority-tier alias.uses_responses_lite: it serves from the Responses Lite lane alongside the gpt-5.6 tier. Verified with a ChatGPT-authenticated Codex account: a request through the proxy returns 200 withstop_reason: end_turn, and a full Claude Code turn withANTHROPIC_MODEL=gpt-6-astracompletes.CODEX_MODELS(registry.rs): so the model and its-fastsibling appear inGET /v1/models.Not changed:
MODEL_ALIASES. The Claude aliases keep pointing at the gpt-5.6 tier; anyone who wants opus or fable on Astra can setANTHROPIC_DEFAULT_OPUS_MODEL=gpt-6-astraon the Claude Code side.full_lane_web_search_modelis also untouched, as Astra is not lite-only.Via Codex the model reports a 272k context window (872k max), so the 1M hint does not apply.
cargo fmt --checkclean.cargo test --lib: 853 of 854 pass in the full parallel run; the one failure,cancellation_while_replacement_startup_is_blocked_aborts_request_state, is a timing test that passes in isolation on both this branch and unpatched v0.1.35.🤖 Generated with Claude Code