feat(llm): support OpenAI Codex endpoints

[linux-devil]

Summary

• add an OpenAI Responses API client selected by /v1/responses endpoints
• send max_completion_tokens for GPT-5/Codex/o-series chat-completions models while preserving max_tokens for legacy models
• make OCR environment variables take precedence over config and accept CI-compatible OCR_LLM_AUTH_TOKEN / OCR_LLM_USE_ANTHROPIC names
• document Codex/GPT-5 setup examples

Tests

• git diff --check
• go test ./internal/llm
• make build
• local ocr llm test against an Azure GPT-5.5 chat-completions deployment

Note: go test ./... currently fails in internal/config/rules because existing tests expect Chinese default-rule snippets while the embedded default rules resolve to English text; that failure is unrelated to this LLM endpoint change.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[github-actions]

🔍 OpenCodeReview found 8 issue(s) in this PR.

• ✅ 8 posted as inline comment(s)
• 📝 0 posted as summary (missing line info)

---

📊 Posting Statistics:

• ✅ Successfully posted: 6 comment(s)
• ❌ Failed to post: 2 comment(s)

❌ Failed Comments Details

• internal/llm/resolver.go: Unprocessable Entity: "Line could not be resolved"
• internal/llm/resolver.go: Unprocessable Entity: "Line could not be resolved"

[vanducng]

exactly what im looking for. hope this will be resolved and merged soon

[lizhengfeng101]

Hi @linux-devil, thanks for this PR!

Since this PR was opened, we've landed a provider-based architecture on main that natively supports configuring different LLM providers (including OpenAI-compatible endpoints). You can see the latest setup in the docs and the ocr llm setup interactive flow.

Would you be able to rebase onto the latest main and rework your changes on top of the new provider system? That way we can avoid duplicating logic and ensure everything fits cleanly into the current design.

Looking forward to seeing your updated design — happy to discuss the approach before you start coding if that would help!

[linux-devil]

Sure, let me look into it Best Regards Harshit Sharma

…

On Thu, Jun 25, 2026 at 7:18 PM kite ***@***.***> wrote: *lizhengfeng101* left a comment (alibaba/open-code-review#62) <#62 (comment)> Hi @linux-devil <https://github.com/linux-devil>, thanks for this PR! Since this PR was opened, we've landed a provider-based architecture on main that natively supports configuring different LLM providers (including OpenAI-compatible endpoints). You can see the latest setup in the docs and the ocr llm setup interactive flow. Would you be able to rebase onto the latest main and rework your changes on top of the new provider system? That way we can avoid duplicating logic and ensure everything fits cleanly into the current design. Looking forward to seeing your updated design — happy to discuss the approach before you start coding if that would help! — Reply to this email directly, view it on GitHub <#62?email_source=notifications&email_token=AACLEYTF6MTRGL5J6VZOQNL5BUUTJA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIOBQGAYDENRXHA22M4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJLDGN5XXIZLSL5RWY2LDNM#issuecomment-4800026785>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACLEYSXZVN6MXBVWJ4BERT5BUUTJAVCNFSNUABGKJSXA33TNF2G64TZHMYTENBRHE3DAMRSGY5US43TOVSTWNBWGAZTAMRQGU2DJILWAI> . Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/AACLEYQM36HBLMSMJ2FHACT5BUUTJA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIOBQGAYDENRXHA22M4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJKTGN5XXIZLSL5UW64Y> and Android <https://github.com/notifications/mobile/android/AACLEYTAS7DMZ6KDDFVGUYD5BUUTJA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIOBQGAYDENRXHA22M4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJLTGN5XXIZLSL5QW4ZDSN5UWI>. Download it today! You are receiving this because you were mentioned.Message ID: ***@***.***>