feat: update reasoning for claude opus 4.7

[hustxiayang]

Description

1 Previously, the display field defaulted to summarized and was dropped during translation. However, on Claude Opus 4.7 and Claude Mythos Preview, display defaults to omitted instead, so users must set display: "summarized" explicitly to receive summarized thinking [1]. This PR exposes the display parameter so users can control thinking content visibility.

2 Add opus-4-7 and mythos-preview to the list of effort-supported models. Also, with Claude Opus 4.7, Anthropic introduced xhigh as a separate effort tier between high and max for long-horizon agentic and coding tasks [2]; Upgrade anthropic-sdk-go from v1.27.1 to v1.38.0 for OutputConfigEffortXhigh constant support; Define self-owned ReasoningEffort constants (none, low, medium, high, xhigh, max) in the internal OpenAI schema package, replacing direct usage of openaisdk.ReasoningEffort* constants across Anthropic and Gemini translators. This decouples the gateway from the upstream OpenAI SDK's constant definitions and allows extending with values like max that are Anthropic-specific.

3 Add OfAdaptive branch to AWS Bedrock thinking map (previously missing).

4 4 outputConfigModels (structured outputs) is not updated for the new models as it is not available on AWS Bedrock for both Converse API and InvokeModel API.

1: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking#summarized-thinking
2: https://platform.claude.com/docs/en/build-with-claude/effort#effort-levels

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.44%. Comparing base (edef4c5) to head (07ccdf8).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2089      +/-   ##
==========================================
+ Coverage   84.42%   84.44%   +0.02%     
==========================================
  Files         134      134              
  Lines       19162    19175      +13     
==========================================
+ Hits        16177    16193      +16     
+ Misses       1998     1995       -3     
  Partials      987      987              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[dosubot]

Related Documentation

1 document(s) may need updating based on files changed in this PR:

Envoy's Space

vendor-specific-fields /ai-gateway/blob/main/site/docs/capabilities/llm-integrations/vendor-specific-fields.md — ⏳ Awaiting Merge

^{How did I do? Any feedback?}  

[hustxiayang]

/gemini review

[gemini-code-assist]

Code Review

This pull request updates the Anthropic SDK to version 1.38.0 and adds a Display field to thinking configurations in Anthropic, OpenAI, and AWS Bedrock schemas. It also introduces a local ReasoningEffort type alias and constants to decouple the implementation from the OpenAI SDK, while extending support for new Anthropic models like opus-4-7 and mythos-preview. Review feedback identifies discrepancies in docstrings regarding the none reasoning effort value and suggests using the local type alias consistently in struct definitions to complete the decoupling.

[hustxiayang]

@aabchoo @nacx Do you mind to take a look? Thanks a lot!

[nacx]

LGTM. Can you fix the conflicts and rebase to latest main?

[hustxiayang]

/retest

[hustxiayang]

@nacx Hi, I have rebased and fixed the conflicts. Could you please help to merge this pr? Thanks a lot!