Skip to content

gproxy v2.6.4

Choose a tag to compare

View all tags
@LeenHawk LeenHawk released this 14 Aug 02:23
v2.6.4
eaf8ad8

v2.6.4

Preserves opaque reasoning data across OpenAI Responses, Claude, and Gemini while improving service-tier support and Console behavior.

English

Added

  • OpenAI ultrafast service tier. Added end-to-end handling for the ultrafast tier, including count-tokens, compact, Claude speed mapping, and fallback behavior.

Fixed

  • Cross-protocol encrypted reasoning. OpenAI Responses encrypted_content, Claude thinking signatures/redacted thinking data, and Gemini thoughtSignature now map directly across request, response, and streaming conversions without passing through a Chat request or response object.
  • Console quota ordering. Provider usage cards now show upstream quota before local quota.
  • Edge update controls. Self-update actions are disabled where the deployment cannot perform a native binary update.

Changed

  • Rust dependencies. Refreshed the Rust dependency set and related compatibility adjustments.

简体中文

新增

  • OpenAI ultrafast 服务层级。 完整支持 ultrafast，覆盖 token 计数、compact、Claude speed 映射及降级行为。

修复

  • 跨协议加密推理。 OpenAI Responses 的 encrypted_content、Claude 的 thinking signature/redacted thinking data 与 Gemini 的 thoughtSignature 现在会在请求、响应和流式转换中直接映射，不再经过 Chat 请求或响应对象。
  • 控制台配额顺序。 Provider 用量卡片现在优先显示上游配额，再显示本地配额。
  • Edge 更新控制。 无法执行原生二进制自更新的部署环境现在会禁用相应操作。

变更

  • Rust 依赖。 更新 Rust 依赖并完成相关兼容性调整。
Assets 59
Loading