問題修復
修復 LLM 閘道與自訂網址連線，提示詞快取正常重用
使用 LLM 閘道（LLM gateway，集中管理模型請求的中繼服務）或自訂基礎網址的工作階段，提示詞快取現在已恢復正常。過去這類工作階段的快取（cache，把先前處理過的文字暫存起來直接用）功能會失效，導致每次對話都得重新計算重複的前文。現在能順暢重用已暫存的內容，替每一次請求省下重複解析的時間與成本。
繁體中文
新功能
內建 Concise 精簡輸出風格，直出成果不講廢話
如果你討厭 AI 每次回話都先講一大串客套前言或碎碎念的工作進度，現在有了更乾淨俐落的回答方式。新版本在設定中加入了內建的「Concise」輸出風格。開啟後，Claude 會直接先給出具體結果，並省略開頭的前言寒暄與多餘的過程旁白，但背後的分析與執行工作依然一樣徹底仔細。想切換這個風格，只要在終端機輸入 /config 指令，並在「Output style」選項下選取「Concise」即可。
English
BUGFIX
Prompt caching restored for LLM gateways and custom base URLs
Prompt caching is now fully functional for sessions using an LLM gateway or custom base URL. Previously, routing your traffic through an intermediate gateway or custom endpoint broke prompt cache reuse, forcing subsequent API turns to re-process identical prompt prefixes from scratch. That cache invalidation issue has now been resolved, allowing sessions using an LLM gateway or custom base URL to save tokens and latency on repeated context.
FEATURE
Built-in "Concise" output style leads with results and skips narration
You can now choose a cleaner, to-the-point response style if you prefer answers without filler commentary. This update adds a built-in "Concise" output style: Claude leads directly with concrete results while skipping preamble greetings and step-by-step narration, without compromising the thoroughness of the underlying work. To switch to this mode, run /config and select "Concise" under the "Output style" menu.
原始 CHANGELOG
Original changelog修復 LLM 閘道與自訂網址連線，提示詞快取正常重用
Fixed prompt caching for sessions using an LLM gateway or custom base URLOriginal changelog內建 Concise 精簡輸出風格，直出成果不講廢話
Added a built-in "Concise" output style: Claude leads with results and skips preamble and narration, while doing the work just as thoroughly. Select it under Output style in /config.