[Bug]:AutoContextMemory的上下文压缩造成思考出现英文

[nilcodes]

Describe the bug
可能是由于上下文压缩默认采用的是英文提示词，在压缩后，产生的总结为英文，导致在上下文中的思考变成了英文了。

如何解决
在默认提示词中约定“使用用户提问的语言”来压缩内容。

[xiaoyuuuuuupeng]

@AlbumenJ 是否可以让大模型以第一条UserMessage使用的语言进行压缩呢

[liuxh613]

我也碰到这个问题了。比如地址什么的，转到英文是拼音，再转回中文完全是错的。完全可以在prompt中约束 和提问的语言保持一致

[Alexxigang]

I traced this through the AutoContextMemory compression flow and confirmed the likely root cause: the compression/summarization prompts are always generated in English, and there is no language-preservation hint derived from the conversation itself. I'm working on a fix that infers the preferred conversation language from the round messages, threads that hint through the compression prompt builder, and adds regression tests for Chinese and mixed-language cases.

[Alexxigang]

Opened a fix here: https://github.com/agentscope-ai/agentscope-java/pull/1145\n\nIt preserves the compression language across the AutoContextMemory compression flow and adds regression coverage for Chinese and mixed-language conversations.