dose fast/deep mode use isolated knowledge clusters?

[loewy43]

when Sirchmunk worked at step of "Searching for similar knowledge clusters" , it didn't reused the answer ever recorded in deep mode, even the same question asked in fast mode.

dose the fast mode and deep mode use isolated knowledge clusters?

[wangxingjun778]

It appears to be a bug; I will fix it in the next minor release. :)

[wangxingjun778]

Fixed, refer to PR: #87

Please reinstall sirchmunk: pip install "sirchmunk[web]" -U
(sirchmunk>=0.0.5post1)

[loewy43]

we had tried, but it's not working.

[wangxingjun778]

Looks like Node.js isn't installed ？

[wangxingjun778]

Regarding independent knowledge clusters, a temporary solution is to set the SIRCHMUNK_WORK_PATH environment variable in the ~/.sirchmunk/.env or pass different work_path values during the initialization of the AgenticSearch class.

[huasheng-zhang]

你的意思是重新修改sirchmunk的工作路径；还是给fast和deep模式分别建立不同的工作路径呢；我们当前的工作路径

[wangxingjun778]

sirchmunk目前设计上还是维持单实例模式，如果支持多用户的话，可以这样做：

1. 安装最新的版本>=0.0.5post2，参考release notes: https://github.com/modelscope/sirchmunk/releases/tag/v0.0.5post2
2. sirchmunk web init --work-path /path/to/user1, 以及 sirchmunk web serve --work-path /path/to/user1
3. 同理，新增更多用户则指定不同的/path/to/user-i 即可，效果是每个用户都有自己独立的工作目录（缓存，配置，模型文件，会话历史等）
4. 以上新版本还没同步到docker镜像中，如需镜像运行，可以进入容器内部更新一下sirchmunk依赖。或者后续大版本很快会同步更新镜像。
5. 备注：在web端的settings页面，直接配置SIRCHMUNK_WORK_PATH原则上也是可以的，对于新的work path会自动新建，但是首次会话会有点慢，自带了init过程； 如果SIRCHMUNK_WORK_PATH切换到已有的work path上，目前逻辑是在点击保存按钮时会自动reload该路径下已经存在的各种配置。