basic-memory support specific LLMs？

[Liuyuan0803]

When I use qwen in china, the MCP server has not been used. Or maybe my usage is wrong?

[phernandez]

Hi @Liuyuan0803, thanks for trying out basic-memory. I haven't tried qwen. You might try adding the ai assistant guide to your context https://memory.basicmachines.co/docs/ai-assistant-guide. You can download the file as markdown and add it. It will explain to the AI how to use all the tools.

if you want to send me logs, you can find them in ~/.basic-memory/*.log and I will take a look at them. You can send them to hello at basicmachines.co.

[Liuyuan0803]

Thanks, it is helpful to me, but I found another question: I use cline as mcp client in vscode, if I want to use basic-memory, I must say: please use basic-memory mcp server and search related knowledge to explain my question and it can use basic-memory, if not, the agent cannot use the server.
this is the correct way to use it?

[phernandez]

It really depends which model you are using. Some have more training with using tools and work better. Claude Sonnent 3.5/3.7 works really well, but I found ollama to not work so well.

One think I've done in the past is to have an "education session" with the AI, where I asked it to use all the basic-memory tools and explore the knowledge base. After that I asked it how I could best prompt it to use the tools, then saved that into the project memory.

[Liuyuan0803]

OK, I will try it. Thank you so much