Local model + sympy as a tool?
Posted by MrMrsPotts@reddit | LocalLLaMA | View on Reddit | 2 comments
Is there a good way to let a local model use sympy when it needs to?
Posted by MrMrsPotts@reddit | LocalLLaMA | View on Reddit | 2 comments
Is there a good way to let a local model use sympy when it needs to?
o0genesis0o@reddit
Attach whatever LLM agent you use into this mcp server could be one way: https://github.com/sdiehl/sympy-mcp
Please do review the MCP server code before using. I just share the search results. I haven't used this before.
Several-Tax31@reddit
Have a genuine question: what is the difference between using mcp vs just telling the model to write the sympy code? I'm always telling the model to use whatever libraries they should use when they need to, but do you think using mcp has any advantages over this? I never understand the point of the skills, or mcp's, but I feel like I'm missing something.