Did some more experimentation with local LLMs. This time looking at how to integrate MCP servers.
As a fun experiment I tried to use tool calling to implement a simple POC of a basic GraphQL-esque response from a series of tool calls inferred from the prompt.
My takeaway is that tool calling works reasonably well, even in small LLMs (7-8B).
As a fun experiment I tried to use tool calling to implement a simple POC of a basic GraphQL-esque response from a series of tool calls inferred from the prompt.
My takeaway is that tool calling works reasonably well, even in small LLMs (7-8B).
Article: https://www.teachmecoolstuff.com/viewarticle/using-mcp-serve...
reply