RAG, RAFT or Fine-Tuning Models

[euroblaze]

What's the community thinking about RAGing, RAFTing or Fine-Tuning opensource models, for the specific purpose of DevOps, Sys-Admin, MLOps or for that matter any such task-domain?

From what I understood so far about gptscript, tools are the space where a user can express needs.

But in my (very) initial tests, just typing ls on the gptscript prompt (llm-basics-demo) is yielding this output.

my@dev1:~$ gptscript github.com/gptscript-ai/llm-basics-demo

  Hello! I'm here to help you with any questions or information you might need. How can I assist you today?                       

> ls

  It looks like you're using a command that is common in Unix-based systems to list directory contents. If you need help with     
  something specific related to this command or anything else, feel free to ask!                                                  

@Your Friendly Chat Assistant>  

Every systems guy would expect a different output to the ls prompt.

What are your thoughts on:

• the LLM "powering" gptscript having some background knowledge related to the intentions of the user?
• which of RAG, RAFT or Fine-tuning could/would work better? Or are foundational models (remember, their knowledge is old) already the best fit?

Thanks for your views!