Agents and CLI need developing further

[Mainframework]

CLI is king

There is a lack of programmers that are willing to develop the Cli instead of the terrible server. This project has a shot to become very important as is dependency free and native, but it can go in a better direction, filling another important gap that llama.cpp is missing, making the CLI rich and adding some simple agents/skills/tools_ this style: https://github.com/gary149/llama-agent.

The reason is quite simple, Fast, Safe, Secure, and very useful, (in fact so much that developing our own not native agents/skills/tools on top of llama.cpp, is on par with proprietary services).

Great job to all of you, thank you.

[farhi]

I agree, the tools are currently only in the WebUI.
However, the server has, in my view, highly improved. I personally like it.

What is currently to be looked at is the security breach brought my the 'shell_exec' tools which when enabled allows everything and represents a major threat to users files and computer stability.
I think a mechanism for better definition of what commands are allowed should be looked after. I suggest to be able to define such commands from a Markdown file, and the commands must strictly keep their syntax to avoid code injection and every variation.

[Mainframework]

We already solved the issues, if you want things well done, do it yourself :)

Test it here (get the most powerful local version worldwide) here: https://hugston.com and prove me wrong.