Make the new llama-server internal tools to work in TUI / WebUI

[devingfx]

This discussion is a follow up from #936 , that deviated from OP to talk about llama-server internal tools...

@aittalam wrotes:
Note that while tool calling is supported by llama.cpp's server, this feature is not taken advantage of in the TUI (which is why I provided the python code to test it). What the TUI should do is implement an agentic loop close to what I sketched above in python:

• get the available tools from the llama.cpp server
• add those tools' descriptions to the request you are sending to the server, together with your prompt
• for any tool call the server returns you, execute it via the server's /tools endpoint (see the python code above) then return the output to the LLM

@devingfx wrotes:
But again, the whole point was to be able to quick test tools with random models using TUI and/or WebUI without having to code something :/

The reason of existence of this llamafile project is easyness, no config, no code right? So it would be cool to have tools working in TUI/WebUI, right ?
I understand though that this internal tools from llama-server.cpp are new features.

Are the TUI / WebUI specific to llamafile? Or are they external (pluggable) projects ?

What is the complexity to make internal tools working in these quick UIs ?

( My guess is that llama-server team would eventualy add some more basic tools, so getting them to work out of the box would be awesome ! )