Didier Villevalois

@Rizzen hey! Sorry for the late answer too... :smile: Well, this PR was just for the demonstration. We would have to discuss if, how, when and where we integrate this....

Hey @DRSchlaubi ! Nice PR. I had to make a few tiny changes to make it work here: - in the `common` module, regenerate code and dump API again -...

@Ololoshechkin would you consider making the `*ClientSettings` data classes `@Serializable`? That would be really practical for people making frameworks and applications above Koog. That would allow to reuse those classes...

Just a note to say that the current persistency feature stores not only the current graph node but also the history and the agent state. I would hope that the...

I tried that example with "hf.co/mradermacher/ToolACE-2-Llama-3.1-8B-i1-GGUF:Q4_K_M" and it doesn't either call all the tools. Sometimes it calls no tools, and sometimes it calls most, but never does all the 5...

> adding the toolChoice parameter to your prompt The thing is that OllamaClient doesn't support toolChoice...

FYI, I use a `PromptExecutor` wrapper that fixes (in a very very quick and dirty way) some of the recurring bad tool calls that I observed using Llama3.1 based models:...

> I can certainly add the `wasmJs` target. iOS targets need to be built on a Mac, which I don't use for development. I could use GH actions to build...