So are you saying then that the way to, from architecturally, the way to do this is that the orchestrator agent that is handling that initial input, let's say that it determines that this agent is the right agent to solve this problem, are you saying that then you won't actually just kind of hand off into that agent, what you would do is you would keep the orchestration layer always processing what's coming in and then sending that request into the individual agent level and then bringing it back up to the orchestration level. Is that what you're saying? Yes.

So actually we provide like, maybe we can see like these orchestrator is a proxy. So it fully isolates the user from the skill. So all the input and output from specific skill goes through the orchestrator. So after some specific skill or specific board produced its own answer, first of all, this answer goes to the orchestration and then the orchestrator decides if this like response appropriate or not, maybe it should be guardrails or or maybe it should be post-protest it. For example, some names I didn't know or and only then it goes to the user. So this orchestration layer is constantly like monitoring what is going on and decides if, for example, we need to switch to some in other skill. Right.