[DOCS] Clarify AI Assistant will spawn ML nodes

[lucabelluccini]

Clarify AI Assistant will spawn ML nodes

To the doc writers:

• please change wording
• backport to all versions back to AI Assistant pages got created
• add an additional paragraph if the use of AI Assistant / Knowledge Base implies ML Nodes will no more be scaled down back to 0

[github-actions]

A documentation preview will be available soon:

• 📚 HTML diff
• 📙 Observability Guide

[dedemorton]

Thanks for updating the docs. Just some minor changes to follow Elastic style and avoid what grammarians call an unclear antecedent.

[dedemorton]

@lucabelluccini Do we want to port this change to the serverless docs, too?

[lucabelluccini]

Hello @dedemorton - thanks for the review.

I think we need some inputs from engineering as:

• It would be valuable to also mention the fact enabling AI Assistant / KB will not scale down ML nodes (I am not certain)
• I think in serverless the autoscaling of ML doesn't work in the same way

[dedemorton]

@dgieselaar Can you respond to Lucas' comments here so we can finalize these changes? Thanks! (My apologies if I'm pinging the wrong person here. I'm still trying to figure out who knows what on the team.)

[CoenWarmer]

Hello @dedemorton - thanks for the review.

I think we need some inputs from engineering as:

• It would be valuable to also mention the fact enabling AI Assistant / KB will not scale down ML nodes (I am not certain)

I assume you meant to say 'disabling AI Assistant / KB' ? The AI Assistant will not explicitly scale down ML nodes if it is disabled.

• I think in serverless the autoscaling of ML doesn't work in the same way

I'm unfamiliar with how autoscaling works in ML, but happy to ask around:

https://elastic.slack.com/archives/C2AJLJHMM/p1711029351315459

[dedemorton]

@CoenWarmer Can you help me interpret the response to the question about autoscaling? The response includes a link to a document that "could be out of date" and the code for autoscaling, so I'm still not sure what, if anything, I need to say in the serverless docs.

[dedemorton]

Suggested change

	====
	In {ecloud} or {ece}, if you have Machine Learning autoscaling enabled, Machine Learning nodes will be started when using the knowledge base and AI Assistant. Therefore using these features will incur additional costs.
	====
	====
	In {ecloud} or {ece}, if you have Machine Learning autoscaling enabled, Machine Learning nodes will be started when using the knowledge base and AI Assistant. Therefore using these features will incur additional costs.
	Note that disabling the AI Assistant will not result in ML nodes being scaled down.
	====

[dedemorton]

@lucabelluccini Is this what you had in mind?

[lucabelluccini]

I think that's great @dedemorton
I would love to get confirmation by engineering whether ML nodes will scale down.
ATM, it seems that once ML jobs are stopped or finish, ML nodes can be automatically scaled down.

But AI Assistant is not using ML Jobs but other ML features...

[bmorelli25]

Following up here. Did we ever figure out if ML nodes scale down? This comment https://github.com/elastic/support-tech-lead/issues/1392#issuecomment-2030864999 says we do.