Hey, even if your n8n is not "working", the deployment itself will still consume resources and therefore count towards your usage as long as it is running (as in active deployment).

Yep, this is normal behavior. To halt spending you'll have to click to remove the deployment of each individual service inside each individual project. So for n8n it'll usually be the worker, the primary, the redis and the postgres services.

then let me ask you this: this rag work (if I can get it to work)...will be used on a blog...then people querying means more resource consumption...in this case the hobby plan will definitely not be enough...is a pro subscription necessary. and is pro enough for such a work...but in this case it is the same price as n8n...

Hmm, I think this reasoning is incomplete.

If you don't need the PRO features, keep the hobby plan. You already get the 5 USD usage. n8n should cost you 5 or 7 USD, so you'd pay just 2 extra bucks if your usage is 7. You don't need to upgrade to the PRO plan if your usage is expected to surpass 5 USD.

If you enable the serverless feature, it'll cost virtually nothing if your usage is low (like 2 USD per month), but you'll still pay 5 coz that's the minimum. Actually several thousand users (for a blog) will likely keep your bill under 10 bucks. Depending on what you need you can even save a lot on AI usage if you use Open Router instead of Open AI. Lots of good free models there.

Don't worry too much about this now, just start, then learn how to decrease costs. That's what soft and hard limits are for, to prevent attacks like DDOS and unplanned usage.

my problem is that the hobby plan started on 11 may...I didn't use the n8n for 3 days...but despite this, with a workflow that doesn't work, more than half of the plan (2.88 :D) has already been used...which makes me think...most probably before 11 june, the hard limit I set will reach 10$ spending...(I am writing with the help of a translation program, please excuse me if I am not understood)

That's because you didn't know you had to enable serverless mode or just remove the deploys. Now you know. Consider it an investment. Just keep doing it, with the Open Router tip I gave you you'll save more on AI than you'd spend on hosting.

I have activated your server in the primary, but if the project is activated, I have to activate it again, right ? (current usage 3.18 $ )

I couldn't use openrouter because no model in embeddings worked as well as open ai embeddings...i use as chat model gemini

No, if it's active then it's active. Even with serverless mode you don't need to activate it again.

And regarding the embedding models, yeah, you'll have to use Open AI directly, but for everything else it's worth trying.