Setting up Prometheus and grafana on separate nodes was purely for resource isolation, as in Prometheus (the metrics store) will need 1GB ram and will do a lot of disk read/write. When you will use the dashboard, Prometheus as well as grafana will consume resources. I am pretty sure reef-pi will run fine in 8Gb /Pi4 setup, its just I have not tested and I will worry about other resources (disk in particular) that may have spiky usage which in turn can impact reef-pi. It is possible to mitigate most of these, but that will be some work (systems cgroup etc). This is more about validation and then recommendation. If any of you want to venture, I am happy to help out. I have done this (including running the pi as its own wifi hotspot) so the installation and configuration is really not that hard, also we know for sure that it works. It just about running it for longer, with realistic load and see if everything works out. As you get to decide how long you want to store metrics, and as you look at longer and longer duration dashboard, this will incur higher and higher resource usage in pi... as time goes by. Again, this too can be addressed by judicious use of dashboard (e.g. keep only 12 months of data), but we have test out some of these bounds. before recommending..
My recommendation will always be running Prometheus and Grafana on separate Pi and preferably with an 2.5" SSD attached to it. They are 20$ and adds 100G of reliable storage. Enough to run and keep all data needed for a reef-pi and more.
I have thought about mobile clients. It's definitely desirable, and there were a couple of independent attempts to make it, but it didn't happen yet. There is interest, it's all about resourcing. If we have folks who are interested in developing/testing... I can chip in. Mobile is definitely a secondary thing compare to web interface. Because of its lower development resource requirement and much higher applicability. A website will work on any mobile (not as good as native apps) as well as all other devices. But the mobile apps won't work for desktops. Mobile app by themselves will not solve the metrics/dashboard challenge, it has to be combined with cloud base storage. Also for apple and android, its extra painful to make two different apps (there are ways to make a single code generating two types of apps,. but that's another complication). Just to add to the pain apple developer certificate itself is 100$ and the app hosting also will cost $$. Same goes for all cloud based things. A pi in the cloud is 60$/month. As of now, reef-pi (app, docs etc) development and distribution do not have incur opex/recurring cost. This is something that discouraged me to not have any cloud-based integration and mobile apps.
reef-pi's strength is in its ecosystem (Pi, Linux). Something that keeps it free, open and affordable. If I get one hour to invest in this, I feel I should spend on reef keeping related features (stepper driver, doser time->volume conversion etc).. and not on that closed source, non-core functionality type things. I do not have any opposition for those other goodies, and if we get interest now or later, we'll definitely embrace those. I would absolutely love an iPhone app. I'll pay 5$ for it