Icon base by Lorc under CC BY 3.0 with modifications to add a gradient
I use stepsy for step counting, energize for macros, openscale for measurement taking, and flexify for workouts. I like then well enough. I know wger combines a lot of that, but I didn't want to move everything over.
As far as devices go, I'd check out gadgetbridge and see what's supported and go with that. I don't use a wrist worn device, so I can't really comment beyond that.
There are some that do steps and heart rate, but not sleep. I have a banglejs v2 that’s pretty cool. It collects the data, but you’ll be responsible for gathering it.
Having done a lot of research into this, the state of things sucks right now. The current open-source options have very bad tracking (like, just a very rough estimate) and are much more focused on smart-watch interfaces, which I get because having made a smart watch development board myself, PPG AFEs are a black hole of NDA-riddled bullshit, manufacturer lies and bad documentation, and no support (looking at you Analog Devices). ZSWatch is the most promising open source project using the best openly - available sensor set with the potential to hit 80% heart rate accuracy, but that is still years away.
Honestly for half-good tracking, your best bet js Gadgetbridge + a proprietary smart device that is entire locally supported.
For example, the 100€ Amazfit Helio Strap (no screen) has great heart rate tracking and better than average sleep tracking (old strap style or bicep strap) and you can run it once in the official app to update firmware and extract the encryption keys, then from then on run it completely locally on gadgetbridge and uninstall the official app completely.
Letting perfect be the enemy of good enough sadly leads to almost useless tracking data in the open source wearables world. For now at least. Making wearables is insanely expensive and to get the best sensors you need NDAs and quantities >10k per year which is unobtainable for community open source projects right now, and you need massive amounts of user data to build good algorithms to analyze the data from the sensors.