Datapoints analytics integration for exploring, investigating and annotating data

Hi all,

I’ve been working on a custom integration called HASS Data Points and thought I’d put it out here to get some real-world feedback before I take it any further.

This came from a gap I kept running into myself. Home Assistant is very good at showing what your sensors are doing, but not particularly helpful when it comes to recording why something happened. If energy usage spikes, heating behaviour changes, or something subtly shifts over time, there’s nowhere to record that context alongside the data. You end up trying to piece it together later, which isn’t ideal.

Data Points is an attempt to address that. It lets you record timestamped events (notes, labels, icons) and overlays them directly onto your history charts, so you can actually see what was happening at the time rather than just the raw values.

There’s also a per-entity analysis layer built into the history view. Things like trend lines, summary stat bands, rate-of-change overlays, and a set of anomaly detection methods (Z-score, IQR, residuals, flatline detection, spikes, comparison windows). You can also overlay different time periods for the same sensor to compare behaviour, and there’s a sidebar panel for browsing and editing recorded events.

This is still fairly early. It works, but I’m sure there are rough edges and probably a few things that only break once someone else runs it on a completely different setup.

What I’m mainly interested in is:
• whether it actually runs cleanly on other setups
• whether the history view renders properly or does something unexpected
• whether the UI makes sense, or just feels like too much going on
• any mobile issues
• performance once there’s a decent amount of historical data
• any sensor types where things behave oddly

Also very open to general “this doesn’t feel right” feedback. That’s usually the most useful at this stage.

Repo is here:

If you try it and it falls over, that’s still useful. Even a “nothing showed up” is helpful right now.

Happy to answer any questions or go into more detail if needed.

I absolutely love this.
There are only two drawbacks I see right now:

  1. It can take what seems like an eternity to plot even a few day’s worth of data (maybe I’m using it wrong?).
  2. I can’t get it to show binary data. I’d really like to be able to plot binary vs continuous to look for things like: did the temperature change trigger the fan to turn on at the correct setpoint.

Sorry, I only just noticed the reply.

What kind of data are you plotting, what device are you running HA on, what is the time range you have selected? Depending on the fidelity of the data, length of data and total number of data points, it could take a while. It also depends on the data processing you have enabled for the entities that determines the speed of load. In the latest version there is a downsampling data option that should default to 1 minute so you don't get huge multi-second datasets. You could try enabling the downsampling time period higher, 10, 15 minutes or something like that.

There seems to be a bug at the moment with binary data, that makes it not appear at all. However also worth bearing in mind that HA only stores a short amount of binary data by design, 10 days I believe. Increase retention times for things like power usage, on off functions, etc