Home Assistant

Why it's ranked here

The verdict: this is infrastructure, not a weekend project, and the repository shows it. Apache licensed, marked production stable in its packaging metadata, with 59 core runtime requirements nearly all pinned to exact versions, and an in-house manifest validator running 29 checks that every integration must pass. Adding device support starts from a scaffold command that generates code and then reruns validation, requirements generation, and translation extraction. The cost of that rigor is bulk: every supported integration lives in this one repository, so contributing means working inside a very large tree with a strict review surface.

What's good

Operational care shows up in small places. Startup takes a lock so only one instance runs per configuration directory, enables a crash handler writing to a fault log, removes that log when it stays empty, and counts stray non-daemon threads on exit. Recovery mode and skipping dependency installs are command line flags; safe mode is read from the configuration directory. The web interface filters states by per-entity read permission for non-admin users, and firing events or opening the event stream requires an admin token. Embedded maintenance scripts install their own declared dependencies before running, and filename and path helpers reject traversal characters outright.

Tradeoffs

Not for every host. The launcher refuses to start outside macOS and Linux and points Windows users at a Linux subsystem, though one flag bypasses the check. Python 3.14.2 or newer is required, which is aggressive. The container build installs an exact pinned dependency set plus a bundled streaming binary, so running outside the official image means matching those versions yourself. The lint configuration disables a long list of checks, most delegated to a second linter and several marked as pending cleanup, so the rules actually enforced are narrower than the tooling list suggests.

How to use it well

Pick this if you want device automation running on hardware you control, with state, events, and services reachable over a local HTTP interface you can script against. It suits people comfortable running a long-lived service and editing a configuration directory by hand. If you plan to add device support, the generator and the manifest validator make the contribution path explicit. What it does not give you is a hosted service, an account you sign up for, or a general application framework: the entity and service model assumes homes, devices, and rooms.

Observed

License

Apache-2.0, declared in pyproject.toml

Language

Python, async throughout

Install surface

Python package exposing a hass console script; container image built from the repo Dockerfile

Interfaces

command line launcher, local HTTP state and service API, server-sent event stream

Platform support

Linux and macOS only; Windows via a Linux subsystem

Structure

one package per integration under a single components directory

Configuration

file-based configuration directory on the host machine, not a hosted account