nvda-addon-testkit
End-to-end testing for NVDA add-ons, against a real NVDA, in CI.
Your add-on’s logic can be unit-tested with stubs. What cannot be stubbed is whether it installs, registers, and behaves inside NVDA itself. This kit provisions a disposable portable NVDA, installs your add-on into it, and lets pytest drive it.
def test_my_addon_announces_itself(nvda, addon_under_test, assert_no_unexpected_errors):
before = nvda.speech.index()
nvda.keys.press("NVDA+shift+m")
assert "my add-on is ready" in nvda.speech.wait_for("ready", timeout=10, since=before).text
assert_no_unexpected_errors(nvda)See Example Project for this same pattern applied to a real, working add-on.
Core concepts
| Fixture | What it gives you |
|---|---|
|
a connected client, reset between tests |
|
the path to your built |
|
that bundle, installed and enabled, NVDA restarted |
Every fixture and namespace has its own page under Guide, most with a real, tested code sample pulled straight from this kit’s own test suite:
| Namespace | Use it for |
|---|---|
what NVDA asked to say, and waiting for it |
|
the raw text sent to the braille display |
|
sending gestures through NVDA’s own input pipeline |
|
reading and writing NVDA’s configuration |
|
structured log records, and |
|
two-phase install, remove, and state |
Requirements
Windows to run the tests. NVDA is downloaded automatically — you do not need one installed, and nothing touches an NVDA you already have.
Tests run serially: only one NVDA can own a desktop session, so pytest-xdist with more than
one worker is refused rather than silently producing nonsense.
Want to help? Learn how to contribute to the ZirekHQ docs ›