Intent matching test syntax

To ensure that the template sentences work as expected, we have an extensive test suite. This test suite is based on YAML files that contain a list of input sentences and the expected matched intent and slots.

The tests are stored on GitHub and are organized by having for each language a directory of files tests/<language>/:

• _fixtures.yaml - Fake entities and areas that can be referenced during testing
• <domain>_<intent>.yaml - Sentences for a single intent and domain. These files should only test sentences that are defined in the match sentences file with the same name.

# Example homeassistant_HassTurnOn.yaml
language: "en"
tests:
  # You can have multiple blocks of tests, each with different expected match data
  - sentences:
      # Multiple sentences can be tested at once
      - "turn on the ceiling fan"
      - "turn the ceiling fan on"
    # Expected match data
    intent:
      name: "HassTurnOn"
      slots:
        name: "fan.ceiling"

Fixtures

When Home Assistant is matching sentences, it will provide a list of areas and entities that can be referenced in the sentence. For tests we define these in _fixtures.yaml.

# Example _fixtures.yaml for English
language: "en"
areas:
  - name: "Kitchen"
    id: "kitchen"
  - name: "Living Room"
    id: "living_room"
entities:
  - name: "Kitchen Switch"
    id: "switch.kitchen"
    area: "kitchen"
  - name: "Curtain Left"
    id: "cover.curtain_left"
    area: "living_room"

Make sure that fixtures do not have generic names like "garage door" or "curtains". Instead, use a unique name like "garage door left" or "curtains left". This is necessary to allow defining matching sentences based on the generic names, like "open the garage door".