feature-flags.md

Feature flags

Feature flags are implemented using OpenFeature with an InMemoryProvider. Flags are declared in YAML files and loaded at startup.

Flag files

Each environment has its own flag file. The application selects the file based on the DEPLOYED_TO environment variable, falling back to flags.yml for local development where DEPLOYED_TO is not set.

Environment	File
dev	flags.dev.yml
review	flags.review.yml
preprod	flags.preprod.yml
production	flags.production.yml
local (no DEPLOYED_TO set)	flags.yml

Adding a flag

1. Declare the flag in each environment's file.

Add an entry to all five flag files (flags.yml, flags.dev.yml, flags.review.yml, flags.preprod.yml, flags.production.yml). Set the value to true to enable or false to disable in that environment.

flags:
  my_flag: false

2. Check the flag in application code.

from manage_breast_screening.core.feature_flags import FeatureFlag

if FeatureFlag.is_enabled("my_flag"):
    # feature is enabled

FeatureFlag.is_enabled returns False if the flag is missing or the provider is unavailable. If a flag name is not found in the YAML file, OpenFeature silently returns the fallback rather than raising an error — so a typo in a flag name will return False without any indication something is wrong.

Enabling a flag in tests

Use the with_flag_enabled fixture to turn a flag on for the duration of a single test:

def test_something(with_flag_enabled):
    with_flag_enabled("my_flag")
    # flag is enabled for this test only

Multiple flags can be enabled by calling it more than once:

def test_something(with_flag_enabled):
    with_flag_enabled("my_flag")
    with_flag_enabled("another_flag")

In class-based tests the fixture must be relayed via an autouse fixture so it is accessible to helper methods:

@pytest.fixture(autouse=True)
def flags(self, with_flag_enabled):
    self.with_flag_enabled = with_flag_enabled

def and_my_flag_is_enabled(self):
    self.with_flag_enabled("my_flag")

After each test the flags are reset to the defaults from the YAML file.