Add a migration
Add a schema change that upgrades and rolls back cleanly, with the checks the migration tests enforce.
Recipes
Step by step for the changes contributors make most often, each pointing to the script or make target that does the work.
Each recipe is the short path through a recurring change. The rules behind the
steps live in the AGENTS.md
at the root and in the directory you change; the recipe tells you which
commands to run and what to check.
- Add a migration
- Record a change: changelog and docs
- Backport to a release branch
- Add a configuration setting
- Add or change an /api/v2/ endpoint
- Add a Helm value
- Add a CLI command or flag
- Add a connector
- Cut a release candidate
- Fix a bug
- Fix a flaky end-to-end job
Reviewing a change rather than making one? See the review checklists.
Record a change: changelog and docs
Write the changelog fragment and the docs a pull request needs to pass the changelog and docs guards and the release cut.
Backport to a release branch
Carry a change merged to main into release-X.Y so the next patch ships it, in the shape release-gap.sh recognizes.
Add a configuration setting
Add a server setting that binds from DEXAFLOW_* and LEOFLOW_*, fails fast on a bad value, and is documented in the configuration reference.
Add or change an /api/v2/ endpoint
Change the public HTTP API without breaking the Airflow UI, starting from the OpenAPI document and ending with a guarded, tenant-scoped handler.
Add a Helm value
Add a chart value with its doc comment, its template, a unit test and a regenerated README, keeping upgrades from the previous chart working.
Add a CLI command or flag
Add a dexaflow command or flag, regenerate the CLI reference, and keep every command written in the docs runnable.
Add a connector
Add a connection type to the generated connector catalog from a pinned Airflow provider, and prove the connection reaches the task.
Cut a release candidate
The short path through an rc cut. RELEASING.md has the full flow and the recovery steps; scripts/cut-release.sh does the work.
Fix a bug
Reproduce the bug in a test that fails, fix it, and prove the test would catch it coming back.
Fix a flaky end-to-end job
Tell an infrastructure transient from a race in the test or the product, reproduce it, and fix the cause instead of retrying past it.