Where tests run
Where a delivery's tests run, how each result names its environment, and where the evidence is kept.
Every check an agent records names its environment: the app on the agent’s machine, a pull request preview, or a deployed step such as Staging or Production. Tests run where the agent runs them. Blume records where each check ran and ties it to the code it ran on.
We recommend proving the outcome locally before review, then checking it again where users are.
Set up
Environments belong to an application. Have your agent describe the project’s applications and their steps before the first delivery. See Applications and environments.
Open an application from the project’s Applications tab to see each environment’s address, how tests sign in, and its test data.
Local runs
Most checks before review run on the agent’s machine, against the app started from its checkout. The Blume plugin reports the code the checkout is at, including uncommitted changes. Blume does not verify that report.
A check against localhost is a local check, whatever the environment is called. It never counts as proof of a deployed step.
Ask for checks that start from a known state: empty or seeded data locally, and a test account the agent creates on a deployed step.
Previews and deployed steps
- A pull request preview is an environment like any other. Its evidence belongs to the pull request’s code.
- A deployed step is checked against its own address. The result shows on that step in the delivery’s Release tab.
- A step is verified when a check of the code going out passed there. Until then it reads “Deployed, not checked yet”.
- A check against the wrong code does not verify a step. See Release.
Where evidence is stored
Results, reports and screenshots are uploaded to Blume and stored with your workspace. Opening an evidence file needs a signed-in member of that workspace. An agent can read only the evidence of its own project. See Local and cloud data.