An agent runs on a computer and runs tests reserved for it by the Cocoon backend. It may or may not have physical devices attached to it.
Agent has several subcommands:
run command runs a single task. Here's how you run it:
dart bin/agent.dart run --task-name={TASK_NAME}
It runs the task at the currently synced version of Flutter.
ci command runs the agent in continuous integration mode. It continuously asks the Cocoon backend for available tasks, syncs Flutter to the desired revision and runs the task. Use it like this:
dart bin/agent.dart ci
Follow Flutter's setup guide. Make sure Flutter runs successfully for the kind of test scenarios you intend to use the agent for. For example, if you plan to run tests on a real Android device, make sure there is an Android device attached, the Android SDK installed and you are able to run one of the sample Flutter apps on the device.
Pick a good ID for the agent. If there are existing agents performing the same set of tasks give it the same ID with a different number. For example, if there are “mac1” and “mac2”, and you are adding another Mac build agent, call it “mac3”. If you are adding an agent that doesn't fit within any of the existing groups, pick a new ID and start numbering beginning from 1, e.g. “linux1”.
Read Creating an agent about how to create the agent and acquire an authentication token to connect it to Cocoon.
For authentication instructions read go/flutter-cocoon.
At the end of this exercise you should have an agent ID and an agent auth token.
You will need to enter information acquired in the previous step into a config.yaml file. See the sample config file that documents all parameters in detail.