Runner samples

These samples are portable starting points for Assets → Runners. Copy a sample's source into a runner, then connect it to a build. They do not contain a product name, a local absolute path, or a background scheduler.

Sample	Use when	Build requirements
browser-user-journey-runner.py	You want recurring, observable browser checks.	Browser base URL and at least one fixed test case.
external-command-runner.py	You already have a command-line automation tool.	ORBIT_ADAPTER_COMMAND configured in the runner environment.
inspect_behavior.py	You want to inspect the bundled behavior definition locally.	A repository checkout with the Python package available.
ai-slo-supportops/	You want a safe local target for the AI SLO and behavior-drift Quick Start.	Python 3 only; no provider credentials for the fixture itself.

Browser user journey

Create fixed test cases with a route (path) and, optionally, text that must be visible (expected_text). The sample rechecks failed cases before rotating to the next fixed case. It stores only bounded planning state in OpenOrbit AppData and attaches screenshots and page evidence to each run.

External command adapter

Configure ORBIT_ADAPTER_COMMAND as a JSON array when arguments contain spaces or special characters:

["/opt/automation/bin/check"]

Or use a shell-like command string:

python -m my_automation

The external program should implement status, prepare, run-once, and collect-evidence as one-shot actions. It must not start a daemon or schedule its own repeat loop. The sample captures each action's output as an immutable OpenOrbit artifact.

Local inspection

Run the behavior-inspection sample from a repository checkout after installing the project dependencies:

uv run python examples/inspect_behavior.py

For every sample, adapt only the configuration and bounded work for your project. Keep lifecycle ownership with OpenOrbit: before_all and after_all run once per process; before_each, execute, verify, and after_each run once per iteration.

AI SLO and behavior-drift demo

ai-slo-supportops/ is a complete local sample: a synthetic support website and an evaluator agent that returns structured SLO evidence. It is deliberately local-only and has no credentials, real customer data, external provider endpoint, or background scheduler.