Machines and files

Machines and Files attach to a task. See Core concepts. Create them under Manage, then attach them on the task. They are not treatments, and you do not pick them as separate launch peers. When you compare treatments, only the treatment should change. The work, the files, and the machine stay the same. That keeps lift comparable.

​

Files

Files are folders you attach to a task: a repo, sample data, or fixtures.

1. Create them under Manage → Files.
2. Open the task Files tab and attach what you need. Each attachment mounts at a path in the machine.

Before a run starts, Oqoqo mounts those folders. Every run of the task begins from the same starting files. Keep secrets out of Files when you can. Prefer Variables on the machine, or dashboard connections that enter secrets securely.

​

Upload or import from GitHub

When you add files or folders, choose Upload or From GitHub.

• Upload — add local folders
• From GitHub — paste a public github.com URL; Oqoqo snapshots at the resolved commit

Later, use Check for updates on a GitHub-sourced folder to replace the live record with a newer snapshot. Private-repo import is not available. Skills support the same GitHub import pattern. See Assets.

​

Machines

A machine is the environment a run executes in: a realistic developer setup for the person you are testing for. Machines run on a standard Ubuntu image today. You set:

• Packages — tools and dependencies you expect the user to have
• Services — anything that runs alongside the agent, such as a database or a local copy of your product
• Variables — values the agent needs, such as an API key

Create machines under Manage → Machines. On the task Machine tab, pin one, or leave the Standard Dev Machine. Each run gets a fresh machine. Nothing carries over between runs. For an example with Redis on a custom machine, start from the Redis template.

​

At launch

Oqoqo snapshots the machine and files from each selected task. Later edits in Manage do not change runs already launched. Runs of the same task share that snapshot across every treatment and agent. You measure treatment, agent, and model.

​

Recommendation

Start with the Standard Dev Machine and only the files the task needs. Add a custom machine when the work needs packages, services, or variables that the default does not provide.