Account and authentication

You need an account before you can run anything. The CLI needs no API key: it holds a session for you, the dashboard signs you in with the same account, and the REST API takes an access token the CLI prints for you. Inference endpoints use separate API keys (Inference endpoint).

Create an account and sign in to the CLI

This is the fastest way, and it leaves you signed in. It needs the CLI. If you do not have it yet, install it first: Ways to use distil labs.

distil signup

The command prints a short code and opens the distil labs sign-in page in your browser. On that page:

1. Create an account, or sign in if you already have one.
2. Check that the code on the page matches the one in your terminal, and confirm it.

The command then prints Logged in as <email> and you are signed in.

Create an account from the browser

Sign up at app.distillabs.ai/sign-up. This creates the same account, but it gives the CLI no session, so bring the account to your terminal with distil auth before you run anything.

Sign in to the CLI

Sign in with an existing account:

distil auth

This is the same flow as distil signup: confirm the code in your browser and the CLI picks up the session.

The browser does not have to be on the same machine. If the CLI cannot open one, a remote shell for example, it prints the address instead. Open it on any device, sign in, and confirm the code.

Check who you are signed in as, or sign out:

distil whoami
distil logout

The CLI keeps its own session, in ~/.config/distillabs/session.json (under $XDG_CONFIG_HOME when that is set), and refreshes it for you. A run spanning several hours will not ask you to log in again.

Sign in to the dashboard

Go to app.distillabs.ai and sign in with the same account.

The dashboard is read-only. It lists your models, datasets and traces, and you start every job from the CLI or the REST API. See Ways to use distil labs.

Authenticate with the REST API

The REST API takes a bearer token from the CLI. Sign in to the CLI first, then print a token:

distil access-token

Each run refreshes the session and prints a new access token, and only the token, so it drops straight into a header:

curl -H "Authorization: Bearer $(distil access-token)" https://api.distillabs.ai/datasets

Access tokens are short-lived, much shorter than a training run, so fetch one per request rather than holding it:

response = requests.get(
    "https://api.distillabs.ai/datasets",
    headers=auth(),
)

Using the REST API has the auth() helper that calls distil access-token, along with the rest of the helpers the API examples depend on. Do not copy the auth block on its own. The surrounding helpers handle the error bodies where validation failures actually live.

Troubleshooting

If the API rejects a token, fetch a new one with distil access-token rather than reusing the old one. Sign-in errors and their fixes: Troubleshooting.

Next

You have an account. Now build a model, or pick your task type first with Task selection.