Quickstart

Create your account

Go to app.datachef.com and sign up with your work email, Google, or your company's SSO.

If you sign up with email and a password, Datachef sends a confirmation link. Your account only exists after you click it. The link is valid for 24 hours.

If your company already uses Datachef and has verified its email domain, you'll join its organization automatically, or after an admin approves you, depending on how it's set up. Some organizations only admit people they've invited. If no organization has verified your domain, you start a new one. See people and roles for how joining works.

Bring in some data

A new organization has nothing in it yet, so the Get started page offers two ways to fill it:

• Load the sample dataset. One click, no credentials. You get three synthetic tables: sample.customers (two years of signups), sample.opportunities (eighteen months of pipeline on either side of today) and sample.usage_events (the last 90 days of product usage). The data is generated relative to today, so trend queries show real trends.
• Connect a real source. See connections.

The sample is an ordinary connection. You can keep it next to your real sources and delete it later from its own connection page.

Run your first query

The tour hands you a query written against the tables you actually have. Run it in the SQL editor with Run (or ⌘↵).

SELECT date_trunc('month', signed_up_at) AS month, count(*) AS signups
FROM sample.customers
GROUP BY 1
ORDER BY 1

Turn the result into a chart

Every result has a Chart tab next to its table. Datachef picks a chart from the shape of the result. Choose Customize to change it. See charts.

Save it as a model

Choose Save as model and give it a name such as monthly_signups. From now on, anyone in the workspace can SELECT * FROM monthly_signups, it shows up in autocomplete, and the agent uses it when answering questions. See models.

Ask the same thing in natural language

Open the Agent page and type your question in natural language, in whichever language you work in. On the sample dataset, the page suggests prompts to start with. See asking questions.

Each step on the checklist ticks itself when the work exists in your workspace, so your progress carries over to a new browser. Get started stays in the sidebar until you finish or skip it.