This tutorial shows the shortest path through Emolysis: a local file, a business question, a tested gold table, a dashboard and a link. You do not need a data platform for any of it.
What you need
- The Emolysis desktop app.
- Your own key from the model provider you choose, on any plan, or a paid plan's model usage.
- A CSV of orders with an order date, a region, an amount and an order id. Any invented data works.
- An Emolysis account. The Free plan is one and is enough to build the dashboard; publishing, in the last steps, needs a paid plan.
Step 1: Create a project
Open the app and choose New project, or run emo init orders in a terminal. Emolysis creates a folder with models/, pipelines/, tests/ and an emolysis.yml. For a local-only project, the platform can stay unset until you connect one. The projects docs describe the layout.
Step 2: Add the file
Drop orders.csv into the project. Emolysis works on local files, CSV and Parquet, on DuckDB alone, with no platform connected; you stay signed in to your Emolysis account, as every use of Emolysis needs. The app shows the file's columns and types as DuckDB reads them. Check that the order date is read as a date and the amount as a number; if not, say so and the agent casts them.
Step 3: Ask the question
Type: "Monthly revenue and order count by region, for the last 12 months." The agent reads the schema and proposes a semantic model:
- Measures: revenue (sum of amount), orders (count of distinct order id).
- Dimensions: month (from order date), region.
- Grain: one row per month per region.
- Filter: the last 12 full months.
Read it carefully. If amounts include refunds as negative rows, or if revenue should exclude tax, tell the agent now. This is where wrong numbers are easiest to prevent.
Step 4: Watch the gold table get built
The agent picks the Aggregation / gold shape and fills its template with the slots from the semantic model. Before anything else, the template's golden tests run on DuckDB, using input with known totals. You see the SQL, the test results and, if a test fails, the fix the agent makes before it runs them again. Then the pipeline runs on your file and writes the gold table to the project's local lakehouse.
Step 5: Look at the dashboard
The agent chooses charts that fit the table. A typical first draft:
- A KPI for revenue over the period, with a small trend line.
- A KPI for orders.
- A line chart of monthly revenue by region.
- A table of revenue and orders by region.
Change anything you want: a chart type, a title, the number format (compact figures, currency, fixed decimals), the order of regions. You can also ask for another view, such as "revenue share by region as a pie". The sample dashboard shows the kind of charts to expect, on synthetic data.
Step 6: Sign in
Publishing needs an account and a paid plan. Choose Publish. If you are not signed in, the app shows a short code. Open the link, sign in with your GitHub, Microsoft or Google account and confirm the code. The code expires after 15 minutes.
Step 7: Choose who can see it
Pick a visibility:
- Link: anyone with the link can open it.
- Organization: signed-in members of your Emolysis organization. This is the default.
- Only me: only you.
The setting is fixed for the link; to change it later, revoke the link and publish again, which makes a new link with a new URL.
Then publish. The app returns the dashboard's stable link on emolysis.com, in the form emolysis.com/d/ followed by the dashboard's id. The sharing docs cover revisions and revoking.
What leaves your machine
Two things, and it is worth knowing both. While the agent works, its prompts go to the model provider you chose, and they can include the file's column names and sample values. With a paid plan's model usage, they pass through Emolysis's metering service on the way; that service is designed not to store or log their content. When you publish, and only then, the dashboard's charts and the gold table behind them are uploaded so the link can be opened. Your CSV itself is not published. See the security page for the full list.
Next steps
- Publish again after a change: it adds a revision under the same link, so the people you sent it to see the change.
- Connect a platform and build the same gold table on real data. The SCD Type 2 and CDC merge previews show how pipelines deploy.