Status
In private development. Early access by invitation.
Dashboards are where most people start with Emolysis. You ask a business question; the agent answers with a draft dashboard at once, then builds and tests the table behind it. The BI page shows the idea, and the sample dashboard on this site draws the same kind of charts from synthetic data.
Ask
In the app, type a question or use /dashboard; from an editor or coding agent, ask through the MCP server.
A good question names the measure, the breakdown and the period: "Monthly revenue by region for the last 12 months, with order counts." The more of the three you give, the fewer questions the agent has to ask back.
What the agent does
- Find the sources. It searches your catalog, or the files in your project when you work locally, for tables that hold the measures and dimensions in the question.
- Propose a semantic model. It lists the measures (revenue, order count), the dimensions (month, region), the grain of the answer, the joins between tables and any filters. You confirm or correct it. This is the step that catches a wrong measure, grain or join, so read it. See semantic model, grain and fact table.
- Answer first. It picks charts that fit the question: a KPI for a headline figure, a line for a trend over time, a bar for a comparison across categories, a table for exact values. It draws them at once from illustrative rows generated from the definitions, so you see the shape of the answer before any table is built. The charts appear in the app, drawn from a chart spec that is the same in the app and in the browser.
- Bind each chart. It checks each chart against the tables in your lakehouse and links it to the Aggregation / gold template for your platform, often after Dedupe / conform for messy sources.
- Build the gold table. It fills the template and runs the golden tests on DuckDB.
- Go live. Once the gold table is tested and deployed, each chart switches from illustrative rows to the gold table's rows.
Why the semantic model comes first
A wrong dashboard usually comes from a wrong definition, not from wrong code: revenue counted before refunds, a join that doubles rows, a month that starts on the wrong day. Putting the definitions in front of you before any table is built is how Emolysis catches these early. The confirmed model also becomes the contract for the gold table, so the tests check totals at the grain you agreed.
Chart types
Bar, line, area, scatter, pie, KPI and table. Bars can be stacked or horizontal; lines and areas can be stacked. Numbers can be shown as compact figures, fixed decimals, percentages or currency. A KPI shows one figure and, when the data has a time dimension, a small trend line. See Chart types for what each type takes.
Edit
You can change chart types, titles, labels, number formats and ordering, or ask the agent for another view of the same table.
Local or on your platform
Without a platform, you can build a dashboard from local CSV or Parquet files on DuckDB, signed in to your Emolysis account like every use of Emolysis. With a connected platform, the gold table can be built from a sample while you iterate, then deployed to your platform like any other pipeline so the numbers come from the full data. See Deploying.
A dashboard built from a sample shows the sample's numbers. Deploy the gold table and refresh it before you share figures for people to act on.
Publish
When you are ready, publish the dashboard as a link. That is the moment it leaves your machine. Publishing shared dashboards is part of the paid plans; building them on your machine does not need a paid plan. See Sharing dashboards and Plans and pricing.
See an example
The sample dashboard on this site uses the same chart spec with synthetic data. It shows every chart type the agent uses.