How it works
The path a dashboard takes from your AI tool through compile, validate and seed, to a scheduled refresh that rewrites only its data.
The pipeline
Your AI tool talks to the Dashies publish service, which compiles a spec into a dashboard, validates it, runs the SQL once to seed it, and stores the result. A scheduler then re-runs that SQL on your cadence and rewrites the data the page carries.
| Stage | What happens |
|---|---|
| Author | Your AI writes a spec: datasets, tiles, and a source. |
| Compile | The server turns the spec into a dashboard file. |
| Validate | Structural and correctness rules run before anything is stored. |
| Seed | The SQL runs once so the published page has real numbers. |
| Refresh | A scheduler re-runs the SQL and rewrites only the data. |
Why only the data is rewritten
A published dashboard is one self-contained HTML file. The numbers live in a single data island inside it. Refresh replaces that island and nothing else, which is what makes the design survive every refresh.
What is on this page later
- The full architecture, end to end, with each component named.
- Island versions and what changed between them.
- Where the runtime comes from, and why it is never persisted.
Placeholder
This page is a skeleton. The written content lands in a later wave.