Skip to content

Start here

What Dashies is

The infrastructure that keeps an AI-built HTML dashboard alive - refreshing on a schedule, versioned, and shareable by link.

Dashies is the agentic BI platform for publishing, refreshing, and sharing AI-built HTML dashboards.

The one sentence worth remembering: a Dashies dashboard re-runs its own SQL on a schedule, with no AI in the loop. Your AI tool authors the dashboard once. The server compiles it, validates it, and runs the query once to fill it with real numbers. From then on a scheduler re-runs that same query and rewrites only the numbers, forever, without a model being asked anything.

The problem this exists to solve

People who work in BI already have plenty of AI in their workflow. BI itself is not there yet. Legacy BI tools are hard to drive from a plugin, and AI is good at writing HTML rather than at assembling a workbook.

So the obvious move is to ask an AI for a dashboard as a self-contained HTML file. That works exactly once. The file is right on the day it is handed over and stale the day after, because a hand-built HTML file cannot re-run its own query. Every update is another prompt, and every prompt is another chance for a metric to come out defined differently than it was last month.

Dashies is the part that was missing: the infrastructure that keeps the file alive. Your AI writes the dashboard; Dashies keeps it refreshing, versioned, and shareable by link.

What you get

  • A URL. A published dashboard lives at https://<workspace-slug>.dashies.ai/<slug>.
  • Numbers that update on their own. Pick a cadence and the scheduler re-runs the query and rewrites the data in place. See Schedules.
  • A design that survives every update. Refresh replaces one block of JSON inside the file and touches nothing else, so your layout, styling, and copy are never regenerated. See The data island.
  • Version history. Every republish snapshots the previous body, so you can preview it and roll back.
  • A refusal instead of a wrong number. Dashies rejects a dashboard that would add up a measure that cannot be added up, rather than publishing a plausible wrong figure. See Measure correctness.

What Dashies is not

It is not a dashboard editor. There is deliberately no authoring UI in the web app: you do not drag a chart onto a canvas. The web app manages, schedules, serves and shares what your AI published. Authoring happens in your AI tool.

It is not a query-time BI server. The incumbents re-query the warehouse on every view, so a bad cache heals itself the next time someone loads the page. Dashies precomputes and ships static bytes, so a published dashboard holds the only copy of those numbers. That is what makes it fast and cheap to view, and it is also why the correctness rules in Core concepts are enforced at publish time rather than left to the author to remember.

The three steps

  1. Install the Dashies plugin, which bundles the authoring skill and the publish service into your AI tool.
  2. Connect a data warehouse: PostgreSQL, BigQuery, Snowflake, Redshift, Databricks, SQL Server, or Oracle Database. Or skip this and use the built-in self connection.
  3. Ask your AI for a dashboard.

Next

Is Dashies right for you? states the fit and the anti-fit before you invest in step 1. How it works is the architecture behind the three steps. Or go straight to the Quickstart.