Serve on localhost and chat with the AI
Run dac serve to deploy your dashboards on localhost:8321, then use the in-browser AI chat to add new charts by typing what you want in plain English.
What you'll do
Boot dac serve, open the live dashboard in your browser at http://localhost:8321, and use the built-in AI chat to add a new chart with a natural-language prompt. Every change you make through chat lands as a YAML edit on disk - reviewable, diffable, and committable.
Why this matters
This is the payoff. Every step before this was setup; once dac serve is running, you have a fully interactive analytics app powered by your YAML, watched by a live-reload daemon, and editable by a built-in agent.
Instructions
1. Start the dev server
dac serve --dir . --open
DAC starts on port 8321 (the next free port if 8321 is taken) and opens your browser. You'll land on the dashboards list (the screenshot at the top of this page). Click any dashboard and the widgets render live against your connection.
2. Edit a YAML, watch it reload
Change the chart title or add a new filter option in dashboards/sales.yml and save. DAC streams the change to the browser over SSE - no refresh needed.
3. Chat with the AI assistant
The right panel of the dashboard view has a chat. Type something like:
Add a pie chart of revenue by channel below the existing charts.
The agent reads dashboards/sales.yml, proposes a YAML diff, applies it, and live-reload paints the new chart. Because it's editing the YAML directly, the change shows up in git diff like any code change.
Tip
Tune the agent's reasoning depth with dac serve --agent-effort low|medium|high. Lower effort is faster; higher effort handles vaguer prompts and multi-widget changes more reliably.
4. Ship it
When you're ready to share a snapshot:
dac build --dashboard "Sales Overview" --output dist/
This produces a single dist/index.html with all query results baked in - no DB needed at view time. Drop it on S3, Netlify, or any static host.
For a live, query-on-demand deployment, run dac serve --password <something> on a server that can reach your warehouse.
What just happened
You took a folder full of YAML and turned it into a live, AI-editable analytics app on your laptop. From here you can:
- Add a
semantic/model so multiple dashboards share a single revenue metric - Switch the connection in
.bruin.ymlto Postgres, Snowflake, or BigQuery and point the same YAML at production - Commit the project to git and let teammates iterate via PR review