Building a Private BI Dashboard That Runs Entirely in the Browser

You can build a full BI dashboard — KPIs, time-series charts, top-N tables — that runs entirely in a browser tab. Drop a CSV or Parquet file into Zro7 Local BI Dashboard, write SQL against it with DuckDB-WASM, and Zro7 renders the results as charts using Recharts. No Metabase install, no Snowflake bill, no data leaving your laptop.

The architecture, in three pieces

1. DuckDB-WASM — the query engine. Reads CSV/Parquet, runs analytical SQL in a Web Worker.
2. Recharts — a React chart library. Consumes the query result and renders bar/line/pie/area charts.
3. The Zro7 UI — a grid of panels; each panel is {sql, chart_type, x, y}. Persisted to localStorage.

Dashboard recipes

• KPI card — SELECT SUM(revenue) FROM sales WHERE date >= current_date - 30. Renders as a big number.
• Time-series — SELECT date_trunc('day', ts) AS d, COUNT(*) FROM events GROUP BY d ORDER BY d. Renders as a line chart.
• Top-N table — SELECT product, SUM(qty) AS units FROM sales GROUP BY product ORDER BY units DESC LIMIT 10.
• Funnel — nested CTE counting distinct users at each step; render as bar chart.

Why 'in the browser' matters for BI

• Compliance — audit exports, medical claims, payroll: legally can't leave the machine.
• Speed of iteration — no ETL, no warehouse loading, no BI-tool learning curve. Drop file → chart.
• Cost — a hosted Looker or Sigma instance runs into thousands/month. This tab is free.
• Air-gapped or client-site work — consultants often can't install anything; a browser tab always works.

Steps

1. Open Local BI Dashboard.
2. Drop CSV or Parquet files.
3. Add panels — one SQL query each — and pick a chart type.
4. Save the layout locally; re-open the tab tomorrow with the same charts.