About generative-bi-dashboard
AI dashboard builder: drop in a spreadsheet, data never leaves the browser.
Title Generative BI Dashboard: AI dashboard builder that keeps data in the browser (full source code) Short pitch Drop in a CSV or Excel file and get an editable dashboard in under a minute. The AI suggests which dashboards fit the data, builds the one you pick, and writes a plain-language analysis under every chart. Unlike most "chat with your data" tools, the file never leaves the browser: queries run locally in DuckDB-Wasm and the AI only sees column statistics and aggregated results. What it does Reads CSV, TSV, Excel, JSON and Parquet files. Suggests four dashboards tailored to the file, or builds one from a written description. Thirteen tile types: headline numbers with comparison and trend line, bar (grouped, stacked, horizontal), line, area, combo, pie, treemap, funnel, waterfall, histogram, heatmap, scatter and table. Dashboard-wide filters, click-to-filter on charts, and multi-level drill-down. Works with the buyer's choice of LLM: Google Gemini, Anthropic Claude, OpenAI, or any OpenAI-compatible endpoint including self-hosted models. Drag-and-resize grid, point-and-click tile builder with a SQL mode, and an assistant that edits the dashboard from plain-language requests. Written analysis per tile and a summary per dashboard. Dashboards and data are saved in the browser. No accounts, no database, no server-side storage. Why it is interesting to a buyer Privacy by design is a real selling point for finance, HR and healthcare teams who cannot upload spreadsheets to a third-party AI service. Cheap to run: no database and no data pipeline; the only running cost is LLM calls on small prompts. Clear next steps: a written roadmap (export and sharing, multi-file joins and calculated fields) and a codebase organised to extend. Stack Next.js 16, React 19, TypeScript, Tailwind CSS 4, shadcn/ui, Vercel AI SDK, DuckDB-Wasm, Recharts, react-grid-layout, Vitest, GitHub Actions. What the buyer gets Full source code and git history, with all rights assigned. Tests, CI workflow, README, handover guide and roadmap. A list of third-party licences (all permissive: MIT, Apache-2.0, ISC, BSD). What the buyer does not get No users, revenue, domain or hosted deployment. No support after the sale. The interface has not been through a full browser test pass; see the handover guide for the exact state of testing. GitHub Stats (at time of listing): 0 stars 0 forks 0 open issues Last commit: October 9, 2026
Project details
Built with: TypeScript,CSS,JavaScript
Website: https://github.com/Chicoo000/generative-bi-dashboard
Video 1: https://vimeo.com/1234580903?share=copy&fl=sv&fe=ci
Next step: <p>Roadmap in the GitHub repo. Right now, we are entering phase 3, and I haven't had time to start. </p><p><br></p><p>Phase 3 — Layout and presentation</p><p><br></p><p>- Text and heading tiles, and tabs or sections within a dashboard.</p><p>- Light and dark mode toggle.</p><p>- Export a tile as PNG and a dashboard as PDF; export and import a dashboard as a JSON file.</p><p>- Undo and redo.</p><p><br></p><p>Phase 4 — Richer data</p><p><br></p><p>- Several files in one workspace, with suggested or user-defined joins.</p><p>- Calculated fields (for example `variance = amount - budget`).</p><p>- Period-over-period helpers: versus previous period, year to date, moving average.</p><p><br></p><p>Not planned yet</p><p><br></p><p>User accounts, server-side storage, scheduled refresh and live database connections.</p>