Problem: A client wanted to replace a proprietary BI tool without losing visual query building, multi-tenant isolation, or real-time analytics. The alternatives on the market were either per-seat licensed or locked to proprietary data formats.

Approach: We designed and deployed a complete BI platform: Cube.js as the semantic layer, React for the dashboards, PostgreSQL for storage. Non-technical users build their own views through the visual query builder.

Technical details:

• Cube.js semantic layer, so a metric is defined once and used everywhere
• React frontend with a drag-and-drop dashboard builder
• Multi-tenant architecture with per-client data isolation
• Real-time analytics over WebSocket connections
• PostgreSQL with tuned query patterns
• Docker Compose deployment with health checks and automated backups
• Deployment and configuration documented in a README

Outcome: The platform replaced the licensed tool, so the per-seat cost is gone. Team members outside the data function now build their own dashboards.

Stack: Cube.js, React, PostgreSQL, Docker Compose, WebSocket

Sandboxed plugin ran for projects/bi-platform and read “Universal BI Platform” through the mediated API.