Aura Workspace

About Aura Workspace

A portfolio-grade spatial business simulator by Fluxis Labs.

Architecture

The browser never talks to the simulation host directly. Every request passes through Vercel route handlers that validate input and output, keep secrets server-side, and forward work to a Python FastAPI engine running Monte Carlo math with NumPy and Pandas.

BrowserHUD + 3D sceneuseSimulationVercelDDoS + WAF + BotIDproxy.ts CSP nonceRoute Handler + ZodRender FreeFastAPI + guards1000 × 12 months
Browser to Vercel (WAF, CSP, Zod) to Render FastAPI engine, then back with validated results.

Monte Carlo model

Think of Monte Carlo as asking the same business question a thousand times with slightly different luck each time. Instead of one optimistic spreadsheet, you get a cloud of plausible futures and a clear sense of how often the plan works.

  1. 1

    You set the knobs

    Ad spend, price, team size, operating costs, and cash reserve describe how the SaaS firm plans to run for a year.

  2. 2

    The engine rolls the dice

    Each of the 1000 runs varies lead cost, conversion, and churn within realistic ranges, then steps month by month for 12 months.

  3. 3

    Capacity is a hard limit

    A fixed team can only serve so many clients. When demand overruns capacity, overload raises churn and squeezes profit.

  4. 4

    You see ranges, not a single guess

    The UI shows percentile bands, risk components, node health, and plain-language insights. The browser never invents replacement numbers if the link drops.

Technical challenges

The hard parts were not the 3D glow. They were trust, latency, and keeping every figure honest under failure.

  • Cold starts on free compute

    Render Free can sleep. The boot screen pings health, shows real wake telemetry, and only unlocks ENTER AURA after the engine answers.

  • Secrets stay server-side

    The browser never learns the Render URL or shared secret. Route handlers validate with Zod, attach the key, and return generic errors on failure.

  • Math never runs in the browser

    When the link is lost, the scene keeps the last Python result marked stale. No client-side fallback arithmetic is allowed.

  • Strict CSP and same-origin fencing

    Nonce-based CSP, BotID on simulate, and Origin checks keep the public demo from becoming an open proxy into the engine.

Links

Work with Fluxis Labs

Aura Workspace is the lead portfolio exhibit for Fluxis Labs - spatial simulation, WebGL craft, and production-grade security in one demo.

Contact on GitHub