How I coded an SSE streaming AI Copilot & 3D canvas entirely from my smartphone during factory line shifts

Hey everyone,

I’m a completely self-taught independent developer. For the past few months, I’ve been coding CoSpatial 3D (https://cospatial3d.xyz)—a%E2%80%94a) mobile-optimized spatial studio and generative concept engine—entirely from my smartphone during my 15-minute breaks and lunch hours while working intensive manual labor shifts on a production line.

I finally got our production deployment stable over our custom domain, but as an amateur coder, the biggest technical roadblock I hit was UI throttling. Whenever I asked our in-app AI Copilot to fix complex 3D math scripts or rewrite long files, the backend server would hit default HTTP timeout limits, buffer, and completely crash the mobile browser window.

To fix this before launch so other self-taught builders can bypass roadblocks smoothly, I ripped out the standard request-response loops and hardcoded Server-Sent Events (SSE) streaming on my Python backend, using this fetch reader loop on my React frontend to consume the chunks token-by-token:

const handleFetchStream = async (prompt) => {
  const response = await fetch('/api/copilot/stream', {
    method: 'POST',
    headers: { 'Content-Type': 'application/json' },
    body: JSON.stringify({ prompt })
  });
  const reader = response.body.getReader();
  const decoder = new TextDecoder();
  let done = false;
  while (!done) {
    const { value, done: readingDone } = await reader.read();
    done = readingDone;
    const chunk = decoder.decode(value, { stream: !done });
    setChatHistory((prev) => prev + chunk);
  }
};

The streaming typewriter effect completely fixed the timeouts, and the 3D viewport canvas is now executing raw trigonometry to draw accurate geometric shapes (like perfect 5-point stars) instead of blurry AI hallucinations.

Here is my current business hurdle: I am completely bootstrapping this project out of my own pockets from my factory paychecks. Because running high-intensity image generators and cloud rendering machines hits external APIs that charge me credits out of pocket, I’ve had to strictly sandbox our free trials inside closed, solo workspaces to protect our baseline balances. Upgrading to our paid Pro tier ($10/mo) is what actually provisions our server lines to unlock our multi-user live pair-programming channels via WebSockets.

I don't want to sound like a desperate corporate ad, but I also won't pretend I don't need your financial support to keep this project alive, pay our hosting bills, and secure our GPU tokens for next month. If you believe in backing raw, independent indie development over multi-million dollar corporations, skipping the trial and locking in a $10 subscription directly keeps our engine alive.

For the senior engineers here: I’m wide open to brutal, honest feedback on our UI responsiveness, our data isolation rules, or our prompt expansion filters. Jump into the sandbox at https://cospatial3d.xyz on your mobile devices and let me know how the streaming text loads on your screen sizes!