Great use case for edge functions and plugin systems. How does the startup latency compare to running the same code in a V8 isolate like Cloudflare Workers?
Rendering Office XML directly to Canvas is clever — avoids the heavy DOM tree for large documents. Does it handle embedded media (images/charts) or just text + formatting for now?
The lesson-style README is a great approach. Breaking down LLM inference into digestible steps makes the codebase approachable even for people who haven't touched CUDA before.