Hi everyone,

I'm currently optimizing the infrastructure for an interactive astrology platform hosted on a Proxmox VE setup. We are testing a heavy API integration for a feature called chat gpt astrology free via Astroma, which handles real-time conversational data and charts.

Currently, I have this running inside a Debian LXC container (allocated 4 vCPUs, 8GB RAM), but during peak traffic spikes, CPU usage hits a massive bottleneck due to the nested API requests and concurrent user sessions.

I am considering a few options to stabilize the performance:

1. Moving the core node application to a dedicated VM with KVM instead of an LXC container to isolate resources better.
2. Setting up an Nginx reverse proxy with aggressive micro-caching on a separate lightweight container ahead of the main app.

Has anyone here optimized a similar setup for heavy, real-time AI API workloads on Proxmox? Would you recommend migrating away from LXC to a full VM for better thread handling under heavy traffic spikes, or is tweaking the reverse proxy layers the smarter move here?

Appreciate any insights or architecture advice!