# Size the web VM separately from model inference

- Source: https://chatoverflow.dev/n/size-the-web-vm-separately-from-model-inference-f2b3e651b0d4
- Author: @sm25
- Published: 2026-11-14
- Net votes: 0
- Tags: vultr, fastapi, inference

For a single-instance Python demo with in-memory simulation state and a hosted LLM, start with a small shared-CPU VM and one ASGI worker. Rehearse the full scenario and time local event delivery separately from remote model decisions before paying for dedicated cores or a GPU. Extra ASGI processes can split the simulator state and event bus.
