Stacks/vLLM/Services

vLLM stack services

This page shows the capabilities provided by the services in the vLLM stack and the connections, configs, settings, cron schedules, actions, volumes, backups, imports, and other customizations attached to them for this specific stack.

Back to the vLLM stack
Stack serviceBase serviceVersionsRole
vLLM
vLLM0.29Primary
  • Additional services can be added or removed.
  • When one of the services in the stack updates, the stack can be upgraded to a new revision.
  • Stack changes can then be rolled out to environments on your schedule.

Settings

Tune service behavior per environment while keeping one reusable stack baseline.

vLLM
Hugging Face model
Default value: Qwen/Qwen3-0.6B.
vLLM
Model revision
Default value: c1899de289a04d12100db370d81485cdf75e47ca.
vLLM
API model name
Default value: model.
vLLM
Maximum context length
Default value: 4096.
vLLM
GPUs used by each replica
Default value: 1.
vLLM
Fraction of GPU memory used by vLLM
Default value: 0.9.
vLLM
Hugging Face token

Volumes

Provision the persistent storage and shared volume relationships the stack expects.

vLLM
Model cache
Size: 50 GB. Optional.