Concepts

Understand the resources and mechanics behind AIM Engine.

• AIM Services — Resolution shapes, overlays, caching, status

• AIM Models — The four model flows

• Profiles — Self-contained runtime configurations

• Inference Engines — vLLM and vLLM-Omni runtime behavior

• AIM Profile Sets — Derivation engine

• Model Sources — Auto-discovery from container registries

• Runtime Configuration — Storage defaults, routing, environment resolution

• Model Caching — Cache hierarchy, ownership, deletion behavior

• Accelerator Detection — How AIM Engine sees GPUs and CPUs

• Resource Lifecycle — Ownership, finalizers, deletion behavior