Forge Pool

Appearance

VMem ​

VMem is Forge Pool's heavier shared working memory surface for reusable execution payloads.

Where KV handles lightweight metadata and coordination state, VMem is oriented toward larger intermediate assets tied more directly to compute workflows.

Typical Uses ​

VMem is suitable for:

  • reusable baselines
  • intermediate tensors
  • stored distributions
  • parameter grids
  • cached execution surfaces
  • workflow handoff payloads

Position in the System ​

VMem is part of the execution platform, not a primitive family.

It supports stateful or staged workflows on top of the otherwise clean canonical execution surface.

Example Endpoint Overview ​

http
POST   /v1/vmem/store
POST   /v1/vmem/load
DELETE /v1/vmem/delete

Why VMem Exists ​

Not every workload should recompute every intermediate asset from zero.

VMem gives Forge Pool a controlled shared memory layer for reusable compute-adjacent payloads while preserving the architectural distinction between:

  • execution contract
  • replay trace
  • persistence surfaces
  • higher-order workflow composition