Products

Runtime and Fleet

Source

Runtime and fleet is Eden's distributed runtime and infrastructure coordination layer. It turns local machines, private servers, cloud GPUs, and managed capacity into one governed fleet for AI and other eligible workloads.

Responsibilities

Runtime and fleet provides:

  • Fleet membership and node health
  • Model, runtime, hardware, and protocol compatibility
  • Placement and route planning
  • Dispatch and execution state
  • Session and connection lifecycle
  • Permission-scoped shared model memory
  • Predictive state hydration
  • Capacity allocation and marketplace lease metadata
  • Runtime, usage, and cost telemetry

Gateway boundary

Runtime and fleet is not an unauthenticated public model endpoint. Client requests enter through an Eden gateway. The gateway authenticates the caller, applies policy, resolves an eligible route, and streams the result. The runtime control plane coordinates the fleet but is not a public inference hop.

Each node has a separate identity from the gateway and control plane. Fleet membership, capability, compatibility, allowed gateway principals, and lease scope are checked before a node can serve work.

Deployment choices

Organizations can enroll:

  • Developer machines running owner-authorized local runtimes
  • Private Linux GPU servers
  • Cloud GPU instances
  • Managed open-source capacity

Model weights, inference payloads, and cache values can remain in customer-controlled infrastructure. Eden coordinates the operating layer and records safe control-plane evidence.

Help improve Eden Docs

Find something unclear or incomplete? Review the source and propose an update.

View on GitLab Updated July 27, 2026