Topology
Every capability is a node.
The Resource Graph maps models, tools, memory stores, and attention budgets. Scheduling and cognition route through topology — not ad-hoc function lists.
Why a graph
Lists do not capture contention. Graphs do. When two plans compete for the same encoder, the scheduler sees the edge.
Nodes
Models, tools, corpora, twin instances, human approvers.
Edges
Affinity, denial, cost, latency, residency constraints.
Policy
Read-only edges cannot promote to write without a new grant.
Operator view
Inspect hot paths, starved nodes, and policy blocks before a cognition run commits budget.
Graph facets
Same underlying store — multiple projections.
Capacity heatmap
Where budgets concentrate.
Policy overlay
Deny and residency edges.
Cost projection
Edge-weighted spend foresight.
Residency map
Region and tenancy constraints.