Query cells against an explicit layer
Like recall, but against an explicit semantic layer (L0/L1/L2). Returns a JSON array of fully hydrated knowledge nodes in rank order.
Authorizations
The API key issued for the calling agent. Required on every endpoint except GET /healthz.
Body
The body for POST /v1/query.
The query text to embed and query against.
Semantic hierarchy. L0 = raw cells, L1 = scene centroids/patterns, L2 = architectural decisions/constraints.
L0, L1, L2 The maximum number of nodes to return.
x >= 1Structural filter for recall/query. source_agent and tenant_id here are client-convenience filters; the enforced read boundary is applied separately by the server.
Response
Hydrated nodes in rank order (nearest-first).
Always set; default in V1.
The cell's embedding vector.
What a graph node represents. Serialized as an externally-tagged enum: the unit variant Scene is the bare string "Scene"; the data-carrying variants are single-key objects (Cell, Agent, Beacon).
Scene The nine cell types the energy and promotion models are calibrated against.
Decision, Constraint, Risk, Pattern, Lesson, Fact, Preference, Context, Task Constrained to [0.0, 1.0].
0 <= x <= 1The agent that produced this cell.
Semantic hierarchy. L0 = raw cells, L1 = scene centroids/patterns, L2 = architectural decisions/constraints.
L0, L1, L2 The owning scene, if any.
Supersession pointer; set once this cell has been abstracted into a higher-layer Pattern. null = live.