OptionalfetchFn: {Optionalinit: RequestInitOptionalinit: RequestInitOptionaloptions: { middleware?: Middleware[]; timeoutMs?: number }ReadonlyidShort identifier (e.g., "sparql", "cypher", "holonic").
ReadonlynameHuman-readable name (e.g., "SPARQL Endpoint").
Expand the neighborhood of a node at the given depth. Returns a UGM containing the discovered subgraph.
DEPTH IS HONORED OR REJECTED, never ignored. An implementation
that cannot express the requested hop count throws
AdapterArgumentError with argument: "depth" before issuing a
request, so a host learns at the call site instead of receiving a
shallower subgraph that looks complete. Of the shipped adapters,
SparqlAdapter, CypherAdapter and GremlinAdapter honor any
depth in [1, MAX_TRAVERSAL_DEPTH]; HolonicAdapter and
RestAdapter honor 1 and reject the rest, each for a reason
recorded on the method.
OptionaledgeTypes: string[]Optional filter to specific edge types.
Get full properties for a specific node. Used for lazy-loading detailed properties.
Get the schema of the data source.
Execute a query and return results as a new UGM. The query language depends on the adapter (SPARQL, Cypher, etc.).
Interface that all data source adapters implement.
ARGUMENT SAFETY.
nodeId,depthandedgeTypestypically come from host state (a node click, a search result, a server payload), so the remote adapters treat them as untrusted: values are bound through the protocol's own mechanism where one exists, and validated by./query-safetywhere it does not. A value that cannot be placed safely raisesAdapterArgumentErrorbefore any request is sent.query(q)is the exception and is NOT sanitized: its argument is query text by definition, so hosts must not build it from untrusted input.