Procedures

This reference is designed for both human developers and AI coding agents. Every entry includes complete CALL syntax and return columns that can be used directly in queries.

140+ CALL procedures: Database Introspection (25), Graph Algorithms (37), Embedding (6), Knowledge Graph (9), Temporal (9), Live Queries (5), System (21), Spatial System (3), GPU (4), Observability (2), Health (3), Auth (7), Extensions (6), and Behavior Graph (3). All return tabular results.

CALL db.procedures()

Returns the full list of available procedures (column: name).

Database Introspection (25)#

Core procedures for inspecting database state, schema, and metadata.

-- Get a schema overview
CALL db.schema()

Returns one row per label with property keys and node count, plus relationship patterns.

-- Run diagnostics
CALL db.doctor()

Returns rows for each check (node_count, relationship_integrity, index_consistency, constraints, generation) plus a summary row with HEALTHY or ISSUES_FOUND status.

Graph Algorithms (37)#

Centrality (5)#

CALL algo.pageRank()

Returns one row per node, sorted by rank. Uses 20 iterations with damping factor 0.85.

Community Detection (7)#

CALL algo.louvain()

Returns one row per node with hierarchical community assignment.

Graph Metrics (6)#

CALL algo.triangleCount()

Returns a single row with the total number of triangles.

Path Analysis (5)#

-- Find confidence-weighted shortest path between node 1 and node 5
CALL algo.confidencePath(1, 5)

Returns the path as "id1 -> id2 -> id3", total cost, and hop count.

Similarity and Spatial (7)#

Vector Search and RAG (7)#

-- Vector similarity search with a query vector
CALL algo.vectorSearch([0.1, 0.9, 0.3])

Returns nodes ranked by cosine similarity to the query vector.

-- Trusted RAG pipeline
CALL algo.graphRAGTrusted()

Returns nodes with trusted_score, observation class (observed > inferred > predicted), filtering low-trust paths.

Embedding Algorithms (6)#

Generate node embeddings for downstream vector search, classification, and similarity tasks.

-- Embed all nodes with 128-dimensional Node2Vec vectors
CALL algo.node2vec(128, 80, 10)
 
-- Find stale embeddings after property updates
CALL algo.staleEmbeddings()

Knowledge Graph Algorithms (9)#

Algorithms for reasoning over facts, confidence, and semantic relationships.

Physical AI Algorithms (3)#

Procedures for multi-agent world models: sensor reliability scoring, multi-agent belief fusion, and multi-sample weighted-centroid fusion.

-- Check sensor reliability after a calibration run
CALL algo.sensorReliability('lidar-front')
YIELD accuracy, drift_rate, sample_count, calibration_recommended

Returns mean confidence of all Entity nodes attributed to lidar-front, plus a calibration flag when accuracy < 0.85.

-- Fuse two agents' conflicting observations of 'player-07' into one inferred fact
CALL algo.beliefReconcile('player-07')
YIELD fused_confidence, source_agents, reconciliation_method

Scans all [:BELIEVES] edges ending at the entity, computes precision-weighted Kalman mean, writes the result as _observation_class: 'inferred' on the entity. See Fleet Coordination guide for the full multi-agent pattern.

-- Find entities that may refer to the same person
CALL algo.entityResolution()

Auth & Governance (7)#

API key management and audit log procedures.

Temporal (9)#

Procedures for temporal queries, clock management, and change tracking.

-- See recent mutations with timestamps
CALL db.mutations()

Returns rows with sequence number, operation description, and timestamp.

-- Point-in-time query (query modifier, not a CALL procedure)
MATCH (n:Person) AS OF seq N
RETURN n.name

Note: AS OF is a query modifier, not a CALL procedure.

Live Queries (5)#

Procedures for live queries: topics, subscriptions, live views, and causal chains.

Topic event sequences are strictly monotonic per topic — retention pruning never re-uses a sequence number. See Live Queries → Topics for the retention policy and subscription cursor replay surfaces.

CALL db.causalChain()

Finds all :Action nodes and traces outgoing CAUSES relationships up to depth 10, returning each hop with confidence scores.

System (21)#

Procedures for backend management, validation, observability, and storage internals.

-- Check available GPU backends
CALL db.backends()

Returns one row per backend (cpu, metal, cuda) with availability and status message.

-- Query by observation class
CALL db.nodesByObservation.observed()

Returns all nodes classified as "observed" with their confidence scores.

Spatial System (3)#

Coordinate frame management and spatial query observability.

-- Inspect current coordinate frame
CALL db.spatialMetadata()
 
-- Check dispatch lane after a spatial query
CALL arcflow.spatial.dispatch_stats()

GPU (4)#

GPU residency and capability inspection. The engine routes algorithms to GPU automatically when available; these procedures expose the routing decisions and stack details.

-- Confirm GPU stack is fully wired before running large-graph algorithms
CALL db.gpuStack()

Observability (2)#

OpenTelemetry policy controls. Span recording is off by default; enable lite or full to surface execution traces in your collector.

Health (3)#

CALL db.replicationStatus()

Returns the current replication mode (standalone/primary/replica), replica count, and whether writes are enabled.

Extension Procedures (6)#

Additional domain-specific procedures.

-- Register a drone agent
CALL swarm.register('drone_01', 'drone')

Behavior Graph (3)#

Behavior tree procedures backed by the world model. Nodes labeled :BehaviorNode form the tree; CHILD relationships define structure.

-- Tick a behavior tree named "patrol"
CALL behavior.tick('patrol')

-- List all behavior nodes
CALL behavior.nodes()

See Also#

• Built-in Functions -- 93 scalar functions (17 math, 14 aggregation, 24 string, ...)
• EXPLAIN -- query plan introspection
• PROFILE -- query execution profiling
• RAG Pipeline Guide -- building RAG with ArcFlow