cuGraph SQL API

DataFusion Nexus exposes NVIDIA RAPIDS cuGraph algorithms as SQL table functions over its Flight SQL server. Each cugraph_* function reads an edge relation, builds a GPU graph, runs an algorithm, and returns rows you can compose with ordinary SQL.

This section is the self-describing contract for discovering and calling those functions — the same metadata an agent reads at runtime, rendered for humans.

The call convention

Every algorithm follows one positional shape:

cugraph_<algorithm>(table_name [, src_col, dst_col [, weight_col [, options_json]]])

• table_name — an edge relation (a registered table or view of edges).
• src_col / dst_col — endpoint columns; default to src / dst.
• weight_col — optional edge weight column. Whether it affects results depends on the algorithm's weight policy (see the cheat sheet below).
• options_json — a JSON string of algorithm and graph options.

A few functions take extra positional arguments (for example, cugraph_bfs takes a source vertex); see each function's page for its exact signature.

Algorithm families

• Centrality — importance scores (PageRank, eigenvector, Katz, HITS, betweenness).
• Community Detection — partition vertices into communities (Louvain, Leiden, ECG, spectral).
• Traversal & Paths — BFS and single-source shortest paths.
• Connectivity — connected components and minimum spanning tree.
• Structure — degrees, triangles, core numbers, k-core.
• Layout — 2-D layout coordinates.

Recommended workflow

1. List — cugraph_list_algorithms() to find enabled functions.
2. Describe — cugraph_describe_algorithm('<fn>', 'verbose') for arguments, options, schemas, and limitations.
3. Validate — cugraph_validate_call('<fn>', '<table>', '<call_json>') to statically check a call.
4. Execute — run the call once validation returns valid = true.

See Discovery & validation for details, and Connecting to start a server and client.

All functions at a glance

Algorithm	SQL function	Family	Source policy	Weight policy	Output kind	Options	Summary
Balanced Cut Clustering	cugraph_balanced_cut_clustering	Community Detection	none	ignored	fixed	many	Cluster vertices with balanced-cut spectral clustering.
Betweenness Centrality	cugraph_betweenness_centrality	Centrality	none	ignored	fixed	many	Compute exact or explicitly seeded/k-sampled approximate betweenness centrality scores.
BFS	cugraph_bfs	Traversal & Paths	required_scalar_or_list_or_relation	ignored	default_mode	rich	Multi-source-capable breadth-first search; source selectors must contain at most one source per connected component.
Core Number	cugraph_core_number	Structure	none	ignored	fixed	few	Compute graph core numbers.
Degrees All	cugraph_degrees_all	Structure	none	ignored	fixed	none	Compute in-degree and out-degree for every vertex.
ECG	cugraph_ecg	Community Detection	none	ignored	fixed	many	Detect communities with ensemble clustering for graphs.
Edge Betweenness Centrality	cugraph_edge_betweenness_centrality	Centrality	none	ignored	fixed	few	Compute exact or explicitly seeded/k-sampled approximate edge betweenness centrality scores.
Eigenvector Centrality	cugraph_eigenvector_centrality	Centrality	none	ignored	fixed	few	Compute eigenvector centrality scores.
ForceAtlas2	cugraph_force_atlas2	Layout	none	optional	fixed	many	Compute ForceAtlas2 two-dimensional layout coordinates.
HITS	cugraph_hits	Centrality	none	ignored	fixed	few	Compute HITS hub and authority scores.
In Degrees All	cugraph_in_degrees_all	Structure	none	ignored	fixed	none	Compute in-degree for every vertex.
K-Core	cugraph_k_core	Structure	none	ignored	fixed	few	Return the graph k-core edge set.
Katz Centrality	cugraph_katz_centrality	Centrality	none	ignored	fixed	few	Compute Katz centrality scores.
Leiden	cugraph_leiden	Community Detection	none	ignored	fixed	few	Detect communities with Leiden optimization.
Louvain	cugraph_louvain	Community Detection	none	ignored	fixed	few	Detect communities with Louvain modularity optimization.
Minimum Spanning Tree	cugraph_minimum_spanning_tree	Connectivity	none	required	fixed	none	Compute a minimum spanning tree.
Out Degrees All	cugraph_out_degrees_all	Structure	none	ignored	fixed	none	Compute out-degree for every vertex.
PageRank	cugraph_pagerank	Centrality	none	optional	fixed	few	Compute PageRank scores.
Personalized PageRank	cugraph_personalized_pagerank	Centrality	none	optional	fixed	many	Compute personalized PageRank scores.
Spectral Modularity Maximization	cugraph_spectral_modularity_maximization	Community Detection	none	ignored	fixed	many	Cluster vertices with spectral modularity maximization.
SSSP	cugraph_sssp	Traversal & Paths	required_scalar	optional	fixed	few	Compute single-source shortest paths.
Strongly Connected Components	cugraph_strongly_connected_components	Connectivity	none	ignored	fixed	none	Compute strongly connected components.
Triangle Count All	cugraph_triangle_count_all	Structure	none	ignored	fixed	none	Compute triangle counts for every vertex.
Weakly Connected Components	cugraph_weakly_connected_components	Connectivity	none	ignored	fixed	none	Compute weakly connected components.