NetworkX Backend Interface

nx-neptune serves as a NetworkX-compatible backend for Neptune Analytics. This enables you to offload graph algorithm workloads to AWS with no code changes — use familiar NetworkX APIs to seamlessly scale graph computations on-demand.

For more on NetworkX backends, see the NetworkX backends documentation.

How it works

You write standard NetworkX code
Add backend="neptune" to any supported algorithm call
nx-neptune transparently syncs your graph to Neptune Analytics, executes the algorithm there, and returns results in NetworkX-compatible format

import networkx as nx

# Build a small social network
G = nx.DiGraph()
G.add_edges_from([
    ("Alice", "Bob"),
    ("Alice", "Carol"),
    ("Bob", "Carol"),
    ("Carol", "Dave"),
    ("Dave", "Alice"),
])

# Runs on Neptune Analytics, returns standard NetworkX result format
scores = nx.pagerank(G, backend="neptune")

# Who's most influential?
top = sorted(scores.items(), key=lambda x: x[1], reverse=True)
for name, score in top:
    print(f"{name}: {score:.4f}")

Configuration

Set the NETWORKX_GRAPH_ID environment variable to point at your Neptune Analytics instance:

export NETWORKX_GRAPH_ID=your-neptune-analytics-graph-id
python ./nx_example.py

Or inline:

NETWORKX_GRAPH_ID=your-neptune-analytics-graph-id python ./nx_example.py

Backend registration

The backend is registered via the nx_plugin entry point in pyproject.toml. When you install nx_neptune, NetworkX automatically discovers the "neptune" backend.

# No explicit import needed — just use the backend parameter
result = nx.pagerank(G, backend="neptune")

Algorithm implementation pattern

Each algorithm follows this pattern:

from nx_neptune.utils.decorators import configure_if_nx_active

@configure_if_nx_active()
def algorithm_name(neptune_graph: NeptuneGraph, **kwargs):
    # Build openCypher query
    # Execute on Neptune Analytics
    # Transform results to NetworkX format
    return result

Supported algorithms

nx-neptune supports algorithms across four categories:

Category	Algorithms
Traversal	`bfs_edges`, `descendants_at_distance`, `bfs_layers`
Link Analysis	`pagerank`
Centrality	`degree_centrality`, `in_degree_centrality`, `out_degree_centrality`, `closeness_centrality`
Community	`louvain_communities`, `label_propagation_communities`, `fast_label_propagation_communities`, `asyn_lpa_communities`

→ Full algorithm reference

Neptune Analytics extensions

Beyond standard NetworkX parameters, nx-neptune exposes Neptune Analytics-specific parameters on each algorithm:

vertex_label — filter by vertex label
edge_labels — filter by edge labels
concurrency — control thread count
traversal_direction — "outbound" or "inbound"
write_property — persist results as node properties
edge_weight_property / edge_weight_type — weighted computation

These are passed as additional keyword arguments:

r = nx.pagerank(
    G,
    backend="neptune",
    vertex_label="Person",
    edge_labels=["KNOWS"],
    write_property="pr_score",
)