Graphistry Integration¶

GPU-accelerated graph visualization with llcuda.

What is Graphistry?¶

PyGraphistry provides: - GPU-accelerated graph rendering - Millions of nodes/edges - Interactive exploration - RAPIDS integration (cuDF, cuGraph)

Split-GPU Architecture¶

flowchart LR
  A[GPU 0: llcuda LLM] --> B[GPU 1: RAPIDS + Graphistry]

Quick Start¶

from llcuda.graphistry import GraphWorkload, register_graphistry

# Use GPU 1 for graph workloads
workload = GraphWorkload(gpu_id=1)

# Register Graphistry (personal keys recommended)
register_graphistry(
    api=3,
    protocol="https",
    server="hub.graphistry.com"
)

# Visualize
g = workload.create_knowledge_graph(
    entities=[{"id": 1}, {"id": 2}, {"id": 3}],
    relationships=[{"source": 1, "target": 2}, {"source": 2, "target": 3}]
)
g.plot()

See: - Knowledge Graphs - RAPIDS Integration - Examples