Skip to content

User Guide Overview

This guide covers how to use GraphForge for graph analysis and construction.

Topics

Cypher Query Language

Learn the openCypher query language — GraphForge's primary interface for working with graphs.

Graph Construction

Build graphs programmatically using the Python API.

Analytics Integration

Export graphs to NetworkX, igraph, and pandas for further analysis.

Ranking Nodes — forge.rank()

Score every node with a graph centrality algorithm (PageRank, betweenness, closeness, degree, clustering coefficient, or triangle count). Returns an Arrow Table with node properties plus a score column. Pass write_property to persist scores back to the graph. See the tutorial for examples.

Clustering Nodes — forge.cluster()

Assign community membership using Louvain or connected-components algorithms. Returns an Arrow Table with node properties plus a community_id column. Pass write_property to persist assignments back to the graph. See the tutorial for examples.

Finding Nodes — forge.find()

Full-text, vector similarity, or hybrid search over node properties. Bring your own vectors — GraphForge stores and queries them but does not generate embeddings. Returns an Arrow Table with node properties plus score and matched_on columns. See the tutorial for examples.

Core Concepts

Graphs

A graph consists of nodes (vertices) and relationships (edges) connecting them.

Nodes

Nodes represent entities in your graph. They can have: - Labels - Types or categories (e.g., Person, Product) - Properties - Key-value pairs with data

Relationships

Relationships connect nodes and can have: - Type - The nature of the connection (e.g., KNOWS, PURCHASED) - Direction - From one node to another - Properties - Additional data about the relationship

Patterns

Cypher uses ASCII-art patterns to describe graph structures:

(a:Person)-[:KNOWS]->(b:Person)

This pattern matches two Person nodes connected by a KNOWS relationship.

Query Flow

  1. MATCH - Find patterns in the graph
  2. WHERE - Filter results
  3. RETURN - Specify what to return
  4. ORDER BY - Sort results
  5. LIMIT - Limit number of results

Result Types

v0.5.0: ALL methods return a PyArrow Table — execute, rank, cluster, find, and schema. There are no CypherValue wrappers and no SearchHit objects. Access values via .as_py() or pass the table directly to pandas, Polars, or NetworkX, which all accept Arrow as input.

table = forge.execute("MATCH (p:Person) RETURN p.name, p.age")
# table is a pyarrow.Table — use Arrow, pandas, or Polars to consume it
import pandas as pd
df = table.to_pandas()

Next Steps