Graph

The GraphResource provides graph-level operations: community detection, duplicate entity finding, and entity merging.

Access

graph = client.graph

Methods

communities(limit)

List detected communities in the knowledge graph. Communities are clusters of related entities found via graph algorithms.

communities = graph.communities(limit=10)
for c in communities:
    print(f"{c.label}: {c.entity_count} entities")
    if c.summary:
        print(f"  {c.summary}")

Returns: List of GraphCommunity objects with fields id, label, summary, entity_count, entity_ids

detect_communities(resolution)

Run community detection on the knowledge graph using the Louvain algorithm. This is a background operation that groups related entities.

result = graph.detect_communities(resolution=1.0)
print(result)  # {"message": "Community detection started"}

Returns: Status dict with message

find_duplicates(threshold, limit)

Find potential duplicate entities based on semantic similarity.

dupes = graph.find_duplicates(threshold=0.95)
for group in dupes.groups:
    print(f"Duplicates (similarity: {group.similarity_score}):")
    print(f"  Entity IDs: {group.entity_ids}")
    print(f"  Suggested keep: {group.suggested_merge_target}")

Returns: DuplicatesResponse with fields:

• groups — List of DuplicateGroup (entity_ids, similarity_score, suggested_merge_target)
• total_groups — Total number of duplicate groups found

merge(keep_entity_id, merge_entity_id)

Merge two duplicate entities. Keeps one entity and soft-deletes the other, redirecting all relationships.

result = graph.merge(
    keep_entity_id="uuid-to-keep",
    merge_entity_id="uuid-to-merge",
)
print(f"Merged into: {result.merged_into}")
print(f"Relationships redirected: {result.relationships_redirected}")

Returns: MergeResponse with fields merged_into, merged_count, relationships_redirected, status