Knowledge Graph

Content doesn’t live in isolation. Rosmarium provides a lightweight knowledge graph layer powered directly inside PostgreSQL.

Building the Graph

✦

Typed Edges

You can define directional, typed edges between content items to map relationships manually.

Article -> AUTHORED_BY -> Person
Company -> COMPETES_WITH -> Company

✦

Auto-Inference

Rosmarium’s AI worker automatically infers edges! When NER extracts an entity (e.g. “Google”), the Engine attempts to resolve it against existing “Company” entries and creates an INFERRED_REFERENCE edge.

Graph Traversal & Analytics

Cypher-lite Traversal

You can traverse the graph via API using a subset of the Cypher query language:

MATCH (a:article)-[:AUTHORED_BY]->(p:person)
WHERE p.name = “John Doe”
RETURN a

The Node.js API transpiles this into recursive CTEs executed natively in PostgreSQL, so no specialized graph database is required.

Deep Graph Analytics

The Python worker leverages NetworkX to run heavy analytics across the entire dataset:

📈

PageRank

Identify the most authoritative content entries.

🕸

Louvain Communities

Group highly interconnected entries into contextual clusters.

[!TIP] Data Portability: You can export the entire graph in standard JSON-LD, RDF, or GraphML formats at any time.