PuppyGraph
PuppyGraph is a real-time, zero-ETL graph query engine. Rather than requiring developers to copy relational tables into a specialized graph database (like Neo4j) via complex ETL pipelines, PuppyGraph queries relational data stores and data lakes directly as a graph model. It supports standard graph query languages like openCypher and Gremlin.
Iceberg Integration Scope
PuppyGraph provides native compatibility with Apache Iceberg catalogs and tables:
- Zero-ETL Schema Mapping: PuppyGraph maps existing Iceberg tables directly into nodes and edges at query time. For example, a
userstable acts as node vertices, and atransactionstable acts as connecting edges, without moving the data. - Catalog Connectivity: PuppyGraph integrates with Iceberg catalogs (including AWS Glue, REST Catalogs, Apache Polaris, and Databricks Unity Catalog). It queries the catalog to discover tables and fetch schemas automatically.
- Vectorized Graph Execution: To support multi-hop graph traversals (such as 10-hop relationship queries), PuppyGraph utilizes a vectorized execution engine. It reads data files directly from cloud object storage (like S3) using path-based credential vending, executing joins and traversals in memory.
- Hybrid Analytics: Because PuppyGraph is read-only on the lakehouse storage, the same Iceberg tables can be queried simultaneously using SQL for traditional BI (via engines like Dremio) and openCypher for graph-based fraud detection.