Creating Your First Knowledge Graph

Before anything gets clicked there is one decision that shapes everything after it. Does this data deserve a new Knowledge Graph, or does it belong in one you already have? This lecture walks through that decision and then shows the full creation flow: connecting a warehouse, selecting tables and columns deliberately, and letting Veezoo generate a first Knowledge Graph.

Deliberate selection is the part worth watching. Technical columns mean columns with no business meaning, such as the timestamp a load tool like Fivetran writes when it imports a row. That value is of no use to your users and of no use to the Knowledge Graph, while ID and name columns are needed. The generated result starts as one file per table and answers its first questions within minutes.

Key points

• Extend by default: if users will connect the new data to data you already model, add it to the existing graph.
• New graph only for a separate domain with its own access needs, such as HR.
• One Knowledge Graph connects to exactly one data source.
• Creation flow: create the graph, pick the database type, enter credentials, give it a proper name.
• Select deliberately: import only the schemas, tables, and columns your users' questions need.
• Technical columns stay out: a Fivetran load timestamp helps nobody, while IDs and names are needed.
• Add to Knowledge Graph imports the warehouse semantics, creates classes, and infers relationships from primary and foreign keys.

Check your understanding