Before You Start

• Define entity types, such as companies, people, institutions, regions, policies, or papers.
• Define relationship types, such as collaboration, citation, investment, affiliation, publication, or influence.
• Prepare data files, DuckDB tables, or PDF, Word, Markdown, and TXT documents that contain extractable entities and relationships.
• For document-based graph building, state the extraction scope, entity types, relationship types, and evidence-retention requirements.

Steps

1. In the Knowledge Graph workspace, review the current schema, node table, and relationship table; use Apply Schema to adjust the model.
2. Click Import Data to import node files, relationship files, or DuckDB tables into the graph. To build a graph from documents, ask the Agent to inventory the materials and preserve evidence locations first.
3. To review capabilities, open Settings → Skills and switch to the Knowledge Graph Agent. Built-in Skills sync automatically; use Import only to add graph-building capabilities.
4. Enter a graph prompt with entity types, relationship types, field mapping, unique identifiers, document evidence requirements, and query goals.
5. Click the send button and let the Knowledge Graph Agent extract relationships, create the model, import data, or run Cypher queries.
6. Click “Open Visualization” to inspect whether the node-edge network matches the research question.

Acceptance Checks

The graph model should describe nodes, relationships, properties, unique identifiers, and data sources. For document-built graphs, it should also retain filenames, page or paragraph positions, text excerpts, and extraction conclusions. Query results should include the Cypher definition and trace back to source materials, node tables, or relationship tables.