Step 4 — Graph Editor
Visualize and edit your knowledge graph (Graph RAG only).
The Graph Editor is a conditional step that only appears when you select Graph RAG as your pipeline type. It provides tools to visualize, explore, and refine the knowledge graph extracted from your documents.
When Does This Step Appear?
The Graph Editor step is visible only when:
- Search Method is set to
graphin Pipeline Configuration - Graph Retrieval is enabled during document processing
For Simple RAG and Hybrid RAG, this step is skipped entirely.
Graph Components
The Graph Editor provides multiple views and interaction modes for working with your knowledge graph. Each component serves a distinct purpose in the graph exploration and editing workflow.
Graph Visualization
Interactive force-directed graph showing entities as nodes and relationships as edges. Nodes are color-coded by entity type and sized by connection count. Click nodes to highlight connected entities and explore relationship paths.
Graph Table View
Tabular representation of all graph data with filtering and sorting. View all entities and relationships in a structured format, search by name, and filter by type. Useful for auditing graph quality and identifying extraction errors.
View Toggle
Switch between Graph Visualization and Table View modes. The toggle preserves your current filter state when switching between views.
Node and Edge Filtering
Filter graph elements by entity type, relationship type, or confidence score. Filters apply to both visualization and table views simultaneously. Clear filters to restore the full graph.
Graph Highlight System
Visual highlighting system that emphasizes specific nodes and their connections. Highlights appear when you click nodes, run queries, or apply filters. Highlighted nodes are styled differently to distinguish them from the rest of the graph.
Graph Chat AI Assistant
The Graph Chat Drawer provides an AI-powered interface for editing your knowledge graph through natural language. Describe what you want to change — such as "Add property 'status' to all Person nodes" or "Show me all entities connected to Project Alpha" — and the AI will preview the changes for you to confirm or cancel before applying.
Entity types and relationship types are determined automatically by the LLM during graph extraction based on your document content. The system identifies what entities and relationships exist in your data — you don't need to define them upfront.
Graph extraction consumes graph credits. The quality of the extracted graph depends on the LLM model used and the structure of your source documents.