Visual Graph Explorer

Overview

The Visual Graph Explorer is a read-only Dataiku webapp designed for business users and analysts to perform in-depth visual exploration of graphs managed in the Dataiku Flow.

../../_images/visual-graph-explorer.png

Note

Get started by following this Explorer webapp tutorial.

Key features

  • Schema overview
    View the graph’s structure, including all node and edge groups.
  • Visual exploration
    Visualize graphs with different layouts, expand neighbors to explore relationships, color nodes based on their property values…
  • Saved queries
    Execute a library of saved Cypher queries prepared by data scientists to answer common business questions.
  • Ad hoc querying
    Write and run custom Cypher queries directly within the interface with schema-aware autocompletion.
  • Query generator
    Use a built-in LLM assistant to generate Cypher queries from natural language questions.

Note

All schema design, data mapping and visual customization are performed beforehand in the Visual Graph Editor webapp.

Settings

Graph databases

  • Folder(s) containing the graph databases: select one or multiple Dataiku Folders containing graph databases.

  • Neo4j connection: if some selected folders represent a Neo4j graph database, configure the connection to be used by the webapp. Only one Neo4j connection can be used for all selected folders.

Note

To access Neo4j databases available through different connections, you will need to create separate Explorer webapps for each connection.

AI assistant

  • LLM connection to use (Optional): select a LLM connection to enable the AI-powered Query generator. It helps users construct complex Cypher queries using natural language.

  • LLM history dataset (Optional): select the dataset storing the history of all questions asked by users in the Query generator. It can be used for auditing or analysis.

Advanced settings

  • Cypher query timeout (seconds): set a maximum execution time for queries run within the Explorer. This acts as a guardrail to prevent resource-intensive queries from impacting performance.

  • Log level: configure the verbosity of the logs. Select INFO for standard operational logging (recommended for production) or DEBUG for detailed diagnostic information while troubleshooting issues.

Note

Not configuring optional settings will simply disable the associated feature.

Visual Graph Explorer Interface

Landing Page

The landing page displays a list of all available graphs present in the selected graph database folders. Key information, such as the last build time, is shown for each graph to indicate data freshness.

Graph Exploration Page

Left-side panel

Graph information

The top of the panel displays the graph’s name, its last build time, and the comment associated with its published configuration.

Groups

All node and edge groups are listed with their respective colors.

Graph statistics

For each group, two types of counts are shown:

  • visible: The number of nodes/edges currently displayed.

  • total: The total number of nodes/edges for that group in the entire graph database.

View controls

For each node and edge group, you can toggle its visibility on or off to show or hide all elements of that group in the current view.

Bottom panel

Saved queries tab

The Saved queries tab contains a list of pre-built queries. They can be executed to see the results rendered as a graph or a table.

Note

These queries are prepared by Data Scientists in the Editor webapp.

New query tab

The New query tab provides a powerful editor with schema-aware autocompletion to write and execute your own Cypher queries.

Query generator tab

For complex queries, the Query generator tab allows you to ask a question in natural language. The AI assistant will translate your question into a Cypher query that you can then execute.

Central panel

In the Graph view, you can click on any node or edge to inspect its details, including its identifier, name, and properties.

When a node is selected, an additional Neighbors tab is available. It provides an overview of all directly connected nodes, displaying the total neighbor count and a breakdown by node group.

From here, you can selectively expand all neighbors or only those belonging to a specific group to explore the graph’s structure interactively.

Graph layouts

Switch between different layouts to optimize the visualization of your graph based on its structure.

Options include gravity (the default layout), tree for hierarchical data and circle or square grid grouping to cluster nodes per group.

Warning

Tree layout is computationally intensive. It is not recommended for large graphs.

Conditional coloring

Enhance the visual analysis of your graph by coloring nodes based on the values of a selected property.

For numerical values, a gradient color scale is applied. For categorical values, you can assign colors to a specific set of values.

It helps identify patterns, clusters, or outliers within the graph based on specific attributes.