From Query Results to Visual Graphs with IRQL

Where this fits

You have a KQL query that returns tabular rows—say, a list of Azure resources and their dependencies—but you need to see that data as an interactive graph in Kusto Explorer. Manually mapping columns to nodes and edges, choosing layout, and adding icons is tedious and error-prone. The azure-kusto-irql-graph skill automates this: it takes your existing query result and your graph description, then generates the Lift_To_Graph mapping and the exact IRQL graph functions (Graph_Render_View, Graph_Fold_By_Property, Extract_Node_*, Enrich_Node_*, Enrich_Graph_*) needed to produce a visual, foldable, icon-decorated graph—without rewriting your query in IRQL.

Why agents benefit

• No KQL rewrite: The source query stays as-is; the agent only adds the graph layer on top of the tabular result.
• Composable function generation: Instead of a monolithic script, the agent emits only the stored functions required for the requested visualization (e.g., fold by property, enrich nodes with icons).
• Handles ambiguous graph descriptions: If you say “show me a graph of resources grouped by region,” the agent infers the appropriate Fold_By_Property and node extraction calls.
• Works in Kusto Explorer directly: The output is a ready-to-run query that renders in the tool, not a separate visualization library.
• Clear routing: It knows when not to use itself—for native graph analysis (shortest paths, graph-match) it directs to azure-kusto-graph, preventing misuse.

Practical scenarios

Visualize dependency chains from a KQL query

You have a query that returns source, destination, and type for all service-to-service dependencies. The agent maps these to graph nodes and edges, then adds Graph_Render_View with a hierarchical layout to show the chain flow.

Fold a large device inventory by location

A query returns thousands of IoT devices with a region column. The agent uses Graph_Fold_By_Property to collapse nodes by region, letting you expand any region to see individual devices—all within Kusto Explorer.

Enrich a network graph with icons and severity

After generating the base graph, the agent applies Enrich_Node_* to add icons based on resource type and Enrich_Graph_* to color edges by latency, giving an at-a-glance operational view without manual formatting.

Add it to your agent workflow

Install the skill from the Microsoft GitHub-Copilot-for-Azure repository:

npx skills add microsoft/GitHub-Copilot-for-Azure --skill azure-kusto-irql-graph

Then, in your agent prompt, provide a query and a graph description. Example:

Run this KQL query: AzureResource | project source=ResourceName, target=DependsOn and show me a graph with icons for each resource type, folded by subscription.

The agent will output the IRQL graph function calls that you can paste into Kusto Explorer.

Compare before adopting

Before integrating this skill, compare it against:

• Quality signals: Check the GitHub repo’s stars (currently 248) and recent commits. The README is clear about scope, which is a good sign.
• Maintenance freshness: The version placeholder (0.0.0-placeholder) suggests it’s early—verify if updates are frequent and issues are addressed.
• Alternatives: For native graph analysis (graph-match, shortest paths, persistent graphs), use azure-kusto-graph instead. This skill is only for visualization of existing tabular results.
• Workflow fit: Does your agent already handle KQL generation? If so, this skill adds a clean post-processing step without overlapping responsibilities.

Why it is worth tracking

This skill fills a real gap: most Kusto users know KQL but not IRQL graph functions. By automating the translation, it lowers the barrier to rich visualizations. The Microsoft authorship and 248 stars provide initial confidence, but the placeholder version means it’s still evolving. Track it if you frequently build Kusto Explorer dashboards or want your agent to produce visual insights from query results—and evaluate it again once it reaches a stable release. For a curated list of similar agent skills, see OpenAgentSkill.