GRAPH · APP · ZERO-INSTALL

Write the query. See the graph. One HTML file.

A browser-based workspace for any Kinetica instance. Browse graphs, read the auto-loaded ontology, run GQL, and explore results as interactive path visualizations — all from a single HTML file, nothing to install.

Open Explorer User Guide

One workspace

Ontology, query, and render — no export step.

Select a graph and its ontology auto-loads as a Graphviz DOT diagram. The Query Helper builds GQL from those labels; results draw themselves as an interactive force graph. The thing you read, the query you build, and the path you see are the same graph on the same server.

Ontology (DOT) → Query Helper (GQL) → live path render

ontologygenerated GQLpath render

NO EXPORT STEP — Click any node to copy its entity ID or pull its full source-table row. Edit the traversal and the render updates with it.

Four actions

Create, Query, Solve, Match — from the header.

Four grammar-driven helpers scaffold valid SQL so you don't hand-write it. Each opens a structured form with column auto-complete, combo grouping, and a Table-or-Constants toggle — then emits a statement you can run or edit.

+ Create

Create

A structured form scaffolds a valid CREATE OR REPLACE GRAPH from NODES and EDGES input tables.

directed / undirected · recreate or modify · parses existing CREATE SQL

+ Query

Query

SQL/GQL editor with the Query Helper above it — generate a traversal from ontology labels, then refine.

multiple panels · path viz · force-graph · node detail lookup

+ Solve

Solve

Grammar-driven helper for /solve/graph — pick a solver type and the panel filters options to fit.

SHORTEST_PATH · PAGE_RANK · STATS_ALL · WEIGHTS · RESTRICTIONS

+ Match

Match

Same shape as Solve, for /match/graph — pick a solve method, configure SAMPLE_POINTS, run.

markov_chain · match_supply_demand · match_isochrone · match_loops

Query helper

Pick two labels. It writes the Cypher.

Choose a source and target label from the ontology and hit Generate Query. The helper finds the shortest path between them through the ontology graph — a BFS through waypoints — and emits a direction-aware GQL MATCH pattern with the correct arrows. Labels with spaces are auto-quoted; graph names are double-quoted per part.

BFS through the ontology → direction-aware MATCH

generated GQL

query_helper.gql

GRAPH "demo"."kg"
MATCH (a:Organization)
      -[:MAKES]->(p:Product)
      <-[:AFFECTS]-(b:Vulnerability)
RETURN a, p, b;

-- found by BFS through the ontology
-- arrows resolved per edge direction
-- labels with spaces auto-quoted
-- helper auto-collapses, path shown green

At scale

Geospatial graphs, rendered on the GPU.

For WKT graphs the renderer toggle switches between Canvas, a Deck.gl WebGL layer, and Kinetica's server-side WMS tiles. Deck.gl uploads Float32Array attributes straight to the GPU and handles tens of millions of edges interactively; WMS renders server-side with no entity fetch at all.

27M

Edges · Deck.gl @ 60fps

WebGL LineLayer with a CartoDB basemap toggle (Off / Light / Color).

6.76M

Nodes · 13.12M edges

ki_home.us_roads, colored by state across 49 distinct edge labels.

622K

Banking demo graph

expero.banking_graph — 16 node labels, 15 edge labels, 845K edges.

EDGE PICKING — Click an edge in any renderer and the full source-table row lands in the detail strip. Deck.gl and Canvas do an id lookup; WMS runs a server-side ST_DWITHIN around the click. A viz-limit guard (10K → ∞) gates very large pulls.

Connect

Any instance. Your session, saved.

Point Explorer at any Kinetica instance and start browsing — graphs and tables in the sidebar, edge-count badges sourced without extra round trips. Query and helper panels persist, so you can pick up where you left off.

Zero install

One HTML file

No build, no server. Open it, enter a connection, and explore. Switch instances or graphs with a guarded confirmation banner.

Session save / restore

Panels persist

Query, Solve, Match, and Create panels save with their form state. Create panels restore without auto-running, so an in-progress build is preserved.

Source-table truth

Every click resolves

Nodes and edges resolve back to the full row in the original table — weights, labels, geometry — not just the graph projection.

Build a graph from text → GraphRAGDon't have a graph yet? kgr induces one from documents — then open it here.

Explore any graph in your browser.

Open Explorer User Guide GitHub

Frequently asked questions

Do I need to install anything to use Graph Explorer?

No. Graph Explorer is a single HTML file — no build and no server. Open it, enter a connection to any Kinetica instance, and start browsing. You can switch instances or graphs from a guarded confirmation banner, and the app is also open source on GitHub.

How does the Query Helper write GQL for me?

Pick a source and a target label from the auto-loaded ontology and hit Generate Query. The helper runs a BFS for the shortest path between them through the ontology graph and emits a direction-aware GQL MATCH pattern with the correct arrows. Labels containing spaces are auto-quoted and graph names are double-quoted per part; undirected graphs generate -[]- with no arrows.

How large a graph can it render?

For WKT/geospatial graphs the renderer toggles between Canvas, a Deck.gl WebGL layer, and Kinetica's server-side WMS tiles. Deck.gl uploads Float32Array attributes straight to the GPU and handles tens of millions of edges interactively — for example 27M edges at 60fps — while WMS renders entirely server-side with no entity fetch at all.

What happens when I click a node or an edge?

Every click resolves back to the full row in the original source table — weights, labels, geometry — not just the graph projection. Deck.gl and Canvas do an id lookup; WMS runs a server-side ST_DWITHIN around the click. A viz-limit guard (10K → ∞) gates very large pulls.

What do the Create, Query, Solve, and Match actions do?

They are four grammar-driven helpers that scaffold valid SQL so you don't hand-write it. Create builds a CREATE OR REPLACE GRAPH from node and edge tables; Query is a SQL/GQL editor with the Query Helper above it; Solve targets /solve/graph (SHORTEST_PATH, PAGE_RANK, STATS_ALL, …); and Match targets /match/graph (markov_chain, match_supply_demand, match_isochrone, …).

Does it remember my work between sessions?

Yes. Point Explorer at any instance and your Query, Solve, Match, and Create panels persist with their form state. Create panels restore without auto-running, so an in-progress build is preserved rather than re-executed.