Query Panel

The Query panel is the everyday workspace: write a command in any supported language, run it against the selected database, and inspect the result as a table, a graph, a chart, a JSON document, or an execution plan.

Query Panel

The header shows the current user and the Database Selector; pick the database to run against here. The status bar at the top right reports Returned X records in Y ms after every execution.

Left icon sidebar

The narrow icon column on the left of the editor switches between auxiliary panes; the main editor and results stay in place.

  • Overview — summary information about the selected database (default view).

  • History — list of recently executed commands. Click an entry to load it back into the editor.

  • Saved — saved queries. Use the Save Query button in the editor toolbar to add the current command to this list.

  • Reference — searchable reference of SQL functions and methods plus quick links to the other query languages.

  • Settings — editor preferences (theme, font size, key bindings).

Query left sidebar

Editor and toolbar

The editor is a CodeMirror instance with syntax highlighting that follows the selected language.

The toolbar above the editor exposes:

  • Language selector with: SQL, SQL Script (multi-statement), Open Cypher (native engine), Apache TinkerPop Gremlin, GraphQL, MongoDB, Redis, PromQL (Time Series), and under Index search: Vector search, Hybrid search, Full-text search (see Index search).

  • Auto Limit dropdown — caps the number of rows returned by interactive runs (20, 50, 100, 500, No Limits). Useful to stay safe on large datasets; the cap is applied as a LIMIT when the language supports it.

  • Clear — empties the editor.

  • Save Query — adds the current command to the Saved pane in the sidebar.

  • Run (or Ctrl+Enter / Cmd+Enter) — executes the command.

Query editor toolbar

The last three entries of the Language selector query an index instead of running a command. Selecting one replaces the editor with a small form, while Auto Limit is hidden because the form carries its own k or limit. Run, Clear, Save Query, the history and the result views (Table, Json, Explain) work as for any other language.

  • Vector search - pick an LSM_VECTOR or LSM_SPARSE_VECTOR index, enter the query vector ([0.12, -0.5, 0.33] or 0.12, -0.5, 0.33), k, an optional SQL WHERE filter, and for a dense index efSearch. For a sparse index you can also enter the sparse dimension ids.

  • Hybrid search - the vector inputs above plus a full-text leg (index and Lucene query), the fusion strategy, per-leg weights and an optional graph expansion (edge types, direction, max depth).

  • Full-text search - pick a FULL_TEXT index, enter a Lucene query (for example +java -python) and a limit.

The form checks what it can before sending: a vector whose length differs from the index dimensions is refused without a round trip, and the limits for k, efSearch, limit and maxDepth are read from the server, so they always match the running version. The index list is read from the database schema; a database without a vector or full-text index shows a note instead.

Each hit is one table row: @rid, @type, then the ranking value (distance, score or fusedScore, and for hybrid the sources), then the record properties. The Json tab holds the raw response, and the Explain tab summarizes which index answered, how it ranks, and whether the result window was filled. When the count reads (truncated), more matches exist: raise k.

The form runs the same operations as the HTTP endpoints POST /api/v1/vector/{database}/search, /hybrid and /fulltext. These searches used to live in a separate Vector tab, which was removed.

Result views

The result area sits below the editor and exposes four tabs. Studio automatically picks the most appropriate one for the result type, but you can switch at any time.

Table

Tabular view backed by DataTables with sorting, paging and per-column search.

  • Click a cell to open the Record Editor in a modal overlay, where you can edit fields and save the record back to the database.

  • The header tools allow exporting the current result to CSV.

Table view

Graph

Force-directed graph view powered by Cytoscape.js. Vertices and edges returned by the command are added to the canvas; existing nodes are kept so you can incrementally explore the graph.

Toolbar actions:

  • Search — locate a node by @rid or a property value.

  • Redraw — reapply the layout.

  • Select — selection helpers: Direct Neighbors, Orphans, Invert Selection, Shortest Path.

  • Cut / Crop — remove the selected nodes / keep only the selected nodes.

  • Import / Export — JSON, GraphML, PNG, JPEG.

  • Settings — per-type style overrides (label, color, size, icon).

  • Fullscreen — expand the canvas to the full browser window.

Interactions:

  • Click on a node selects it; Cmd/Ctrl-drag rubber-bands a block selection.

  • Hold the mouse on a node opens its context menu (load neighbors, edit, delete).

  • Double-click on a node loads its direct neighbors.

The node context menu offers four ways to expand:

  • the three arrows load all neighbours - both directions, outgoing only, incoming only;

  • the funnel opens the relationship picker, where you choose exactly what to add.

The picker lists one row per relationship type and direction with its edge count, largest first, so you can add only the Bought edges going out and the Knows edges coming in, rather than everything at once. A Max elements per direction box caps how much a single expansion may add. This is what keeps a node with thousands of connections from filling the canvas with a hairball; the counts are computed on the server, so opening the picker does not download those edges just to count them. Available since 26.10.1.

Graph view
Graph node context menu

Chart

The Chart tab plots the rows of the last result, whatever language produced them: SQL, Open Cypher, Gremlin, PromQL and so on. The role of each column is inferred from its values, and every choice can be changed in the tab:

  • X axis - the first time column (epoch milliseconds or an ISO date string), otherwise the first text column, otherwise the row number. A time axis is sorted; a text axis keeps the row order, so a GROUP BY result becomes a bar-friendly category chart.

  • Measures - the numeric columns to plot, one line, area or bar series each.

  • Split by - a text column to turn into one series per value. On a time chart it defaults to the first text column with at most 12 distinct values (a host, a metric label), so a result with ts, host, usage draws one line per host.

  • Line, Area, Bar - the chart type.

Record metadata (@rid, @type) and nested values are not plotted, and a value that is not a number leaves a gap rather than a zero. Chart does not run anything by itself: run a query as usual, then open the tab.

SELECT ts, host, usage FROM Cpu

PromQL

Selecting PromQL (Time Series) in the language selector lets you type a PromQL expression, for example Cpu{host="a"}. Auto Limit is replaced by a Range (last 5 minutes to 7 days) and a Step (1 second to 1 hour). The expression runs as a range query against the database’s Prometheus-compatible endpoint, and the points arrive as rows timestamp, metric, value, so they can be read in the Table and Json tabs; the Chart tab opens automatically, with one series per metric.

JSON

The raw response is rendered as pretty-printed JSON in a read-only viewer. A Copy to Clipboard button copies the full payload.

JSON view

Explain

Shows the execution plan for the current command. Tick Profile Execution before running to collect per-step timings; the result is rendered as a flame graph on top of the textual plan, so you can spot the heaviest steps at a glance. Profiling times the query as it normally runs, without changing how it is executed, so the numbers describe the same work the query does when profiling is off.

Explain view