Skip to main content
Turn a saved Explorer query into charts, assemble them into a dashboard, and publish a public link — all through tool calls.
Visuals and dashboards both read from saved Explorer queries. Create the query first with create_explorer_query — an ephemeral run_sql_query result can’t back a chart. See SQL & queries.

Available tools

Fetch the relevant skill before you construct a spec. Both set_dashboard and create_explorer_visual take a typed JSON config, and the exact schema per type lives in the skill — see Skills.

Visuals

A visual belongs to a query. spec.type selects the variant, and the remaining fields depend on that type. Types: chart, value, table, pie, sankey, treemap, scatter, map, chord.
1

Load the schema

get_skill(name="explorer-visuals") for general guidance, then get_skill(name="explorer-visuals", reference_title="<type>") for that type’s exact JSON schema.
2

Create the visual

Returns a visual_id and the visual’s URL. Pass an existing visual_id to replace that visual in place instead of adding a new one.
get_explorer_query returns a visuals list for a query — use those IDs with get_explorer_visual, delete_explorer_visual, or the visual_id argument of share_explorer_query.
Editing a visual requires edit access to its parent query. Visuals inherit permissions from the query they belong to.

Dashboards

A dashboard is a tree: dashboard → page → section → element. set_dashboard upserts or deletes any node in that tree, and the ID path you pass locates the target. The rule: on create, the IDs are the parent path and the target’s own ID is absent. On update, the IDs include the target’s own ID.
Element config.type covers the nine visualization types above plus layout and control elements: markdown, spacer, variable-select, switch, date-range, and radio-group. Call get_skill(name="dashboard-design") for the build order and element catalog, then get_skill(name="dashboard-design", reference_title="<config.type>") for a type’s full schema.
Pass the literal shown above as config.type — a dropdown is variable-select, not select. The skill’s element catalog is the authoritative list.
If the request touches a topic Terminal already covers, run search_terminal and get_terminal_results first and use that chart manifest and SQL as your baseline. See Knowledge & discovery.

Reading a dashboard

read_dashboard returns the targeted node plus summaries of its immediate children — never a full subtree.

Deleting

delete_dashboard takes a dashboard_id only and removes the whole dashboard. To remove a page, section, or element, call set_dashboard with spec: null and that node’s ID path.

Sharing

Returns a share URL (https://app.allium.so/s/{share_id}) and an embed URL (https://app.allium.so/embed/{share_id}).What gets shared depends on visual_id:
  • Omitted — the query. The shared page shows the results table plus one tab per saved visual. Usually the best choice when there’s more than one visual.
  • Provided — that one visual, pinned. This is the only form whose embed URL renders the visual; without it the embed shows just the table.
exclude_sql: true hides the SQL on the shared page. It defaults to false, so the SQL is visible.
Anyone with a share link can view the content without logging in. There is no unshare tool: to stop sharing a dashboard, open it at https://app.allium.so/analyze/dashboards/{dashboard_id} and toggle off public access.

SQL & queries

Create the saved queries that back your charts

Skills

Fetch the exact spec schema for every visual and element type

Dashboards in the App

Build visuals and dashboards by hand in the App

Terminal dashboards

Start from an analyst-built baseline