Ontology Map
A Map That Builds Itself
Every source, table, pipeline, metric, dashboard, and business object in your workspace, connected and computed live from what you've already built. No diagram to maintain, nothing that can drift.
What It Is
Nothing to keep up to date, because nothing is hand-written.
The ontology map is a live view of your workspace: every source you've connected, every table you have, every pipeline that built one, every metric and dashboard that reads from one, and every business object (like Customer or Country) your tables represent. It's assembled fresh, every time you open it, straight from the same records the platform already keeps. There's no separate lineage document sitting off to the side going stale. Change your workspace, and the map changes with it.
The map also fills itself in. It drafts links between your tables and the business objects in your data, using the keys in your Raw tables. You approve or reject each draft. Nothing here is used by the AI team until you approve it.
Reading the Map: Node Kinds
Seven kinds of nodes, each with its own color.
Reading the Map: Connections
Every line on the map is labeled with what it means.
Produces
A pipeline built this table.
Ingests to
A source loads its data into this table.
Reads
A dashboard queries this table.
Uses
A metric is calculated from this table.
Transforms into
SQL turned one table into another.
Conforms to
Entity matching turned this table into a standardized, canonical one.
Represents
This table's rows describe an object or a business object you approved.
Joins to
You approved a link between these two tables, so the AI team joins them on that key.
Using the Page
Five ways to work the map.
Click a node
Opens a side panel with the node's details, when it was created and last updated (when known), and "Depends on this": everything downstream that would be affected by a change.
Search
Type a name and press Enter, or click a result, to pan and zoom straight to that node.
Unconnected nodes hidden
Nodes with no edges are hidden by default. A chip shows how many, and toggling it lays them out in a grid below the connected map.
Legend
A collapsible key in the corner maps each icon back to its node kind.
Pan, zoom, fit
Drag to pan, scroll to zoom, and the map fits itself to the screen the moment it loads.
Suggestions You Approve
The map drafts. You decide. Nothing is used until you approve it.
Links, Objects, Dimensions
With no node selected, the side panel has three tabs. Each one lists the drafts waiting for you.
Declared or Maybe
A Declared link is a key your source declared itself. A Maybe link was measured in your data and scored for how sure we are.
Approve or Reject
Approve a draft and the AI team starts using it. Reject it and it stays out. Approve all declared takes every Declared link in one click. Viewers can look but not change anything.
Update suggestions
Drafts refresh on their own when you connect a source, sync its schema, or new column profiles land. Click Update suggestions to run it now.
Metrics and dimensions
After a pipeline run, metrics are drafted from your Ready SQL and dimensions from its GROUP BY columns. Approve dimensions in the Dimensions tab. Approve or Dismiss metric drafts on the semantic layer page, or click Update drafts there.
Every business object gets a page
Open it from the Objects tab or from "Open object page" on its node. It shows Facts, Sample rows, and Linked objects. Ask the Analyst opens a chat with a question about it ready to send.
Why Does a Node Have No Connections?
An unconnected node just means nothing registered a link to it yet. The most common case is a pipeline that never ran, or ran but never registered an output table. The map isn't broken; it's telling you the truth about your workspace. Once that pipeline produces a real table, the connection shows up on its own.
The Ontology Beyond the Map
The same graph powers the rest of the platform.
Guardrails before you break something
Deleting a data source or saving Clean SQL checks the map first. If the change would touch metrics or dashboards downstream, you get a warning naming exactly how many, before you confirm. A transform on the canvas shows the same warning in its node panel. It warns, it never blocks.
The AI team uses what you approve
The Data Engineer joins tables on approved links. The Quality Engineer proposes checks that every key has a match across an approved link. The Analyst, Analytics Engineer, and BI Developer read approved metrics, dimensions, links, and objects as context.
Ask the Manager
Ask "what feeds what?" or "what breaks if I change this table?" in plain English. The Manager reads the same live map to answer.
Metrics stay human-approved
When the Manager proposes a metric on your behalf, or one is drafted from your Ready SQL, it lands as a draft. Nothing becomes certified or governed until a human approves it in the semantic layer.
Availability
The ontology map is a Scale plan feature. On other plans, the page explains what's needed to unlock it instead of showing an error.
What's Next
See who reads the map, and where it leads.
One map, built from what you already have. No diagram to maintain.