A source feeds a context graph with real records from a connected system. Connecting one walks you through a short wizard: pick the connector and connection, choose what to read, map it onto your graph's nodes, and set how it syncs. This page covers that general setup — the optional step for splitting and enriching records before mapping has its own page.
Overview
From a graph's Sources tab, Connect source opens the Add data sources wizard (reopening an existing source shows Edit source). Every connector walks the same rail: Source, Connection, Objects, Discover & Enrich, Map, and Settings — what differs by app is what's inside each step, not the sequence. Reopening a saved source resumes at its first unfinished step.
There's no unsaved-changes warning: closing the wizard discards anything not saved with Continue or Save draft, and closing before picking a connection loses the connector choice entirely. If a Continue save is rejected, you'll see which step failed and why — a "version is stale" message means someone else saved it meanwhile, so reopen and redo your change.
Choosing the connector and connection
The Source step lists every connector available to context graphs, with search and a category filter; picking one collapses the step to a summary card and moves you on. The Connection step lists that app's active connections — Add new connection creates and selects one without leaving the wizard, and a footer note flags any inactive connections needing reconnection. Picking a connection saves the source as a draft and advances you; changing the source or connection later resets everything downstream, so settle these two first.
Selecting objects and scope
The Objects step lists everything the connection can read, with a checkbox per object and a Select all shortcut; at least one must be selected. Some objects carry an inline Scope — filter fields from the connector narrowing which records it uses, with a custom-value option where a lookup list falls short. Required scope fields must be filled before Continue unlocks, and changing a scope, or reselecting a deselected object, resets its configuration.
Discover & Enrich
Between Objects and Map sits an optional step where an agent or automation sorts each object's records into entities and enriches them before they reach the graph. It has its own configuration, covered in the Discover & Enrich documentation.
Mapping fields into your graph
The Map step wires source data into the graph. With Discover & Enrich off, you map each selected object; with it on, you map your discovered entities instead. For each item, map it to nodes and edges, then map source properties to destination properties — add transformed fields, use Auto Map for matching names, and optionally filter which rows get ingested. Enrichment outputs, if any, appear as extra fields under the enricher's name. At least one node must be mapped across the source before you can continue or publish.
Pipeline and sync settings
The final Settings step configures how the source behaves once live. A base set applies to every source, and your connector may add more:
Knowledge Retention — removes ingested knowledge after a retention period in days, months, or years.
Skip User-Level Source Permission Checks — lets any user access this source's knowledge regardless of their own permissions.
Schedule Knowledge Sync — automatic re-syncs on a recurring interval or cron expression.
Settings save whenever you continue or publish.
Publishing and source status
Once every step is complete, the wizard's final action switches from Continue to Publish. Publishing validates that at least one node is mapped, snapshots your configuration, and sets the source's status to Syncing, starting ingestion for every configured object; republishing after edits takes a fresh snapshot, and any object deselected beforehand stops being ingested.
The Sources tab lists each source's app, connection, status, last sync time, owner, and last update date. You'll mostly see Draft (saved, never published) or Syncing (published) — there's no sync log or "sync now" button, so the clearest signal that data is flowing is new records on the Records tab for the node types you mapped.
Notes
The wizard never warns before discarding unsaved changes — save often if stepping away.
Changing the source or connection wipes every step that follows it, so lock those in early.