Data and integrations

What it is

An integration connects Semogram to one of your systems. Integrations come in classes: sources bring data in, destinations receive results, and supporting capabilities handle access and external tools. Workflows reference integrations; they never embed connection details.

What it holds

• Sources: a named read point at a system — which account, which scope, how often it syncs. Reading happens in two modes: a full read of the scope, or an incremental read of what changed since last time.
• Destinations: a named write point — target, shape of writes (add, replace, or update-or-insert), and the approval guarding it.
• Configuration: what the integration needs to work, previewed before anything is saved and validated before anything commits.
• Health and sync state: connection status, last sync time, and samples — the facts you check when outputs look stale.

How it relates

• Secrets stay with the integration. Workflows, plans, and shared views only ever see the reference — never credentials. If access changes, you fix it once at the integration, not in every workflow.
• Workflows express usage intent, not installation state: which source to read, which destination to write. Installing and configuring happens in Integrations, using happens in Pipeline Studio.
• Ontology and predictions read through sources, so stale or broken connections surface downstream as stale evidence — always traceable to the same sync state.

What this means for you

• Start narrow (one system, one scope) and widen after the first result.
• When outputs look wrong, check sync state before blaming the workflow.
• Grant writes the minimum scope the job needs; reads can be broader.

Image: diagram of source → workflow → destination with secrets pinned at the integration layer.

Next

• Connect one: Connect data.
• Field-level detail: Data endpoints, Integrations.