Skip to main content
Every connector is made of taps. A tap is a single stream of records from the source: Linear issues, GitHub pull_requests, Slack channels. The Linear connector, for example, includes sixteen taps, from issues and projects to comments and workflow_states. Taps are the unit you control: each one can be enabled or disabled per connector, and each one lands as its own set of tables in your data lake. The connector definition describes how Parable fetches that stream. You inspect that on connector.taps[] of an instance. You do not set transport, pagination, or sync mode; they are how that source API works. How ingestion walks a source follows those fields through a real Linear query and a GitHub parent-child fan-out. A tap looks like this (fields omitted):

Transport

Each tap has one or more transport options. An option is one way to call the source: priority chooses among options: 1 is highest. If the preferred API is missing on your plan or edition, Parable can fall through to the next option. GitHub list endpoints are typically rest; Linear list endpoints are typically graphql.

Pagination

Sources page results in different ways. The transport option’s pagination.type is how Parable walks pages until the stream is exhausted: You do not pick this. It is why a GitHub tap and a Linear tap can share the same sync engine and still talk to two different paging contracts.

Sync mode

sync.mode is how much is fetched each run: Incremental sync is why connectors stay fast against rate-limited APIs even for large workspaces. The first incremental run may still walk a lookback window (maxLookbackDays when the source retains only recent data). How the watermark is sent to the source is incrementalParams.filterType: Some APIs use a create-poll-fetch job (sync.asyncQuery). You will see that block on the tap when the source works that way.

Record shape

schema on the platform tap is the shape of records the stream emits. Incoming data is checked against it during ingestion, so malformed source records are caught at the boundary instead of corrupting lake tables. Flags on that schema (x-transformDedupKey, x-transformOrdering, identity tags) are how the row lands uniquely in the providers pool. See tap projection. Your workspace can narrow that shape with outputSchema on a workspace tap overlay. For an override, the overlay must stay a structural subset of the platform tap’s schema. enabledByDefault is the definition default. Your overlay’s enabled flag is what actually runs.

Managing taps via the API

PATCH can change enabled, name, description, outputSchema, and config. It cannot change sync mode, pagination, or transport. Those stay on the platform tap. See the API Reference for full request and response shapes.