Pipelines

A pipeline takes data from a source, applies a transform, and loads the result into a Pact object. Pipelines are managed at /v1/etl/pipelines (backed by the etl_pipelines + etl_runs tables, alembic 0098_etl_pipelines) and listed in the app under Data → Pipelines.

Each pipeline row holds its definition plus a rolled-up snapshot of its most recent run (last_run_at, last_run_status), and every run is its own etl_runs row.

Pipelines can't run yet

Pipeline CRUD is real and tenant-isolated, but Pact has no ETL executor: nothing reads a pipeline's source, applies its transform or loads its destination. So Run is not offered. POST /v1/etl/pipelines/{id}/runs answers status: "unsupported" with a gap that says what is missing and what you can do instead, and it writes nothing: no run row and no last-run snapshot.

If the pipeline's source is Salesforce, HubSpot or Pipedrive, that source's Sync now (Data → Sources) runs the real connector and reports the records it imported and updated.
Otherwise, the import wizard (Admin → Import) loads a file and reports the rows it wrote.

An earlier build recorded simulated runs with invented counts. Those rows are kept, but they are reported as Not measured, with no counts, timings or log. Seeded sample pipelines keep their figures, labeled Sample.

Destinations

A pipeline's destination is the Pact object it loads into, validated against a fixed allowlist (ALLOWED_DESTINATIONS):

companies · contacts · events · opportunities · tags · tracking_events · custom

Pipelines carry a schedule field (default manual) and a user-settable status of active or paused. The schedule value is stored, but nothing dispatches scheduled runs, and nothing executes a run.

Endpoints

code

POST   /v1/etl/pipelines               create
GET    /v1/etl/pipelines               list
GET    /v1/etl/pipelines/{id}          fetch one
PATCH  /v1/etl/pipelines/{id}          update
DELETE /v1/etl/pipelines/{id}          delete + cascade runs
GET    /v1/etl/pipelines/{id}/runs     paginated run history
POST   /v1/etl/pipelines/{id}/runs     run, or refuse with a gap (no executor yet)

bash

curl -X POST https://api.pact.place/v1/etl/pipelines \
  -H "Authorization: Bearer $PACT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"name":"Salesforce → Contacts","destination":"contacts","transform":{}}'

A pipeline optionally references a source_id; the transform is stored as transform_json. Deleting a pipeline cascades its etl_runs rows.

Pipeline detailRun history, what a run can do today, and how unmeasured runs read.Data sourcesThe upstream connections pipelines read from.Data vaultThe outbound counterpart — export Pact data to your warehouse.