Connect a data source
OpenDome ingests structured sources through Singer taps (Meltano). The OSS distribution ships standard taps; premium ERP connectors are an Enterprise add-on. For unstructured sources (documents, mail, drives) see Ingest documents.
In the standalone profile these connector routes are served by the cell’s own
tenant-mgmt-api — operational control, not an OSL data surface. Keep it
private behind your own authenticated administrative gateway; in
OpenDome-operated deployments the central control plane fronts the same model.
Built-in taps Apache-2.0
Section titled “Built-in taps ”tap-postgrestap-mysqltap-salesforce- Custom taps — any Singer-spec tap works.
Enterprise Premium connectors (SAP, Workday, banking ERP) ship in the commercial tier.
The flow
Section titled “The flow”-
Provide credentials. Credentials split across a Secret (sensitive) and a ConfigMap (non-sensitive):
PUT /v1/tenants/{t}/connectors/{c}/credentials -
Select streams. Choose which tables/objects to replicate:
PUT /v1/tenants/{t}/connectors/{c}/streams -
Set the replication method per stream (full table, key-based incremental, log-based CDC):
PUT /v1/tenants/{t}/connectors/{c}/replication -
Launch a sync:
POST /v1/tenants/{t}/connectors/{c}/runs
The tap writes Parquet to the tenant bucket
(s3://<bucket>/landing/<connector_id>/…), the ingest Job lands it into
Iceberg (iceberg.raw.*), and dbt curates it into iceberg.curated.* — ready
to be modeled in OSL. Curated data is consumed only through
POST /osl/query; no consumer reads Trino
directly. Orchestration runs in-cell on Dagster —
sync, ingest, dbt and identity-resolution runs chain automatically, with no
control-plane callbacks in standalone mode.
Source preconditions
Section titled “Source preconditions”CDC-based replication needs the right privileges and configuration on the source
(e.g. logical replication on Postgres). Check your source’s requirements before
selecting a log-based method — the repository’s
docs/source-database-setup.md
covers Postgres, MySQL, SQL Server and Oracle, and
docs/connectors.md
is the full catalog + credential-key convention (note: the Postgres credentials
key is user, not username).
Try it with the Segura fixture
Section titled “Try it with the Segura fixture”The repository ships a seeded test Postgres (the fictional Segura
insurance dataset) you can sync end-to-end:
bash scripts/postgres-test-setup.sh deploys it into a test-services
namespace standing in for a customer system. See
docs/segura-demo.md.