Skip to main content

Create a First Sync

The recommended authoring path is to install the RETL skills and ask your agent to use retl-start-project or retl-create-sync in your project:

retl install-skills .

Those skills should produce code with the same shape shown below. Use this page to understand and review the generated sync.

The normal sync shape is:

  1. Create a source backend.
  2. Declare a Source query.
  3. Map rows into State or Event intent.
  4. Load a destination connector.
  5. Bind the declaration and destination into a Sync.
  6. Run with dry_run=True before writing to the destination.

Source backend​

from retl.backends.duckdb import DuckDBSqlBackend

db = DuckDBSqlBackend(
    database="warehouse.duckdb",
    source_schema="main",
    runtime_schema="retl",
)

Source and State​

import retl

customers = retl.source(
    name="active_customers",
    mode="snapshot",
    backend=db.source_backend(),
    query="""
    select customer_id, email, plan
    from customers
    where email is not null
    """,
)

customer_state = retl.state(
    name="customer_profiles",
    source=customers,
    key={"customer_id": "customer_id"},
    identifiers=[{"type": "email", "value": "email"}],
    payload={"plan": "plan"},
)

Destination and Sync​

destination = retl.destinations.load(
    "retl/klaviyo",
    binding_name="klaviyo_primary",
    credential_namespace="destinations.klaviyo",
    config_namespace="destinations.klaviyo",
)

sync = retl.sync(
    name="customer_profiles_to_klaviyo",
    declaration=customer_state,
    destination=destination,
    surface="profiles",
)

Runner​

result = retl.runner(
    name="customer_profile_run",
    runtime_store=db.runtime_store(),
).run(sync, dry_run=True)

print(result.to_text())

When the dry-run output matches your intent and destination configuration is ready, run the same Sync with dry_run=False.