Skip to main content
A connection stores reusable access details for an external system. An import uses that connection to read a particular slice of data into a Dawn dataset. For example, a Postgres connection identifies a server and database. An import supplies the query and output dataset. Several imports can reuse the same connection.

​
Choose your integration

The Source selector in Connections → New connection lists the choices available in the platform.

​
Before you start

You need access to Connections in Dawn, permission to read the source data, and an appropriate credential. A source administrator may need to create a service account, grant access, or permit network connectivity. Decide which tables, files, or accounts Dawn needs. Review permissions and credentials before granting access.

​
1. Create the connection

  1. Open Connections in the Agent sidebar.
  2. Select New connection.
  3. Enter a descriptive name and select the source.
  4. Complete the source fields, such as host and database, bucket, or spreadsheet URL.
  5. Select an existing credential or use Create credential. OAuth sources use Authorize account.
  6. Add a description that explains what the source contains, then select Create connection.
Store passwords, tokens, and keys in the credential form. Keep them out of descriptions, ordinary configuration fields, and chat messages. Expected result: the connection appears in the list. A saved connection is not proof that an import has succeeded.

​
2. Inspect the source

Add the connection to a conversation using its Add to chat action, or select it through the conversation context picker. Ask for a small, specific inspection:
Discovery capabilities vary by source. If broad discovery is unavailable, provide a known table, file, or endpoint. A failed discovery request can reflect missing listing permissions even when a specific resource is readable.

​
3. Define and run an import

Tell Dawn the source selection, destination dataset name, and any filters:
Review the generated pipeline, including its connection, query or path, date filters, and output. Run it when ready. Open the output dataset and compare counts, dates, and sample values with the source.

​
4. Make refreshes repeatable

Run the pipeline manually first. Before scheduling it, confirm the intended replace, append, or update behaviour; how late-arriving records are handled; and who will check failures. See pipelines.

​
Disable or remove access

To stop a connection being used for new work, open its Edit form, expand Advanced settings, and set Status to Disabled. Review schedules and other users of the credential before retiring it. Disabling or deleting a Dawn connection is separate from revoking the credential at the provider. It also does not remove datasets already imported into Dawn. See credential lifecycle and workspace cleanup.