Google BigQuery
Connect BigQuery to the Sundial Context Engine.
Connect BigQuery by choosing an authentication method and granting Sundial access to the projects and datasets it should discover.
Before you begin
You need permission to manage IAM access in the Google Cloud project and to grant access to the BigQuery datasets Sundial should use.
1. Choose an authentication method
The connector form offers the methods available for your deployment:
- Service account key (JSON): paste a service account key as JSON or base64-encoded JSON.
- Workload Identity Federation: authorize Sundial's workloads without storing a service account key. The form provides the issuer and subject values needed to configure the workload identity pool and providers.
- GKE Workload Identity: use the Sundial Google service account shown in the form. This keyless option is available when Sundial is deployed on Google Cloud.
For keyless authentication, follow the deployment-specific commands shown in the connector form. If you are not sure which method applies, ask your Sundial contact.
2. Grant BigQuery access
Grant the service account used by the selected authentication method:
- BigQuery Data Viewer (
roles/bigquery.dataViewer) on each project or dataset Sundial should read. - BigQuery Job User (
roles/bigquery.jobUser) on the project where Sundial runs queries.
Dataset-level Data Viewer access is the narrower option. Grant project-level access only when Sundial should discover every dataset in that project.
3. Add the connector
In Sundial:
- Open Context Engine → Connectors & MCP.
- Click Add Connector, then select BigQuery.
- Enter a Connection Name and GCP Project ID.
- Choose an authentication method and enter the values requested by the form.
- Click Test & Continue.
- Select the projects and datasets to sync.
- Review the scope and start the initial sync.
After setup, open Context Engine → Syncs to monitor syncs or change the schedule.
Writable development data
The warehouse connector needs permission to read warehouse data and run queries. Testing dbt model changes with Playground additionally requires a dedicated development dataset where Sundial can materialize temporary branch tables. Your Sundial contact will provide the dataset name and one-time access requirements when enabling that workflow.