For BigQuery
AI analytics for BigQuery
Last updated
Short answer
exelens connects to BigQuery with a read-only service account, preferably through Workload Identity Federation so no JSON key is stored, reads only the views you import and answers questions in plain language with an AI model on your own key.
| Detail | exelens on BigQuery |
|---|---|
| Preferred sign-in | Workload Identity Federation (no key file stored). A service account JSON key is the fallback |
| What exelens can read | roles/bigquery.dataViewer only on the datasets that hold the views you import |
| Compute | roles/bigquery.jobUser on the project, so queries run as jobs in your project |
| Network | Federation trusts the exelens AWS account; no inbound firewall change to your datasets |
| What you provide | Project ID, dataset location and the federation config or key |
| Where the data lives | In BigQuery. Rows are loaded temporarily only while a question is answered |
How it works on BigQuery
- Create a reader service account. Grant
roles/bigquery.jobUseron the project androles/bigquery.dataVieweronly on the datasets that hold the views. - Set up federation. Create a workload identity pool with an AWS provider that trusts the exelens account. The setup script in Data → Connect generates the commands.
- Paste the config. Add the project ID, location and the federation JSON.
- Test, then import views. exelens lists the views in your datasets; you choose which ones people may ask about.
A good fit for marketing and product data
BigQuery often holds web, ad and product data. Point exelens at a few curated views, such as spend by channel, and a marketing leader can ask CAC by channel for the last 90 days without a ticket.
Keeping cost in check
BigQuery bills the queries, so keep imported views lean and partitioned. See cost control. To see it on your own project, book a demo.
Frequently asked questions
- Do I have to upload a service account key?
- No. Workload Identity Federation is preferred, and nothing long-lived is stored. A key file is available as a fallback.
- Who pays for BigQuery queries?
- Your Google Cloud project, as with any query it runs. AI usage is billed separately by your AI provider.
- Which dataset location should I enter?
- The region of the dataset, such as US or europe-west1.
See exelens on your BigQuery
A short walkthrough with the founder: connect read-only, ask a real question, pin the answer.