AI and cloud costs
BetaConnect Google Cloud billing to Sondola
What Sondola reads from Google Cloud billing: Gemini and Vertex AI spend, and the rest of your cloud. Read-only, with the source and period beside every number.
In beta. You can connect it today and it is built to the vendor's documented API, but it has not yet run for long across many workspaces. Check its first readings against the tool itself.
Read-only. Sondola never changes your data in Google Cloud billing.
- Credential
- Google service account key
- Refresh
- Once a day
- How it connects
- Collected automatically
- Setup time
- About 10 minutes
What Sondola reads from Google Cloud billing
- Gemini and Vertex AI spend
- The rest of your Google Cloud bill
What you need
- Someone who can configure billing export and manage IAM in the project that holds the dataset. Check that this project has billing enabled and is linked to the billing account you want to report.
Collection timing and the period a reading covers are separate. Check the source's data-through date as well as the metric's reporting period.
Connect Google Cloud billing in 5 steps
- Turn on billing export
In the Google Cloud console open Billing, then Billing export, then Standard usage cost, then Edit settings.
The page may start on a project Google made for you, with a name like Cloud Setup Host Project. Pick your own project in the Projects list before you choose a dataset.Google Cloud: Billing export ↗ - Create or choose the dataset
Already exporting? Keep the existing dataset and match its location here. For a new dataset, choose Create new dataset and use the name below. Pick one multi-region, US or EU, then save. This form supports those two locations. A dataset’s location cannot be changed after creation.
Dataset name for a new dataset: billing_export. US multi-region option: US. EU multi-region option: EU.When export is first enabled to US or EU, Google includes data from the start of the previous month. The initial backfill can take up to five days. Single-region exports start on the day you enable them. Re-enabling export does not backfill the gap. - Create a service account
In the same project open IAM and admin, then Service accounts, then Create service account. Use the name below. Skip the optional role step for now.
Service account name: sondola-billing-reader. - Let it read the dataset
In IAM press Grant access, paste the service account email, and give it BigQuery Job User on the project. In BigQuery open the dataset, then Share, then Manage permissions, then Add principal, and give the same email BigQuery Data Viewer on the dataset.
Role on the project: BigQuery Job User. Role on the dataset: BigQuery Data Viewer. - Create a key
Open the service account, then Keys, then Add key, Create new key, JSON, Create. Upload the downloaded file here with the project ID and dataset name.
If your organisation blocks service account keys, ask its administrator whether a project-level exception is permitted. Keep setup within your organisation’s policy.
Enter credentials only in your Sondola workspace. Keep keys out of screenshots, public repositories and support messages.
What Sondola asks you for
IAM and admin, Service accounts, sondola-billing-reader, Keys, Add key, JSON. Upload the file Google downloads.
If you delete this key in Google Cloud, AI spend falls back to your books and bank until you add a new one.The project that holds the billing export dataset. Sondola runs its query in this project.
The BigQuery dataset you chose when you turned on billing export.
Leave empty and Sondola finds the export table.
Where the dataset lives. Most billing exports are in US.
Secrets are stored encrypted in your workspace, in a vault kept separate from the application database. Sondola never shows them again.
Permissions
- BigQuery Job User on the project and BigQuery Data Viewer on the billing export dataset.
- Only reads. Sondola never writes to your Google Cloud project.
Refresh and first reading
Once a day. Most tools refresh once or twice a day, and every number shows when it was read.
Save and test runs one small query on your billing table. BigQuery counts the data each daily read scans against your project's query usage. Only reads. Sondola never writes to your tools. The table can take a few hours to appear after you turn export on. If Save and test cannot find it, wait and test again. Initial backfill can take up to five days to catch up.
Troubleshooting
Save and test rejects the credential.
If you delete this key in Google Cloud, AI spend falls back to your books and bank until you add a new one. The page may start on a project Google made for you, with a name like Cloud Setup Host Project. Pick your own project in the Projects list before you choose a dataset.
The test passes but no reading appears.
The collector runs on its schedule, so a successful test does not mean a reading has landed. Check the source's data-through date and the metric's reporting period, and treat a missing reading as missing, not zero.
A number differs from Google Cloud billing.
Compare the definition, unit and reporting period with the same number in Google Cloud billing before treating a difference as an error. When a stored reading changes, its history keeps the old value, the new value and the reason.
Provider documentation
Google Cloud billing documentation
Readings missing or old? Read the setup and troubleshooting guide and the worked freshness example.
Related integrations
See your company clearly by Monday.
Apply in two minutes. Once you are in, your first numbers arrive minutes after you connect a tool.
Free for qualifying startups. Read-only.