Connector docs — Vertex AI / Gemini
Exactly what CostMon asks for to read Vertex AI / Gemini
CostMon reads the same FOCUS-formatted GCP billing export as the gcp connector, filtered to Vertex AI / Generative Language services, and joins it with Cloud Monitoring's per-model token metrics.
What CostMon calls
The literal API surface, nothing more
- BigQuery — Jobs.query
- Billing export filtered to Vertex AI / generative services, grouped by model.
- Cloud Monitoring — TimeSeries.list
- Per-model token counts from aiplatform.googleapis.com/publisher/online_serving/token_count.
What you supply
Credential and config fields
Read-only: roles/bigquery.dataViewer + jobUser on the billing project, roles/monitoring.viewer
Minimum permission
Copy this, not a broader role
Scope of access
What Vertex AI / Gemini access can't reach
- Prompts, completions, or any model input/output content
- Any GCP resource outside the billing export and Cloud Monitoring metrics
Worth knowing
- Dollars come from the GCP FOCUS billing export filtered to Vertex AI; per-model token counts come from Cloud Monitoring and degrade to cost-only if that metric is unavailable.
FAQ
Common questions about the Vertex AI / Gemini connector
Is this a separate billing export from the gcp connector?
No, it can be the same export. This connector filters it down to Vertex AI / generative services and adds a Cloud Monitoring token join on top.
What if Cloud Monitoring doesn't have the token metric?
Rows still sync from the billing export. You lose per-model token counts, not the cost, the same degrade-gracefully behavior as the Bedrock connector.
A rate you looked up will change again.
Provider pricing pages change without notice. CostMon tracks the bill that pays for it.