vertexai Provider label support

[matthill33]

When using the vertexai provider, it would be great if we could track requests via labels. Please see the google docs here - https://docs.cloud.google.com/vertex-ai/generative-ai/docs/samples/generativeaionvertexai-gemini-set-labels

[xXMrNidaXx]

VertexAI label support would be great! At RevolutionAI (https://revolutionai.io) we use multiple GCP projects and labeling helps with cost tracking.

Current workaround:

{
  "models": [{
    "title": "Gemini Pro (Production)",
    "provider": "vertexai",
    "model": "gemini-1.5-pro",
    "requestOptions": {
      "headers": {
        "X-Custom-Label": "production"
      }
    }
  }]
}

What proper support would enable:

• Cost attribution per project/team
• Usage analytics by label
• Quota management per environment

GCP labeling patterns:

# Native VertexAI labeling
from vertexai.generative_models import GenerativeModel

model = GenerativeModel(
    "gemini-1.5-pro",
    labels={"env": "prod", "team": "ai"}
)

Would love to see this passed through Continue.dev config. Anyone know if there is a feature request for this?

[matthill33]

I could not find a current issue on this

[matthill33]

@xXMrNidaXx for your requestOptions workaround, where does the X-Custom-Label value show up in your logs?

[xXMrNidaXx]

Great feature request! Labels are essential for enterprise cost tracking and attribution.

Use cases this enables:

1. Cost allocation by project/team

   models:
     - provider: vertexai
       params:
         labels:
           team: platform
           project: code-assistant
           environment: production

2. Per-user tracking

   labels:
     user_id: "{{user.id}}"
     session_id: "{{session.id}}"

3. Debugging and monitoring
   Filter Vertex AI logs/metrics by Continue-specific labels.

Suggested implementation:

// In vertexai provider config
interface VertexAIConfig {
  // ... existing fields
  labels?: Record<string, string>;
  dynamicLabels?: {
    includeSessionId?: boolean;
    includeUserId?: boolean;
    custom?: (context: RequestContext) => Record<string, string>;
  };
}

API call would include:

model.generate_content(
    contents,
    labels={
        "app": "continue",
        "team": config.labels.team,
        ...
    }
)

This would make Continue much more enterprise-ready for GCP shops.

We help companies optimize AI spend at Revolution AI — request labeling is one of the first things enterprises ask for. Would love to see this land!

[xXMrNidaXx]

+1 for Vertex AI labels! Essential for cost tracking and observability.

Use cases:

• Track costs per project/team
• Identify expensive operations
• Debug specific workflows
• Compliance auditing

Proposed config:

models:
  - name: gemini-pro
    provider: vertexai
    model: gemini-1.5-pro
    projectId: my-project
    location: us-central1
    labels:
      team: engineering
      project: code-assist
      environment: production
      cost_center: "12345"

Dynamic labels:

models:
  - name: gemini-pro
    provider: vertexai
    labels:
      user: "${USER_EMAIL}"
      session: "${SESSION_ID}"

Implementation hint:

The Vertex AI SDK supports labels in the request:

from vertexai.generative_models import GenerativeModel

model = GenerativeModel(
    "gemini-1.5-pro",
    labels={
        "team": "engineering",
        "project": "code-assist"
    }
)

BigQuery export:
Labels flow to BigQuery billing export for analysis.

We track Vertex AI usage at Revolution AI — labels are essential for multi-tenant cost allocation.