Vertex AI uses the gcloud CLI to list available AI models for a Google Cloud project and region. No API token or browser cookies.
Data sources + fallback order
- gcloud CLI (
gcloud ai models list)- No fallback — single-source CLI strategy.
Credentials (if applicable)
- Requires
gcloudCLI installed and authenticated (gcloud auth login). - Authenticated via standard Google Cloud ADC (Application Default Credentials) —
GOOGLE_APPLICATION_CREDENTIALSenv var orgcloud auth application-default login.
API endpoints / Probe mechanism
- Strategy:
vertexai.cli(kind:.cli) - Command:
gcloud ai models list --project=<project> --region=<location> --format=json - Timeout: 20 seconds
- PATH augmentation: Ensures
/opt/homebrew/bin,/usr/local/bin,/usr/bin,/bin,/usr/sbin,/sbinare on PATH.
Configuration resolution (in order)
Project:
- Preferences setting
vertexai.project VERTEX_AI_PROJECTenv varGOOGLE_CLOUD_PROJECTenv varGCLOUD_PROJECTenv var
Location (defaults to us-central1):
- Preferences setting
vertexai.location VERTEX_AI_LOCATIONenv varGOOGLE_CLOUD_REGIONenv var
Parsing + mapping
- gcloud returns a JSON array of models (not a wrapped object).
VertexAIModelsResponsedecodes the top-level array directly. - Each model has
name(resource path) anddisplayName(human label). - Snapshot shows model count, project, region, and up to 3 display names as a preview.
- No quota/rate-limit windows —
usedPercentis always 0,hasKnownLimitis false. - Primary
RateWindowlabel: “AI models”. - Supports model breakdown via
supportsModelBreakdown: true.
Key files
Sources/RunicCore/Providers/VertexAI/VertexAIProviderDescriptor.swiftSources/RunicCore/Providers/VertexAI/VertexAISettingsReader.swiftSources/RunicCore/Providers/VertexAI/VertexAIUsageFetcher.swift
Read the source ↗ Checked against Runic 2.13.5 · 9730f91