Skip to content

google-vertex.md

Latest commit

 

History

History
75 lines (64 loc) · 2.38 KB

google-vertex.md

File metadata and controls

75 lines (64 loc) · 2.38 KB

Google Vertex AI Integration Guide for kconsole

Prerequisites

  • GCP Project active.
  • Billing enabled.
  • Vertex AI API enabled.
  • Service Account with "Vertex AI User" role.

Step 1: GCP Setup

  1. Go GCP Console.
  2. Create or select existing Project. Note PROJECT_ID.
  3. Go APIs & Services. Search "Vertex AI API". Click Enable.

Step 2: Authentication Configuration

  1. Go IAM & Admin -> Service Accounts.
  2. Create Service Account. Name: kconsole-vertex-sa.
  3. Grant Role: Vertex AI User.
  4. Keys -> Add Key -> Create New Key -> JSON.
  5. Download JSON key file.
  6. Set environment variable for kconsole runtime: export GOOGLE_APPLICATION_CREDENTIALS="/path/to/kconsole-vertex-sa.json"

Step 3: API Endpoint Definition

Vertex AI requires specific region endpoints. kconsole routing logic must handle this.

Base URL Pattern: https://[REGION]-aiplatform.googleapis.com/v1/projects/[PROJECT_ID]/locations/[REGION]/publishers/google/models/[MODEL_ID]:streamGenerateContent

Common Regions: us-central1, europe-west4, asia-southeast1. Common Models: gemini-3-flash-preview, gemini-3.1-flash-lite-preview, gemini-3.1-pro-preview, and image models like imagen-3.0-generate-001 (or latest preview).

Step 4: Token Generation (Gateway side)

kconsole must dynamically generate OAuth tokens using the Service Account JSON.

  • If Node/Python: use official Google Auth SDK to fetch token.
  • HTTP Header required for request: Authorization: Bearer <OAUTH_TOKEN>

Step 5: Payload Translation (Adapter layer)

kconsole must adapt standard OpenAI-style payloads to Vertex Gemini format.

kconsole Input (OpenAI style):

{
  "model": "gemini-3.1-pro-preview",
  "messages": [{"role": "user", "content": "Hello"}]
}

kconsole Output (Vertex style):

{
  "contents": [
    {
      "role": "user",
      "parts": [{"text": "Hello"}]
    }
  ],
  "generationConfig": {
    "temperature": 0.7
  }
}

Step 6: Integration Test

  1. Start kconsole with Vertex adapter enabled.
  2. Send test cURL to kconsole gateway port:
curl -X POST http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <kconsole_api_key>" \
  -d '{
    "model": "vertex/gemini-3.1-pro-preview",
    "messages": [{"role": "user", "content": "Test vertex integration"}]
  }'
  1. Verify successful response format matches OpenAI standard.