vertex-ai

name	vertex-ai
description	Vertex AI API reference for Gemini and Claude endpoints. Use when user says "vertex", "gemini api", "claude api", "vertex ai", or needs to make API calls to Google AI models or Anthropic models via Vertex AI.
user-invocable	false

Vertex AI API Reference

Project: gen-lang-client-0234791928 | Region: us-central1

Gemini Endpoints

Generate Content:

POST https://us-central1-aiplatform.googleapis.com/v1/projects/gen-lang-client-0234791928/locations/us-central1/publishers/google/models/{MODEL}:generateContent

Models: gemini-2.0-flash, gemini-2.0-pro, gemini-1.5-pro, gemini-1.5-flash

Auth: Authorization: Bearer $(gcloud auth print-access-token)

Request body:

{
  "contents": [
    {
      "role": "user",
      "parts": [{"text": "Your prompt here"}]
    }
  ],
  "generationConfig": {
    "temperature": 0.7,
    "maxOutputTokens": 8192
  }
}

Response path: $json.candidates[0].content.parts[0].text

Claude via Vertex AI

Raw Predict:

POST https://us-central1-aiplatform.googleapis.com/v1/projects/gen-lang-client-0234791928/locations/us-central1/publishers/anthropic/models/{MODEL}:rawPredict

Models: claude-sonnet-4-20250514, claude-haiku-35-20241022

Auth: Authorization: Bearer $(gcloud auth print-access-token)

Request body (MUST include anthropic_version):

{
  "anthropic_version": "vertex-2023-10-16",
  "messages": [
    {"role": "user", "content": "Your prompt here"}
  ],
  "max_tokens": 4096
}

Response path: $json.content[0].text

n8n Integration

When using these endpoints in n8n HTTP Request nodes:

• Use Google Cloud OAuth2 credential for auth
• The ENTIRE jsonBody must start with = for expression mode
• Gemini response: {{ $json.candidates[0].content.parts[0].text }}
• Claude response: {{ $json.content[0].text }}

Quick Test (bash)

# Test Gemini
curl -X POST \
  "https://us-central1-aiplatform.googleapis.com/v1/projects/gen-lang-client-0234791928/locations/us-central1/publishers/google/models/gemini-2.0-flash:generateContent" \
  -H "Authorization: Bearer $(gcloud auth print-access-token)" \
  -H "Content-Type: application/json" \
  -d '{"contents":[{"role":"user","parts":[{"text":"Hello"}]}]}'

Common Errors

• 403 Forbidden: Token expired. Run gcloud auth print-access-token to refresh.
• 404 Not Found: Wrong model name. Check available models with gcloud ai models list --region=us-central1.
• 400 Bad Request (Claude): Missing anthropic_version field in request body.