n8n-nodes-vertex-ai-rag

Custom n8n nodes for Google Vertex AI RAG (Retrieval Augmented Generation) with isolated credentials.

Features

• Query Embedding Generator: Generate embeddings from text queries using Vertex AI
• Vertex Vector Search: Search for similar documents in Vertex AI Vector Search
• Isolated Credentials: Separate service accounts for embedding and vector search operations
• Flexible Configuration: Configure project, location, models, and search parameters per node

Installation

Community Nodes (Recommended)

1. Go to Settings > Community Nodes in n8n
2. Select Install
3. Enter n8n-nodes-vertex-ai-rag
4. Agree to the risks and install

Manual Installation

npm install n8n-nodes-vertex-ai-rag

Setup

Prerequisites

1. Google Cloud Project with Vertex AI enabled
2. Two Service Accounts (recommended for security):
   • Embedding SA: For generating embeddings
   • Vector Search SA: For searching indexed documents
3. Service Account JSON keys

IAM Permissions

Embedding Service Account:

gcloud projects add-iam-policy-binding PROJECT_ID \
  --member="serviceAccount:embedding-sa@PROJECT_ID.iam.gserviceaccount.com" \
  --role="roles/aiplatform.user"

Vector Search Service Account:

gcloud projects add-iam-policy-binding PROJECT_ID \
  --member="serviceAccount:vector-search-sa@PROJECT_ID.iam.gserviceaccount.com" \
  --role="roles/aiplatform.indexEndpointUser"

Usage

1. Configure Credentials

Vertex AI Embedding API

• Go to Settings > Credentials
• Add Vertex AI Embedding API credential
• Fill in:
  • Service Account Email
  • Private Key (from JSON key file)

Vertex AI Vector Search API

• Add Vertex AI Vector Search API credential
• Fill in:
  • Service Account Email
  • Private Key (from JSON key file)

2. Workflow Example

Webhook (receive user query)
  ↓
Query Embedding Generator (convert text to vector)
  ↓
Vertex Vector Search (find similar documents)
  ↓
AI Agent (generate answer with context)
  ↓
Respond to Webhook

3. Node Configuration

Query Embedding Generator

• Project ID: Your GCP project ID
• Location: Region (e.g., us-central1)
• Query Text: Text to convert to embedding (e.g., {{ $json.body.chatInput }})
• Embedding Model: Choose model (default: textembedding-gecko@003)

Vertex Vector Search

• Project ID: Your GCP project ID
• Location: Region (e.g., us-central1)
• Index Endpoint ID: Your Vector Search index endpoint ID
• Deployed Index ID: Your deployed index ID
• Query Embedding: Vector from previous node (e.g., {{ $json.embedding }})
• Number of Results: How many similar documents to return (default: 5)

Options:

• Text Metadata Field: Metadata field containing document text (default: text)
• Source Metadata Field: Metadata field containing document source (default: source)
• Context Separator: Separator between documents (default: \n\n---\n\n)

Output Format

Query Embedding Generator Output

{
  "queryText": "How to reset password?",
  "embedding": [0.123, -0.456, 0.789, ...],
  "embeddingDimension": 768
}

Vertex Vector Search Output

{
  "context": "[Document 1]\nSource: manual.pdf\n\nTo reset password...\n\n---\n\n[Document 2]...",
  "documents": [
    {
      "rank": 1,
      "text": "To reset password, go to settings...",
      "source": "manual.pdf",
      "distance": 0.15,
      "id": "doc_123"
    }
  ],
  "documentsCount": 2
}

Security Best Practices

• ✅ Use separate service accounts for each operation
• ✅ Grant minimal required permissions (principle of least privilege)
• ✅ Never hardcode credentials in workflows
• ✅ Store credentials securely in n8n's credential manager
• ✅ Rotate service account keys regularly

Troubleshooting

Authentication Errors

• Verify service account has correct IAM roles
• Check private key format (should include \n for line breaks)
• Ensure service account is enabled

Vector Search Not Found

• Verify index endpoint ID is correct
• Confirm deployed index ID matches your deployment
• Check that index is deployed and online

Embedding Dimension Mismatch

• Ensure the embedding model matches the one used during indexing
• Common dimensions: 768 (gecko@003), 768 (gecko@002)

Development

# Clone repository
git clone https://github.com/yourusername/n8n-nodes-vertex-ai-rag.git
cd n8n-nodes-vertex-ai-rag

# Install dependencies
npm install

# Build
npm run build

# Link for local development
npm link
cd ~/.n8n/custom
npm link n8n-nodes-vertex-ai-rag

License

MIT

Contributing

Contributions are welcome! Please open an issue or submit a pull request.

Support

• n8n Community Forum
• Google Vertex AI Documentation
• GitHub Issues