Semantic Search with Neo4j and Google Vertex AI

A semantic search engine built with Neo4j and Google Cloud Vertex AI that enables intelligent searching of executive profiles. The system uses vector embeddings generated from executive biographies (using Google's textembedding-gecko@003 model) stored in a Neo4j graph database.

Setup

1. Clone the repository

git clone <repository-url>
cd neo4j-vertex-semantic-search

2. Install dependencies

npm install

3. Configure environment variables Create a .env file in the root directory:

NEO4J_URI="neo4j+s://xxxxx.databases.neo4j.io"
NEO4J_USER="neo4j"
NEO4J_PASSWORD="your-password"
GOOGLE_APPLICATION_CREDENTIALS="/path/to/service-account.json"
PROJECT_ID="your-google-project-id"

4. Load sample data

npm run load-data

Technical Implementation

1. Core Components

   • vertex-ai-client.js - Handles Vertex AI embedding generation
   • executive-bio-vectorizer.js - Manages bio vectorization and index creation
   • query-client.js - Handles semantic search queries

2. Project Dependencies

   {
     "dependencies": {
       "@google-cloud/aiplatform": "^3.31.0",
       "axios": "^1.7.7",
       "google-auth-library": "^9.14.2",
       "neo4j-driver": "^5.26.0",
       "dotenv": "^16.4.5"
     }
   }