Building an Intelligent Recommendation System: Leveraging Node.js, PGVector, and Google Gemini for Personalized Content Discovery

This project provides a service to embed text using Google's Generative AI and store the embeddings in a PostgreSQL database. It also includes functionality to find similar embeddings and recommend content based on user preferences.

Prerequisites

• Node.js
• PostgreSQL
• Google Generative AI API Key

Setup

1. Clone the repository:

   git clone <repository-url>
   cd <repository-directory>

2. Install dependencies:

   npm install

3. Environment Variables: Create a .env file in the root directory and add the following:

   API_KEY=your_google_generative_ai_api_key
   DATABASE_URL=your_postgresql_database_url
   ACCESS_TOKEN_SECRET=your_jwt_secret

4. Setup the database:

   npx ts-node src/setupDatabase.ts

Running the Service

1. Start the server:

   npx ts-node src/index.ts

2. The server will run on http://localhost:9000.

API Endpoints

User Management

Create User

• URL: /user
• Method: POST
• Body:

  {
      "name": "username",
      "password": "userpassword",
      "preferences": { /* user preferences */ }
  }

• Response:

  {
      "userId": "generated_user_id",
      "name": "username",
      "preferences": { /* user preferences */ }
  }

User Login

• URL: /login
• Method: POST
• Body:

  {
      "username": "username",
      "password": "userpassword"
  }

• Response:

  {
      "accessToken": "jwt_access_token"
  }

Embedding and Similarity

Embed Text

• URL: /embed
• Method: POST
• Headers:

  {
      "Authorization": "Bearer jwt_access_token"
  }

• Body:

  {
      "text": "Your text to embed"
  }

• Response:

  {
      "userId": "user_id",
      "text": "Your text to embed",
      "embedding": [/* embedding vector */]
  }

Find Similar Embeddings

• URL: /similarity
• Method: POST
• Headers:

  {
      "Authorization": "Bearer jwt_access_token"
  }

• Body:

  {
      "text": "Your text to find similarity"
  }

• Response:

  [
      {
          "text": "Similar text",
          "vector": [/* embedding vector */],
          "distance": 0.123
      },
      // more results
  ]

Recommendations

Recommend Similar Texts

• URL: /recommend
• Method: POST
• Headers:

  {
      "Authorization": "Bearer jwt_access_token"
  }

• Body:

  {
      "text": "Your text to get recommendations"
  }

• Response:

  [
      {
          "text": "Recommended text",
          "similarity": 0.877
      },
      // more recommendations
  ]

Postman Documentation Here

License

This project is licensed under the MIT License.