Firestore Vector Search

A Node.js application that demonstrates vector search capabilities using Firebase Firestore and OpenAI embeddings. This project allows you to generate embeddings for city names and perform semantic searches to find similar cities based on natural language queries.

Features

• Generate embeddings for city names using OpenAI's text-embedding-ada-002 model
• Store embeddings in Firebase Firestore using vector fields
• Perform semantic similarity searches using cosine distance
• Command-line interface for easy searching
• Batch processing with error handling

Prerequisites

• Node.js 16 or higher
• Firebase project with Firestore enabled
• OpenAI API key
• Firebase service account credentials

Setup

1. Clone the repository:

git clone <your-repo-url>
cd city-vector-search

2. Install dependencies:

npm install

3. Set up environment variables:
   • Copy .env.example to .env
   • Add your OpenAI API key
   • Add path to your Firebase service account JSON file

OPENAI_API_KEY=your_openai_api_key_here
FIREBASE_SERVICE_ACCOUNT_PATH=service-account.json

4. Set up Firebase:

   • Go to Firebase Console
   • Create a new project or select existing one
   • Go to Project Settings > Service Accounts
   • Click "Generate New Private Key"
   • Save the JSON file as service-account.json in your project root

5. Create a vector index in Firestore:

   • Navigate to your Firebase Console
   • Go to Firestore
   • Create a new collection called city_embeddings
   • Create a vector index for the city_embedding field with dimension 1536

Usage

Generate Embeddings

To generate and store embeddings for the predefined list of cities:

npm run generate

This will:

• Generate embeddings for each city using OpenAI
• Store the embeddings in Firestore
• Show progress and success/failure counts

Search Cities

To search for similar cities:

npm run search "<search query>" [limit]

Examples:

npm run search "modern city with great technology"
npm run search "historic European city with art museums" 10

The search will return the most semantically similar cities along with their similarity scores.

Environment Variables

Variable	Description	Required
OPENAI_API_KEY	Your OpenAI API key	Yes
FIREBASE_SERVICE_ACCOUNT_PATH	Path to Firebase service account JSON	Yes