@propickler/localvectordb v1.0.0

LocalVectorDB

A lightweight local vector database implementation in Node.js, similar to ChromaDB. This project provides a simple and efficient way to store, index, and query high-dimensional vectors with similarity search capabilities.

Features

• Vector storage and indexing using HNSW algorithm
• Similarity search (cosine similarity)
• Collection-based organization
• REST API interface
• Metadata support for each vector
• Automatic ID generation

Prerequisites

• Node.js (v14 or higher)
• npm (Node Package Manager)

Installation

1. Install the package from npm:

npm install localvectordb

2. Or clone the repository and install dependencies:

git clone https://github.com/yourusername/localvectordb.git
cd localvectordb
npm install

Quick Start

1. Start the server:

node src/index.js

The server will start on port 3000 by default. You can change this by setting the PORT environment variable.

API Usage

1. Create a Collection

Collections are used to organize your vectors. Each collection requires a name and dimension size for the vectors it will store.

curl -X POST http://localhost:3000/collections \
  -H "Content-Type: application/json" \
  -d '{
    "name": "my_collection",
    "dimension": 3
  }'

2. Add Vectors

Add vectors to your collection along with optional metadata:

curl -X POST http://localhost:3000/collections/my_collection/add \
  -H "Content-Type: application/json" \
  -d '{
    "ids": ["vec1", "vec2"],
    "embeddings": [
      [1.1, 2.2, 3.3],
      [4.4, 5.5, 6.6]
    ],
    "metadatas": [
      {"description": "first vector"},
      {"description": "second vector"}
    ]
  }'

3. Query Similar Vectors

Search for similar vectors using cosine similarity:

curl -X POST http://localhost:3000/collections/my_collection/query \
  -H "Content-Type: application/json" \
  -d '{
    "queryEmbeddings": [[1.0, 2.0, 3.0]],
    "nResults": 2
  }'

Response format:

{
  "status": "success",
  "results": [
    [
      {
        "id": "vec1",
        "distance": 0.0023,
        "metadata": {
          "description": "first vector"
        }
      },
      {
        "id": "vec2",
        "distance": 0.0156,
        "metadata": {
          "description": "second vector"
        }
      }
    ]
  ]
}

Node.js Client Usage

You can also use LocalVectorDB directly in your Node.js applications:

const VectorDB = require('localvectordb');

// Create a new VectorDB instance
const db = new VectorDB();

// Create a collection
const collection = await db.createCollection('my_collection', 3);

// Add vectors
await collection.add({
  ids: ['vec1', 'vec2'],
  embeddings: [
    [1.1, 2.2, 3.3],
    [4.4, 5.5, 6.6]
  ],
  metadatas: [
    { description: 'first vector' },
    { description: 'second vector' }
  ]
});

// Query similar vectors
const results = await collection.query({
  queryEmbeddings: [[1.0, 2.0, 3.0]],
  nResults: 2
});

API Reference

Collections

Parameters

Create Collection

• name (string): Name of the collection
• dimension (number): Dimension of vectors to be stored

Add Vectors

• ids (string[]): Optional array of IDs for the vectors
• embeddings (number): Array of vectors to add
• metadatas (object[]): Optional array of metadata objects

Query Vectors

• queryEmbeddings (number): Array of query vectors
• nResults (number): Number of similar vectors to return

Error Handling

The API returns appropriate HTTP status codes:

• 200: Success
• 400: Bad Request (invalid parameters)
• 404: Collection not found
• 500: Server error

Error responses include a message explaining the error:

{
  "status": "error",
  "message": "Collection not found"
}

Limitations

• Maximum vectors per collection: 100,000 (configurable)
• Maximum dimension size: No hard limit, but performance may degrade with very high dimensions
• Distance metric: Currently only supports cosine similarity

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

MIT License - feel free to use this in your own projects!