Skip to content

Latest commit

 

History

History
178 lines (127 loc) · 3.59 KB

README.md

File metadata and controls

178 lines (127 loc) · 3.59 KB
Raw

relate-text

relate-text is a versatile library for creating, storing, updating, and querying text embeddings using neural embeddings. It supports operations via a REST API, a CLI, or programmatic interfaces, making it a powerful tool for text similarity, semantic search, and other embedding-based applications.

Features

  • Create and manage text embeddings.
  • Query for similar text based on embeddings.
  • REST API server with predefined endpoints.
  • Command-line interface (CLI) for embedding management.

Installation

Install the library using npm:

npm install relate-text

For global CLI usage:

npm install -g relate-text

Usage

Programmatic Usage

You can directly use relate-text in your Node.js applications.

Example:

import { TextToEmbeddingController } from 'relate-text';

(async () => {
  const controller = new TextToEmbeddingController();
  await controller.ready();

  // Create a new embedding
  await controller.create('example-id', 'This is a sample text', { tag: 'example' });

  // Query for similar text
  const results = await controller.retrieveSimilar('sample text', 5);
  console.log('Similar results:', results);

  // Update the embedding
  await controller.update('example-id', 'Updated text content', { tag: 'updated' });

  // Delete the embedding
  await controller.destroy('example-id');

  // Delete all embeddings
  await controller.destroyAll();
})();

REST API

You can start the REST API server to interact with embeddings.

Starting the Server

Run the server with:

relate-text start-server --port 3000

Available Endpoints

  • Create an Embedding
    POST /embeddings
    Body:

    {
  "id": "example-id",
  "text": "This is a sample text",
  "metadata": { "tag": "example" }
}

  • Update an Embedding
    PUT /embeddings
    Body:

    {
  "id": "example-id",
  "text": "Updated text content",
  "metadata": { "tag": "updated" }
}

  • Delete an Embedding
    DELETE /embeddings/:id

  • Delete All Embeddings
    DELETE /embeddings/all

  • Query Similar Text
    POST /similar
    Body:

    {
  "text": "sample text",
  "limit": 5
}

Command-Line Interface (CLI)

The CLI provides easy access to all major operations.

Commands

  1. Create an Embedding

    relate-text create -i "example-id" -t "This is a sample text" -m '{"tag":"example"}'

  2. Update an Embedding

    relate-text update -i "example-id" -t "Updated text content" -m '{"tag":"updated"}'

  3. Delete an Embedding

    relate-text delete -i "example-id"

  4. Delete All Embeddings

    relate-text delete-all

  5. Find Similar Text

    relate-text similar -t "sample text" -l 5

  6. Start the Server

    relate-text start-server --port 3000

Configuration

Metadata

Metadata can be provided as a JSON object during create or update operations. It is optional and useful for tagging or storing additional information about embeddings.

Development

Clone the repository:

git clone https://github.com/yourusername/relate-text.git

Install dependencies:

npm install

Build the library:

npm run build

Contributing

Contributions are welcome! Please fork the repository and submit a pull request with your changes. Ensure all tests pass and your code follows the existing style.

Acknowledgments

  • Built with Node.js, Express, and Commander.js.
  • Uses @xenova/transformers for generating embeddings.