Embeddings-Redis

This is a Node.js application that uses OpenAI's API and a Redis database to create and store embeddings of text data. It provides a command-line interface for interaction and supports a mode for ingesting new text data.

Requirements

• Node.js 14.x or later
• OpenAI API key
• Redis instance running, with the Redisearch module installed.

This application uses TypeScript, and thus requires ts-node and typescript packages for execution. It also relies on dotenv for environment variable management, the official openai npm package to interact with the OpenAI API, and redis npm package to interact with a Redis database.

Installation

Clone the repository and install the dependencies:

git clone https://github.com/yourusername/embeddings-redis.git
cd embeddings-redis
npm install

Duplicate the .env.sample file in the project root, rename it to .env and replace the placeholders with your actual OpenAI API key and Redis URL.

cp .env.sample .env

Then edit the .env file with your preferred text editor.

Ensure that you have a running Redis instance with the Redisearch module installed. If you're using Docker, you can use the redislabs/redisearch image which comes with the Redisearch module preinstalled.

Usage

There are two scripts defined in the package.json file:

1. ingest: Ingests new text data and stores the corresponding embeddings in the Redis database.
2. start: Starts the application in an interactive mode.

Data Ingestion

The ingest script ingests new text data, creates embeddings using the OpenAI API, and stores these embeddings in the Redis database.

Run the following command to start the data ingestion:

npm run ingest

It processes text data stored in src/cv.ts file. You may modify this file to ingest your own text data.

Interactive Mode

Run the following command to start the application in interactive mode:

npm start

Once started, the application will prompt you to ask a question. It will then process the input text and respond with the corresponding answer.

Contributing

Contributions are welcome. Please submit a pull request or create an issue to propose changes or additions.

License

This project is licensed under the MIT license.