@seekdb/openai

OpenAI embedding function for SeekDB.

OpenAI embedding models are a core tool in natural language processing, efficiently converting text (words, sentences, paragraphs) into high-dimensional vectors (embeddings). These vectors are generated by high-performance pre-trained models (such as text-embedding-3-large), capturing deep semantic information without requiring users to perform any additional training. The key idea is to map raw text into a dense vector space, where semantic similarity between texts can be conveniently measured by computing distances between vectors (e.g., cosine similarity). This technology is widely used in semantic search, intelligent Q&A, text clustering and classification, enhanced recommender systems, and retrieval-augmented generation (RAG) scenarios.

Installation

npm i seekdb @seekdb/openai

Usage

import { OpenAIEmbeddingFunction } from "@seekdb/openai";

const ef = new OpenAIEmbeddingFunction({
  modelName: "text-embedding-3-small",
  // baseURL: "https://api.openai.com/v1",
});

Configuration

• apiKey: API key (optional; can be provided via env var)
• apiKeyEnvVar: API key env var name (default: "OPENAI_API_KEY")
• modelName: model name (default: "text-embedding-3-small")
• dimensions: output dimensions (optional)
• organizationId: organization id (optional; default: process.env.OPENAI_ORG_ID)
• baseURL: custom base URL (optional)