Embeddings

Generate text embeddings.

OpenAI Embeddings

import { embed } from "@threaded/ai";

const vector = await embed("openai/text-embedding-3-small", "hello world");

Returns an array of numbers.

With Dimensions

const vector = await embed(
  "openai/text-embedding-3-small",
  "hello world",
  { dimensions: 256 }
);

Reduces output dimensions for smaller vectors.

Hugging Face Models

const vector = await embed("Xenova/all-MiniLM-L6-v2", "hello world");

Uses @huggingface/transformers for local inference.

Requires system dependencies (ONNX runtime).

Semantic Search Example

import { embed, cosineSimilarity } from "@threaded/ai";

const documents = [
  "the cat sat on the mat",
  "dogs are great pets",
  "javascript is a programming language",
];

const docVectors = await Promise.all(
  documents.map(doc => embed("openai/text-embedding-3-small", doc))
);

const queryVector = await embed("openai/text-embedding-3-small", "pets and animals");

const scores = docVectors
  .map((vec, i) => ({ document: documents[i], score: cosineSimilarity(queryVector, vec) }))
  .sort((a, b) => b.score - a.score);

Finds the most similar documents to a query.

Caching

Hugging Face models are cached after first load.

const vector1 = await embed("Xenova/all-MiniLM-L6-v2", "text 1");
const vector2 = await embed("Xenova/all-MiniLM-L6-v2", "text 2");

The second call reuses the loaded model.