Run open-source AI in serverless functions, up to 67x faster. Compatible with Tensorflow.js.
EnergeticAI is a distribution of TensorFlow.js optimized for serverless functions:
- small module size (~3 MB vs. 146 MB - 513 MB for stock TensorFlow.js)
- fast cold-start inference (~50 ms vs. 2000+ ms for stock TensorFlow.js)
- incredible ease-of-use (pre-trained models for common cases)
It's intended to be a condensed replacement for TensorFlow.js, so you can use it with much existing code and models.
EnergeticAI is available on NPM.
Install the core package, along with any model libraries you want to use (e.g. embeddings):
npm install @energetic-ai/core @energetic-ai/embeddings @energetic-ai/model-embeddings-enYou can see a full list of packages below.
Here's an example of using the embeddings model:
import { initModel, distance } from "@energetic-ai/embeddings";
import { modelSource } from "@energetic-ai/model-embeddings-en";
(async () => {
const model = await initModel(modelSource);
const embeddings = await model.embed(["hello", "world"]);
console.log(distance(embeddings[0], embeddings[1])));
})();This repository contains the following packages:
| Package | Description |
|---|---|
| @energetic-ai/core | The core EnergeticAI library, comparable to @tensorflow\tfjs-core. |
| @energetic-ai/classifiers | Classifiers model library. |
| @energetic-ai/embeddings | Embeddings model library. |
| @energetic-ai/model-embeddings-en | English embeddings model weights. |
See the readmes in the packages directory for more information.
See the examples directory for examples.
This repository uses Lerna to manage packages, and Vitest to run tests.
Apache 2.0, except for dependencies.
This project is derived from TensorFlow.js, the Universal Sentence Encoder model library and Universal Sentence Encoder Lite model weights, which are also Apache 2.0 licensed.