@themaximalist/ai.js

<!DOCTYPE html> <html xmlns="http://www.w3.org/1999/xhtml" lang="" xml:lang=""> <head> <meta charset="utf-8" /> <meta name="generator" content="pandoc" /> <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes" /> <meta name="description" content="Simple interfaces to LLMs, image generators, embeddings, vector databases and more" /> <title>AI.js — AI Toolkit for Node.js</title> <style> code{white-space: pre-wrap;} span.smallcaps{font-variant: small-caps;} div.columns{display: flex; gap: min(4vw, 1.5em);} div.column{flex: auto; overflow-x: auto;} div.hanging-indent{margin-left: 1.5em; text-indent: -1.5em;} /* The extra [class] is a hack that increases specificity enough to override a similar rule in reveal.js */ ul.task-list[class]{list-style: none;} ul.task-list li input[type="checkbox"] { font-size: inherit; width: 0.8em; margin: 0 0.8em 0.2em -1.6em; vertical-align: middle; } .display.math{display: block; text-align: center; margin: 0.5rem auto;} /* CSS for syntax highlighting */ pre > code.sourceCode { white-space: pre; position: relative; } pre > code.sourceCode > span { line-height: 1.25; } pre > code.sourceCode > span:empty { height: 1.2em; } .sourceCode { overflow: visible; } code.sourceCode > span { color: inherit; text-decoration: inherit; } div.sourceCode { margin: 1em 0; } pre.sourceCode { margin: 0; } @media screen { div.sourceCode { overflow: auto; } } @media print { pre > code.sourceCode { white-space: pre-wrap; } pre > code.sourceCode > span { text-indent: -5em; padding-left: 5em; } } pre.numberSource code { counter-reset: source-line 0; } pre.numberSource code > span { position: relative; left: -4em; counter-increment: source-line; } pre.numberSource code > span > a:first-child::before { content: counter(source-line); position: relative; left: -1em; text-align: right; vertical-align: baseline; border: none; display: inline-block; -webkit-touch-callout: none; -webkit-user-select: none; -khtml-user-select: none; -moz-user-select: none; -ms-user-select: none; user-select: none; padding: 0 4px; width: 4em; background-color: #2a211c; color: #bdae9d; } pre.numberSource { margin-left: 3em; border-left: 1px solid #bdae9d; padding-left: 4px; } div.sourceCode { color: #bdae9d; background-color: #2a211c; } @media screen { pre > code.sourceCode > span > a:first-child::before { text-decoration: underline; } } code span.al { color: #ffff00; } /* Alert */ code span.an { color: #0066ff; font-weight: bold; font-style: italic; } /* Annotation */ code span.at { } /* Attribute */ code span.bn { color: #44aa43; } /* BaseN */ code span.bu { } /* BuiltIn */ code span.cf { color: #43a8ed; font-weight: bold; } /* ControlFlow */ code span.ch { color: #049b0a; } /* Char */ code span.cn { } /* Constant */ code span.co { color: #888888; font-style: italic; } /* Comment */ code span.do { color: #0066ff; font-style: italic; } /* Documentation */ code span.dt { text-decoration: underline; } /* DataType */ code span.dv { color: #44aa43; } /* DecVal */ code span.er { color: #ffff00; font-weight: bold; } /* Error */ code span.ex { } /* Extension */ code span.fl { color: #44aa43; } /* Float */ code span.fu { color: #ff9358; font-weight: bold; } /* Function */ code span.im { } /* Import */ code span.in { color: #0066ff; font-weight: bold; font-style: italic; } /* Information */ code span.kw { color: #43a8ed; font-weight: bold; } /* Keyword */ code span.op { } /* Operator */ code span.pp { font-weight: bold; } /* Preprocessor */ code span.sc { color: #049b0a; } /* SpecialChar */ code span.ss { color: #049b0a; } /* SpecialString */ code span.st { color: #049b0a; } /* String */ code span.va { } /* Variable */ code span.vs { color: #049b0a; } /* VerbatimString */ code span.wa { color: #ffff00; font-weight: bold; } /* Warning */ </style> <link rel="stylesheet" href="style.css" /> <script defer data-domain="aijs.themaximalist.com" src="https://s.cac.app/js/script.outbound-links.js"></script> <meta property="og:url" content="https://aijs.themaximalist.com/"> <meta property="og:type" content="website"> <meta property="og:title" content="AI.js — AI Toolkit for Node.js"> <meta property="og:description" content="Simple interfaces to LLMs, image generators, embeddings, vector databases and more"> <meta property="og:image" content="https://aijs.themaximalist.com/social.png"> <meta name="twitter:card" content="summary_large_image"> <meta property="twitter:domain" content="aijs.themaximalist.com"> <meta property="twitter:url" content="https://aijs.themaximalist.com/"> <meta name="twitter:title" content="AI.js — AI Toolkit for Node.js"> <meta name="twitter:description" content="Simple interfaces to LLMs, image generators, embeddings, vector databases and more"> <meta name="twitter:image" content="https://aijs.themaximalist.com/social.png"> </head> <body> <a class="fork-me" href="https://github.com/themaximal1st/ai.js"><img decoding="async" width="149" height="149" src="https://github.blog/wp-content/uploads/2008/12/forkme_right_darkblue_121621.png?resize=149%2C149" class="attachment-full size-full" alt="Fork me on GitHub" loading="lazy" data-recalc-dims="1"></a> <div id="content" class="gap-md"> <header id="title-block-header"> <h1 class="title">AI.js — AI Toolkit for Node.js</h1> </header> <nav id="TOC" role="doc-toc"> <ul> <li><a href="#ai.js" id="toc-ai.js">AI.js</a></li> <li><a href="#features" id="toc-features">Features</a></li> <li><a href="#installation" id="toc-installation">Installation</a></li> <li><a href="#llm" id="toc-llm">LLM</a></li> <li><a href="#images" id="toc-images">Images</a> <ul> <li><a href="#imagine-concepts" id="toc-imagine-concepts">Imagine Concepts</a></li> </ul></li> <li><a href="#embeddings" id="toc-embeddings">Embeddings</a></li> <li><a href="#vector-search" id="toc-vector-search">Vector Search</a></li> <li><a href="#api" id="toc-api">API</a></li> <li><a href="#debug" id="toc-debug">Debug</a></li> <li><a href="#projects" id="toc-projects">Projects</a></li> <li><a href="#license" id="toc-license">License</a></li> <li><a href="#author" id="toc-author">Author</a></li> </ul> </nav> <h2 id="ai.js">AI.js</h2> <img src="logo.png" alt="AI.js — AI Toolkit for Node.js" class="logo" style="max-width: 300px;" /> <div class="badges" style="text-align: center; margin-top: 0px;"> <a href="https://github.com/themaximal1st/ai.js"><img alt="GitHub Repo stars" src="https://img.shields.io/github/stars/themaximal1st/ai.js"></a> <a href="https://www.npmjs.com/package/@themaximalist/ai.js"><img alt="NPM Downloads" src="https://img.shields.io/npm/dt/%40themaximalist%2Fai.js"></a> <a href="https://github.com/themaximal1st/ai.js"><img alt="GitHub code size in bytes" src="https://img.shields.io/github/languages/code-size/themaximal1st/ai.js"></a> <a href="https://github.com/themaximal1st/ai.js"><img alt="GitHub License" src="https://img.shields.io/github/license/themaximal1st/ai.js"></a> </div> <code>AI.js</code> is the easiest way to add AI text, images, embeddings and vector search to your Node.js app. <div class="sourceCode" id="cb1"><pre class="sourceCode javascript"><code class="sourceCode javascript"><a href="#cb1-1" aria-hidden="true" tabindex="-1"></a>await AI("the color of the sky is"); // blue <a href="#cb1-2" aria-hidden="true" tabindex="-1"></a> <a href="#cb1-3" aria-hidden="true" tabindex="-1"></a>await AI.Image("a red rose"); // <image buffer: red rose> <a href="#cb1-4" aria-hidden="true" tabindex="-1"></a> <a href="#cb1-5" aria-hidden="true" tabindex="-1"></a>await AI.Image.Concept("a red rose"); // {prompt: a red rose in realist style, watercolor ...", <image buffer>} <a href="#cb1-6" aria-hidden="true" tabindex="-1"></a> <a href="#cb1-7" aria-hidden="true" tabindex="-1"></a>await AI.Embeddings("hello world"); // Array(384) <a href="#cb1-8" aria-hidden="true" tabindex="-1"></a> <a href="#cb1-9" aria-hidden="true" tabindex="-1"></a>const db = new AI.VectorDB(); <a href="#cb1-10" aria-hidden="true" tabindex="-1"></a>await db.add("red"); <a href="#cb1-11" aria-hidden="true" tabindex="-1"></a>await db.add("blue"); <a href="#cb1-12" aria-hidden="true" tabindex="-1"></a>await db.search("redish") // red</code></pre></div> Under the hood <code>AI.js</code> seamlessly integrates easy to use local and remote APIs <ul> <li>Text: <a href="https://llmjs.themaximalist.com">LLM.js</a> use <code>GPT-4</code>, <code>Gemini</code>, <code>Claude</code>, <code>Mistral</code> or <code>LLaMa</code> in same simple interface</li> <li>Image: <a href="https://imaginejs.themaximalist.com/">Imagine.js</a> supports local <code>Stable Diffusion</code> and remote services like <code>Replicate</code> and <code>Stability AI</code></li> <li>Embeddings: <a href="https://embeddingsjs.themaximalist.com/">Embeddings.js</a> create <code>local</code>, <code>OpenAI</code> or <code>Mistral</code> embeddings</li> <li>VectorDB: <a href="https://vectordbjs.themaximalist.com/">VectorDB.js</a> searches similar embeddings in memory</li> <li>Audio: Coming Soon</li> <li>Video: Coming Soon</li> </ul> <h2 id="features">Features</h2> <ul> <li>Easy to use</li> <li>Same simple interface for hundreds of models (OpenAI, Google, Anthropic, Mistral, LLaMa, Replicate, Stability AI, Hugging Face and more)</li> <li>Works locally and offline by default!</li> <li>Offers best-in-class options through remote APIs</li> <li>Advanced LLM features like one-shot prompts, chat history, streaming and JSON schema and more</li> <li>Image generation using the best open and paid models</li> <li>Image concepts to easily combine LLMs with Image generators for impressive results</li> <li>Easy to use text embeddings and in-memory vector search</li> <li>MIT license</li> </ul> <h2 id="installation">Installation</h2> Install <code>AI.js</code> via npm <div class="sourceCode" id="cb2"><pre class="sourceCode bash"><code class="sourceCode bash"><a href="#cb2-1" aria-hidden="true" tabindex="-1"></a>npm install @themaximalist/ai.js</code></pre></div> Enable at least one service by setting its environment <code>API_KEY</code> <div class="sourceCode" id="cb3"><pre class="sourceCode bash"><code class="sourceCode bash"><a href="#cb3-1" aria-hidden="true" tabindex="-1"></a>export OPENAI_API_KEY=sk-... <a href="#cb3-2" aria-hidden="true" tabindex="-1"></a>export ANTHROPIC_API_KEY=sk-ant-... <a href="#cb3-3" aria-hidden="true" tabindex="-1"></a>export GOOGLE_API_KEY=sk-ant-... <a href="#cb3-4" aria-hidden="true" tabindex="-1"></a>export STABILITY_API_KEY=sk-... <a href="#cb3-5" aria-hidden="true" tabindex="-1"></a>export REPLICATE_API_KEY=sk-.... <a href="#cb3-6" aria-hidden="true" tabindex="-1"></a>export MISTRAL_API_KEY=...</code></pre></div> <h2 id="llm">LLM</h2> The default interface is text. <code>AI.js</code> lets you send one-off requests or build up complex message histories with the Large Language Model (LLM). <div class="sourceCode" id="cb4"><pre class="sourceCode javascript"><code class="sourceCode javascript"><a href="#cb4-1" aria-hidden="true" tabindex="-1"></a>const AI = require("@themaximalist/ai.js"); <a href="#cb4-2" aria-hidden="true" tabindex="-1"></a>await AI("what is the codeword?"); // i don't know any codewords <a href="#cb4-3" aria-hidden="true" tabindex="-1"></a> <a href="#cb4-4" aria-hidden="true" tabindex="-1"></a>const ai = new AI("the codeword is blue"); <a href="#cb4-5" aria-hidden="true" tabindex="-1"></a>await ai.chat("what is the codeword?"); // blue</code></pre></div> The default <code>AI.js</code> mode is <code>LLM</code>, and is running on top of <a href="https://llmjs.themaximalist.com">LLM.js</a>. Please see that site for full documentation, that also applies to <code>AI.js</code>. There are many features supported across dozens of popular models, like streaming, JSON support, max_tokens, temperature, seed and more. <h2 id="images">Images</h2> <code>AI.js</code> provides powerful image generation functions through <code>Automatic1111</code>, <code>StabilityAI</code> and <code>Replicate</code>. Make sure you have each service setup as needed, either running locally or a valid environment variable. <div class="sourceCode" id="cb5"><pre class="sourceCode javascript"><code class="sourceCode javascript"><a href="#cb5-1" aria-hidden="true" tabindex="-1"></a>const image = await AI.Image("a red rose"); <a href="#cb5-2" aria-hidden="true" tabindex="-1"></a>fs.writeFileSync("rose.png", image);</code></pre></div> See <a href="https://imaginejs.themaximalist.com/">Imagine.js</a> for all image generator documentation. <h3 id="imagine-concepts">Imagine Concepts</h3> <code>AI.js</code> also provides a concept generator—a way of using LLMs together with image generators. <div class="sourceCode" id="cb6"><pre class="sourceCode javascript"><code class="sourceCode javascript"><a href="#cb6-1" aria-hidden="true" tabindex="-1"></a>const { prompt, buffer } = await AI.Image.Concept("a red rose"); <a href="#cb6-2" aria-hidden="true" tabindex="-1"></a>console.log(prompt); // a red rose in realist style, watercolor ..." <a href="#cb6-3" aria-hidden="true" tabindex="-1"></a>fs.writeFileSync("complex-rose.png", buffer);</code></pre></div> This hits your <code>LLM</code> provider and generates a complex image prompt before sending it off to the image generation service. <h2 id="embeddings">Embeddings</h2> <code>AI.js</code> let’s you easily generate local or remote embeddings using <code>Embeddings.js</code>. <div class="sourceCode" id="cb7"><pre class="sourceCode javascript"><code class="sourceCode javascript"><a href="#cb7-1" aria-hidden="true" tabindex="-1"></a>const embeddings = await AI.Embeddings("hello world"); // embedding array</code></pre></div> It works with local embeddings, OpenAI and Mistral. Embeddings can be used in any vector database like Pinecone, Chroma, PG Vector, etc… See <a href="https://embeddingsjs.themaximalist.com/">Embeddings.js</a> for all options. <h2 id="vector-search">Vector Search</h2> <code>AI.js</code> let’s you quickly find similar text strings using a vector database. It runs in-memory and can use embeddings from <code>AI.js</code>. To find similar strings, add a few to the database, and then search. <div class="sourceCode" id="cb8"><pre class="sourceCode javascript"><code class="sourceCode javascript"><a href="#cb8-1" aria-hidden="true" tabindex="-1"></a>const db = new AI.VectorDB(); <a href="#cb8-2" aria-hidden="true" tabindex="-1"></a> <a href="#cb8-3" aria-hidden="true" tabindex="-1"></a>await db.add("orange"); <a href="#cb8-4" aria-hidden="true" tabindex="-1"></a>await db.add("blue"); <a href="#cb8-5" aria-hidden="true" tabindex="-1"></a> <a href="#cb8-6" aria-hidden="true" tabindex="-1"></a>const result = await db.search("light orange"); <a href="#cb8-7" aria-hidden="true" tabindex="-1"></a>// [ { input: 'orange', distance: 0.3109036684036255 } ]</code></pre></div> See <a href="https://vectordbjs.themaximalist.com/">VectorDB.js</a> for full documentation. <h2 id="api">API</h2> <code>AI.js</code> includes these sub-projects: <ul> <li><a href="https://llmjs.themaximalist.com">LLM.js</a> — Simple interface to dozens of Large Language Models</li> <li><a href="https://imaginejs.themaximalist.com">Imagine.js</a> — Local and remote image generation library</li> <li><a href="https://embeddingsjs.themaximalist.com">Embeddings.js</a> — Simple local or OpenAI text embeddings</li> <li><a href="https://vectordbjs.themaximal1st.com">VectorDB.js</a> — Local text similarity search</li> </ul> Check out each individual project for full API documentation. They can be used individually or together like in <code>AI.js</code>. <h2 id="debug">Debug</h2> <code>AI.js</code> uses the <code>debug</code> npm module across all of it’s sub-projects. The namespace is the lowercase version of the project name. View debug logs by setting the <code>DEBUG</code> environment variable. <div class="sourceCode" id="cb9"><pre class="sourceCode bash"><code class="sourceCode bash"><a href="#cb9-1" aria-hidden="true" tabindex="-1"></a>> DEBUG=llm.js* <a href="#cb9-2" aria-hidden="true" tabindex="-1"></a>> node src/run_ai.js <a href="#cb9-3" aria-hidden="true" tabindex="-1"></a># debug logs</code></pre></div> You can combine multiple logs with a comma. <div class="sourceCode" id="cb10"><pre class="sourceCode bash"><code class="sourceCode bash"><a href="#cb10-1" aria-hidden="true" tabindex="-1"></a>> DEBUG=llm.js*,imagine.js*,embeddings.js*,vectordb.js* <a href="#cb10-2" aria-hidden="true" tabindex="-1"></a>> node src/run_ai.js <a href="#cb10-3" aria-hidden="true" tabindex="-1"></a># debug all logs</code></pre></div> <h2 id="projects">Projects</h2> <code>AI.js</code> is currently used in the following projects: <ul> <li><a href="https://infinityarcade.com">Infinity Arcade</a> — Play any text adventure game</li> </ul> <h2 id="license">License</h2> MIT <h2 id="author">Author</h2> Created by <a href="https://twitter.com/themaximal1st">The Maximalist</a>, see our <a href="https://themaximalist.com/products">open-source projects</a>. <script src="/app.js"></script> </div> </body> </html>