Dalai

Dead simple way to run LLaMA on your computer.

Github Twitter

---

1. Powered by llama.cpp and llama-dl CDN
2. Web app included
3. Super simple JavaScript API

---

Quickstart

Install the 7B model (default) and start a web UI:

npx dalai install
npx dalai serve

Then go to http://localhost:3000

Above two commands do the following:

1. First installs the 7B module (default)
2. Then starts a web/API server at port 3000

---

Install

Basic install (7B model only)

npx dalai install

Install all models

npx dalai install 7B 13B 30B 65B

The install command :

1. Creates a folder named dalai under your home directory (~)
2. Installs and builds the llama.cpp project under ~/dalai
3. Downloads all the requested models from the llama-dl CDN to ~/dalai/models
4. Runs some tasks to convert the LLaMA models so they can be used

---

API

Dalai is also an NPM package:

1. programmatically install
2. locally make requests to the model
3. run a dalai server (powered by socket.io)
4. programmatically make requests to a remote dalai server (via socket.io)

Dalai is an NPM package. You can install it using:

npm install dalai

---

1. constructor()

Syntax

const dalai = new Dalai(url)

• url: (optional)
  • if unspecified, it uses the node.js API to directly run dalai
  • if specified (for example ws://localhost:3000) it looks for a socket.io endpoint at the URL and connects to it.

Examples

Initializing a client that connects to a local model (no network):

const dalai = new Dalai()

Initializing a client that connects to a remote dalai server (a dalai server must be running at the URL):

const dalai = new Dalai("ws://localhost:3000")

---

2. request()

Syntax

dalai.request(req, callback)

• req: a request object. made up of the following attributes:
  • prompt: (required) The prompt string
  • model: (required) The model name to query ("7B", "13B", etc.)
  • threads: The number of threads to use (The default is 8 if unspecified)
  • n_predict: The number of tokens to return (The default is 128 if unspecified)
  • seed: The seed. The default is -1 (none)
  • top_k
  • top_p
  • temp: temperature
  • batch_size: batch size
• callback: the streaming callback function that gets called every time the client gets any token response back from the model

Examples

1. Node.js

Using node.js, you just need to initialize a Dalai object with new Dalai() and then use it.

const Dalai = require('dalai')
new Dalai().request({
  model: "7B",
  prompt: "The following is a conversation between a boy and a girl:",
}, (token) => {
  process.stdout.write(token)
})

2. Non node.js (socket.io)

To make use of this in a browser or any other language, you can use thie socket.io API.

Step 1. start a server

First you need to run a Dalai socket server:

// server.js
const Dalai = require('dalai')
new Dalai().serve(3000)     // port 3000

Step 2. connect to the server

Then once the server is running, simply make requests to it by passing the ws://localhost:3000 socket url when initializing the Dalai object:

const Dalai = require("dalai")
new Dalai("ws://localhost:3000").request({
  model: "7B",
  prompt: "The following is a conversation between a boy and a girl:",
}, (token) => {
  console.log("token", token)
})

---

3. serve()

Syntax

Starts a socket.io server at port

dalai.serve(port)

Examples

const Dalai = require("dalai")
new Dalai().serve(3000)

---

4. http()

Syntax

connect with an existing http instance (The http npm package)

dalai.http(http)

• http: The http object

Examples

This is useful when you're trying to plug dalai into an existing node.js web app

const app = require('express')();
const http = require('http').Server(app);
dalai.http(http)
http.listen(3000, () => {
  console.log("server started")
})