Get Started

The Norlax Typescript SDK allows you to easily integrate Norlax in your Typescript based project.

  npm install @Norlax/sdk

In order to use the Norlax SDK, you need to have a launchpad. If you haven't created a launchpad yet, you can read how to create one here.

Basic Usage

import { fetchNorlax } from '@Norlax/sdk'

const response = await fetchNorlax({
  launchpadSlugId: process.env.LAUNCHPAD_ID,
  message: "What is a LLM?",
  cacheTtlSeconds: 'forever',
})

const message = data.message.content // "A LLM is a ..."

This will start a new Norlax chat with the message "What is a LLM?" and cache the response forever. It returns not only the message the LLM sent but also lots of operational data like the id and the url of the chat that it created, or how many TeamTokens it used.

Fetch with in a specific format

If you want to enforce the LLM to respond in a specific format, you can use the schema option.

import { fetchNorlax } from '@Norlax/sdk'
import { z } from 'zod'

const response = await fetchNorlax({
  launchpadSlugId: process.env.LAUNCHPAD_ID,
  message: "Give me a list of the top 5 LLMs",
  cacheTtlSeconds: 'forever',
  schema: z.object({
    llms: z.array(z.object({
      name: z.string(),
      creator: z.string(),
    })),
  })
})

const data = response.message.data // { llms: [{ name: 'LLM1', creator: 'Creator1' }, ...] }

This will do the same as the previous example but enforce the LLM to respond in your specified format. Note that we parse the response with zod and if the response does not match the schema, it will throw an error.

Generating media

Norlax can not only generate text or data but also generate, manipulate or download media. If you instruct the LLM to generate media and it generates it successfully, your response will have mediaAttachments which will contain a url to the media. Note that the launchpad needs to have media generation functions activated in order to be able to generate or manipulate media for you.

import { fetchNorlax } from '@Norlax/sdk'

const imageResponse = await fetchNorlax({
  launchpadSlugId: process.env.IMAGE_GENERATOR_LAUNCHPAD_ID,
  message: "Generate an Image of a cat flying on a donut in space",
  cacheTtlSeconds: 'forever',
})

const imageUrl = imageResponse.mediaAttachments?.[0]?.url // "https://d39ob9hwkmfin1.cloudfront.net/af037574-207c-41df-b1a4-f9a82cfae70a.png?Expires=4070908800&Key-Pair-Id=K1J9Z1M14GSH74&Signature=lsHZkE2GvAnxLYKh-gXjqOFEd--HG~dGvSDkWAcG503RLCt4WonPBeMbX0K21kJYVtSu5n0xtY~KJWzuOuhTksNQ0-T7rpZY7hkzq6tkA3sYB5A5XUyRt-7EbLJCWkiBRYSfxr44Rvdp3kEkOtItNLnFlhDKi8Mp4iihF85Bko8fUHJYEatpB~3hGANLQuvHpKXgo88WkRMHVJLTOnly-CGNjkmd3QcXEQUMeCuMljQh4wLNKKLZ5Jz3YnV2YILtAeIia6-K2Rd89uHkNCwQOsS7oJFgB6eyYAaQqCm1hq1wXtnazjL9lFZuTn3YLCeEhl6gTuVSSOz8bVQTbd6zYQ__"

const audioResponse = await fetchNorlax({
  launchpadSlugId: process.env.AUDIO_GENERATOR_LAUNCHPAD_ID,
  message: "Generate a short poem about a cat flying on a donut in space and read it out loud",
  cacheTtlSeconds: 'forever',
})

const audioUrl = audioResponse.mediaAttachments?.[0]?.url // "https://d39ob9hwkmfin1.cloudfront.net/c1005d12-7027-4d4a-8075-43cb3f07ded0.mpeg?Expires=4070908800&Key-Pair-Id=K1J9Z1M14GSH74&Signature=e9-HUUMLxzo7rKRs71UBzamjmuhvdEuSk-Z8TU4gBmn9FH50PQsee9oTelYluBo6JodFMm07XpLeyOZGPDp2Siu5ykxiPuAJiGxv3uk7flrETlq1B~A92qytgF-UsOAKOB-mPqD6yi-eiBHiMWprFliUhZJQyBvQ6feNp6AqysD0FWFqkT3qTgl765M3pfoqBEPr~Kf9ZWjf4J1SlOquWQCz7~ihWN8nJ2GYzTxtsrHesfU8P001128KDEyZLU94ZHAeqeaFQqblvxmIcFrpnY78NrbgI7IFqE9wZd6R0DG~NWnn6Y3MwYVbdCM3YOLrpmYqZODd0~HTyIXah7kObw__"

Convenience Functions

With the Norlax SDK, you can fetch text, data, or media from the LLMs. The SDK provides convenience functions for each of these, which are just a wrapper around the base fetchNorlax function that just return the message, data, or media from the response.

import { fetchNorlaxText, fetchNorlaxData, fetchNorlaxMedia } from '@Norlax/sdk'

const message = await fetchNorlaxText({
  launchpadSlugId: process.env.LAUNCHPAD_ID,
  message: "What is a LLM?",
  cacheTtlSeconds: 'forever',
})

console.log(message) // "A LLM is a ..."

const data = await fetchNorlaxData({
  launchpadSlugId: process.env.LAUNCHPAD_ID,
  message: "Give me a list of the top 5 LLMs",
  cacheTtlSeconds: 'forever',
  schema: z.object({
    llms: z.array(z.object({
      name: z.string(),
      creator: z.string(),
    })),
  })
})

console.log(data) // { llms: [{ name: 'LLM1', creator: 'Creator1' }, ...] }

const media = await fetchNorlaxMedia({
  launchpadSlugId: process.env.IMAGE_GENERATOR_LAUNCHPAD_ID,
  message: "Generate an Image of a cat flying on a donut in space",
  cacheTtlSeconds: 'forever',
})

console.log(media) // "https://d39ob9hwkmfin1.cloudfront.net/af037574-207c-41df-b1a4-f9a82cfae70a.png?Expires=4070908800&Key-Pair-Id=K1J9Z1M14GSH74&Signature=lsHZkE2GvAnxLYKh-gXjqOFEd--HG~dGvSDkWAcG503RLCt4WonPBeMbX0K21kJYVtSu5n0xtY~KJWzuOuhTksNQ0-T7rpZY7hkzq6tkA3sYB5A5XUyRt-7EbLJCWkiBRYSfxr44Rvdp3kEkOtItNLnFlhDKi8Mp4iihF85Bko8fUHJYEatpB~3hGANLQuvHpKXgo88WkRMHVJLTOnly-CGNjkmd3QcXEQUMeCuMljQh4wLNKKLZ5Jz3YnV2YILtAeIia6-K2Rd89uHkNCwQOsS7oJFgB6eyYAaQqCm1hq1wXtnazjL9lFZuTn3YLCeEhl6gTuVSSOz8bVQTbd6zYQ__"