gpts

gpts = gpt-3 + typescript

typescript wrapper for gpt-3 api

example

import { GpTs } from 'gpts';
const brain = new GpTs(OPENAI_APIKEY); // dont publish your api key!

const thoughts = brain.completion({
	engineId: 'ada',
	prompt: 'whats for lunch?',
});

console.log(thoughts.choices[0].text); // "maybe a banana?"

also see demo/index.ts (need to insert your api key to run)

install

npm

npm i gpts

github

package.json

	"dependencies": {
		"gpts": "thencc/gpts",
		...
	},

features

• engines
  • list ✅
  • retreive ✅
• completions
  • create ✅
  • stream ⚠️ (TODO)
• searches
  • create ✅
• classifications
  • create ✅
• answers
  • create ✅
• files
  • list ✅
  • upload ✅ (server-side only)
  • retrieve ✅
  • delete ✅
• embeddings
  • create ✅
• fine-tunes
  • prepare dataset ⚠️ (TODO)
  • upload dataset ⚠️ (TODO)
  • check upload results ⚠️ (TODO)

notes

general

• uses openai's REST api
• for the /classifications and /answers endpoints, openai seems to switch the syntax from engineId -> model so if you specify both in the options argument, options.model takes precedence

client-side use

• this library works client + server side using axios for http requests
• DO NOT share your api key in public client-side frontend code
• one way to hide your openai api key for client-side use is by hosting an api wrapper endpoint that enforces your own authentication, then updating the origin this library looks to use like the below.

api wrapper example:

import { GpTs } from 'gpts';
/*
    gives your api wrapper authorization like this:
        headers: {
            'Authorization': 'Bearer ASuperSecretPassword'
        }

    the constructor takes 2 arguments
        1. the authorization bearer value
        2. the api origin
*/
const brain = new GpTs('ASuperSecretPassword', 'https://company.api-wrappers.io/gpt3');

TODO

• support completion streaming SSE (https://beta.openai.com/docs/api-reference/completions/create-via-get)
• support fine-tuning management (https://beta.openai.com/docs/guides/fine-tuning/preparing-your-dataset)