January 31, 2021

GPT-3 Programming 101

Here is a primer on how to get started using OpenAI GPT-3 API programmaticaly. OpenAI provides a Python API client, but other clients can be easy build on other languages like Node.

Basic Python Setup

This summary was extracted from many of Twilio’s great GPT-3 examples.

Create working folder.

$ mkdir my-gpt3-app
$ cd my-gpt3-app

Setup a virtual environment to install some Python packages in a sandbox.

$ python3 -m venv my-gpt3-venv
$ source my-gpt3-venv/bin/activate

Install required packages via pip (the Python package installer):

(my-gpt3-venv) $ pip install openai

Using a lightweight API client

I’ll be using Minimaxir/gpt-3-experiments python client for this example. See below for other client implementation and other languages. I picked this one since it has already build in saving API responses to formatted markdown files.

Install packages:

$ source my-gpt3-venv/bin/activate
(my-gpt3-venv) $ pip3 install httpx pyyaml fire tqdm

Setup configuration in config.yml

Usage (this will generate a file for each specified temperature in the config.yml.):

(my-gpt3-venv) $ python3 openai_api.py "Once upon a time"
(my-gpt3-venv) $ python3 openai_api.py "prompt.txt"

(Bonus) Saving OpenAI Secret key in an environment variable

Install packages:

(my-gpt3-venv) $ pip install python-dotenv

Setup OpenAI API key. Create a .env file in your project’s root directory

OPENAI_KEY= <SECRET-OPENAI-KEY>

Resources

Libraries

See a full list at OpenAI API

Python:

Node:

Articles

January 31, 2021

Comments

Discuss on Twitter ↗



Sign up for the mailing list




Previous:Drug Machine
Next:My First Exocortex