poc-chatgpt

This repository is a proof-of-concept project that lets you poke at the OpenAI API directly. It gives you ready-to-run commands to see what AI models OpenAI has available, fetch details about specific ones, and test out basic chat functionality — all from your terminal using simple shell scripts.

At its core, the project is a collection of shell commands that talk to OpenAI's servers. You provide your own API key (a password that tells OpenAI you're authorized), and the scripts fetch information about every model or engine available to your account. The results get saved as JSON files, and there are helper scripts that filter those down into plain text lists — so instead of wading through raw data, you get a clean list of model names sorted by who owns them. There's also a JavaScript file you can run for a basic chat interaction.

The project also documents something interesting: leaked model names. The author found references to models like text-chat-davinci-002-20230126 that appeared to be internal or unreleased OpenAI models. The README links to running notes about how these models were discovered and what happened after. Beyond that, it serves as a reference guide, pointing readers to OpenAI's official documentation for everything from generating text and images to creating embeddings and moderating content.

This would appeal to someone exploring the OpenAI ecosystem hands-on — a founder scoping out what models exist, a developer wanting a quick reference for API calls, or anyone curious about what was happening behind the scenes at OpenAI in early 2023. It's more of a personal playground than a polished tool: the instructions assume you already have Node.js and command-line tools installed, and there's no friendly interface beyond the terminal.

The tradeoff is simplicity over polish. You get raw, transparent commands you can read and modify yourself, but you're on your own if something breaks.