Skip to main content

Deploy a plugin in Autocode

Autocode is a platform that allows you to deploy your plugins from a web-based IDE.

Build a NASA image search plugin in Autocode

Benefits of using Autocode

  • Pearl - Programming AI assistant integrated so you can catch and fix errors as you code.

Deploying a plugin in Autocode

Ok, let's get started!

Step 1: Create a New API

Start by navigating to your Autocode dashboard and clicking on 'New API'. This will prompt you to name your plugin. Once named, you'll be presented with various files and folders, including functions, readme, package.json, payloads.json, stdlib.json, and env.json. The functions folder will contain the endpoints for your ChatGPT Plugin.

Step 2: Create the Plugin Manifest

Before proceeding to create the endpoints for your plugin, it's crucial to first create a manifest. The manifest will define the plugin's metadata and the plugin's expected behavior.

To do this, create a www folder at the root of your Autocode project. Think of this www folder as equivalent to a public folder in a traditional web project. Inside this www folder, create another folder named .well-known. This is where the ai-plugin.json manifest file will be placed. Use the ai-plugin.json file to define your plugin's metadata.

Step 3: Define the OpenAPI Specification

Next, return to the www folder and create an openapi.yaml or openapi.json file. This file will outline the OpenAPI specification for your ChatGPT Plugin. The specification will define the routes of your API and the responses it should return.

Step 4: Develop the API Endpoints

Finally, head back to the functions folder and start creating the necessary endpoints for your plugin as defined in your OpenAPI specification. In Autocode, each JavaScript file in the functions folder represents a different endpoint that your plugin can interact with.