Conversational App Engine is an open-source engine that allows developers to quickly implement chat based apps using OpenAI's completion API. With Conversational App Engine, developers can focus on their app's use case while leveraging the power of Conversational App Engine to handle the chat management.
Conversational App Engine was initially created as a demonstration of how OpenAI's completion API can be used to build a form designer. However, we quickly realized that the same technology could be applied to a wide range of use cases. Our goal is to provide developers with an easy-to-use engine that can improve their app's user experience with conversational AI.
To use Conversational App Engine, you will need to have Node.js and npm installed on your computer. You will also need an OpenAI API key to access the completion API. Once you have these prerequisites, you can clone this project using Git or download it as a zip file.
Next, set the environment variable OPENAI_API_KEY to your OpenAI API key. This will allow Conversational App Engine to authenticate your requests to the API.
To implement your own app, create a new class in the apps directory that extends the ConversationApp class. You will need to implement several methods, such as getDefaultMessages, getChatNameFromMessage, and getTextMessage, to define your app's behavior. You can also overwrite any of the default messages, such as appName and chatListTitle, to customize the user interface.
Once you have created your app, import it into the index.js file and use it to initialize a new instance of ConversationalAppEngine. Then, run node ./index.js in the project directory and navigate to http://localhost:3000/ to see your app in action.
You need to obtain an “OpenAI” API key from https://platform.openai.com/account/api-keys and set the OPENAI_API_KEY environment variable. This will allow the app to make API calls to OpenAI.
git clone https://github.com/david-m-s/conversational-app-engine.git
cd conversational-app-engine
npm install
If you don't have git installed. You can download the project here
1- Crate a class that extends ConversationApp in the apps directory 2- Implement the following methods:
- getDefaultMessages: an array of instructions that tells GPT about its role and the expected response. Format [{"role": "system", "content": "You are ABC assistant, you provide help in..."}, {"role": "user", "content": "Your responses should be structured in yaml format"}, ...]
- getChatNameFromMessage: The name of the chat, it can be derived from the message
- getTextMessage: the response text that should be shown in the chat dialog
- getAppContent: HTML (usually) based content, generated based on extracted content from the message based on the app business need
3- Overwrite any of the following messages if needed
- appName
- chatListTitle
- newChatLabel
- newChatName
- contentPreviewPlaceholder
- chatStartInstruction
4- Overwrite the temperature property if needed, default is 1, valid values are between 0 and 2 inclusive
1- Run node ./index.js in the project directory
2- Navigate to http:/localhost:3000/
3- Check the available apps in the apps drop down menu beside the current app name.
Currently this project includes the following demo apps:
The Quiz Generator is an application that harnesses the power of GPT technology to effortlessly create a quiz based on any given paragraph of text. Users can easily generate a comprehensive quiz that tests their knowledge and comprehension of the subject matter contained within the text. The generated quiz can then be exported in both HTML and GIFT formats:
The Statistical Data Analyzer is an application that extract Statistical Data from the provided text and visualize it as chart and as a data table. It will describe any trends in the data.
The Quote Image Generator is an application that uses GPT to create wallpapers in SVG format. These wallpapers feature suggested quotes that are relevant to a specified topic. It's important to note that sometimes the generated images may have rendering issues. However, you can simply request to fix them in the conversation. To give you an idea of what these wallpapers look like, here's an example of a few generated SVGs:
The Todo Assistant is an application that uses GPT to manage your tasks as a todo list, where you can tell the app about your new tasks and completed ones and the app will render them in a list:
