About

A simple wrapper for OpenAI's ChatGPT API completion HTTP endpoint. Usable on any target. An API key from OpenAI will be required, obviously.

Example usage:

var client = new ChatGPTClient(/*API KEY HERE*/, GPT_4);
var result = client.completion("This is a test, please acknowledge.");

switch result {
	case Success(result):{
		trace("success: " + result.content);
	}
	case Error(errorMessage): {
		trace("error: " + errorMessage);
	}
	case Promise(success, error): {
		success.listen((data) -> {
			trace("async success: " + data.content);
		});
		error.listen((data) -> {
			trace("async error: " + data);
		});
	}
}

client.completion() returns an enum value:

enum ChatGPTClientResult {
    Success(result: ChatGPTClientPayload);
    Error(errorMessage: String);
    Promise(success: HxPromise<ChatGPTClientPayload>, error: HxPromise<String>);
}

On concurrent targets client.completion() will block and return ChatGPTClientResult.Success or ChatGPTClientResult.Error. If concurrency is not supported or client.async == true then ChatGPTClientResult.Promise will be returned, in which case data can be retrieved by success.listen().

Other features:

• client.systemMessage can be used to add a message with role: "system" to the request.
• client.maxTokens throttles tokens returned by the request.
• client.enrichRequest can be set to change the request JSON object before it is sent, thus allowing you to add custom params to the request.
• ChatGPTClientPayload.content returns the simple content of the result, but the full payload can be accessed through ChatGPTClientPayload.fullTextPayload or ChatGPTClientPayload.fullJsonPayload.