GDScript OpenAI (0.3.0)

Interfacing with the OpenAI API's using their API Key

With the gd-openai addon you can quickly create scenes interacting with the API of OpenAI. See the examples below.

Download the ZIP file from GitHub or install through the AssetLib when available.

Usages

• GDScript OpenAI coding

Examples

Chat

Image generation

Browse the saved resources

User Data

The settings file open_ai_user_data.tres and all requests and responses are saved in user://.

The settings file for security reasons as it may not become part of the build.

The requests and responses for browsing back possibilities.

What calls are available?

• /v1/chat/completions to mimic ChatGPT sessions. Note: only model, system, assistent, user are available.
• /v1/models to find available models.
• /v1/images/generations for generating inspiring images.

API Key

The tool needs an API key from your OpenAI account.

Run addons/gd-openai/user_data.tscn to set your API key.

Then run one of the examples to validate your key is correct.

• addons/gd-openai/examples/scenes/models.tscn
• addons/gd-openai/examples/scenes/image_strip.tscn
• addons/gd-openai/examples/scenes/chat_completions.tscn

The key is stored in Project > Open User Data Folder.

Examples

The are some examples available to list the models, start a chat/completions or fiddle with images/generations.

Building a scene

You need to add 3 parts to your scene:

OpenAiApiRequest

As a child in your tree.

@onready var openai_api_request:OpenAiApiRequest = $OpenAiApiRequest

func _ready():
	# Connect openai_api_request signals through the UI or code
	openai_api_request.connect("data_received", _on_open_ai_api_request_data_received)
	openai_api_request.connect("error_response", _on_open_ai_api_request_error_response)

func _on_open_ai_api_request_data_received(data):
	response = data

func _on_open_ai_api_request_error_response(error):
	print(error)

RequestData sub-class

RequestData sub-class like ModelsRequest, ChatCompletionsRequest, ImageGenerationsRequest, etc

@export var request:ChatCompletionsRequest

Then set it's value through the UI.

ResponseData sub-class

ResponseData sub-class like ModelsResponse.

var response:ChatCompletionsResponse = ChatCompletionsResponse.new()

Request/response

Initiate the call ie by a Button click

openai_api_request.do_post(request, response)

and wait for one of the two signals to appear.

Roadmap

• Add all properties to /v1/chat/completions
• /v1/models/{model}
• your wishes / PRs / support

References

• OpenAI API's
• API Key
• API

Some other implementations

• godot-openai
• GPTIntegration
• GodotGPT