AI Explorer (Experimental) Extension for SketchUp

Description

This experimental tool can use various AI (artificial intelligence) services like OpenAI, Google, Anthropic (and others) for a variety of purposes in SketchUp: It can generate SketchUp Ruby code that – when executed – can draw geometry (e.g. with the prompt “Draw a box”). It can also manipulate drawn objects (e.g. with a command like “Color all objects in the selection randomly using five blue tones”). You can also use it to interact with the AI services using chat-like, text-only prompts like “Tell me a joke” or “How does the follow-me tool work in SketchUp”. With current models, you can even chat about what is visible in your SketchUp model.

PLEASE NOTE: This is an experimental extension, which is able to fully automate SketchUp using its Ruby scripting engine (if code execution is enabled). This can even affect your entire file system because SketchUp has unrestricted, user level access to that. Use with caution and at your own risk! When in doubt, do not enable code execution in the extension’s settings (it is disabled by default).

Any AI generated content may contain errors, fabrications, or may not be usable at all. Use any generated content at your own discretion. Therefore, use this tool with caution and experiment with prompts if you run into errors. Sometimes it helps to simply re-word the prompt.

This extension is not affiliated with or endorsed by OpenAI or any of the other services mentioned.

Up to version 2.8, this extension was called “OpenAI Explorer (Experimental)”.

Usage

Before You Start

Before you can use this tool, you must sign up for OpenAI API access and generate an API Key here. This requires you to set up an account with OpenAI and may lead to costs (after a free trial period). You then need to enter the API key that OpenAI provides in this extension’s settings dialog and you are good to go. All of the other settings are pre-configured for OpenAI’s services and can stay as they are for now.

You can find the settings dialog under the Extensions > AI Explorer (Experimental) > AI Explorer Settings menu item.

AI Explorer Dialog

The actual AI prompt can then be found in a different dialog, which opens up when you click the Extensions > AI Explorer (Experimental) > AI Explorer menu item.

This extension features a chat-like interface where you can enter a prompt (e.g. “Draw a line” as shown in the following image) and then the full user prompt (as it was submitted to the AI service) as well as the response are shown (and retained as long as the dialog remains open). This makes it easy to add subsequent queries. You can also use the Settings button to bring up the settings dialog and make changes on the fly (e.g. if you want to switch code execution on or off between requests). Make sure you read the Use Cases section below to understand what can be entered in this dialog.

TIP:
The input field in this dialog should work well with your operating system’s built-in speech-to-text functionality. For example, in Windows, use the WinKey + H key combination to bring up a microphone input.

As of version 2.6, this dialog includes a selection dropdown where you can change use cases quickly on the fly. Changing this dropdown only affects the relevant settings options (e.g. whether code gets executed) and leaves others (e.g. the AI model name) unchanged.

Settings Dialog

This dialog provides several options to customize this extension. Edit those as needed:

System Message: This text is being sent to the AI service’s API as part of the system message portion of a request. This field can basically contain a set of instructions that tell the AI what to do with the actual request (and what not to do). Feel free to experiment with different system messages such as:

• “Generate only valid, self-contained SketchUp Ruby code without any method definitions.”
• “Generate only SketchUp Ruby.”
• “Respond in Shakespearean English.”
• You can leave this empty if you just want to send the user request, e.g. for plain chatbot behavior.

Chat Completion Model: This is preset for a current, well-working OpenAI model (gpt-4o-mini) but you can enter a different model than the default here anytime. A list of valid models can be found in the OpenAI API documentation. I have also discussed some of those in various blog posts.

Max. Tokens: This number controls the length (and thereby potential cost) of the response. It is likely good to keep this number low unless you need it higher. The extension will output the exact number of tokens used for each request, which allows you to adjust this limit as needed.

Temperature: This parameter controls the “creativity” in the answer and therefore is more deterministic at zero and more variable (and possibly less reliable) at higher values.

Execute Code: By default this is set to not execute code, but if you want to automatically execute the generated code, then you can change that here. Two safety features have been implemented when this option is set to yes:

• Any request that contains a destructive word (“delete”, “erase”,…) will generate an additional user prompt before any code is executed.
• A system message is added that aims to prevent code generation for any requests that involve file system access.

Submit Model View with Request / Set Model View Quality: Newer chat completion models like gpt-4o can accept image uploads. If you use such a model, you can turn this on and decide on the image quality level you want to use. This will then upload the SketchUp model view each time you enter a prompt. You can then use this to ask questions like the ones listed below. See this post for more information about this.

• “How can I improve this model?”
• “Critique my design in the context of contemporary architecture.”
• “Is this model 3d printable?”

Submit # of Prompts: If this is set to 1, then each prompt gets answered on its own (i.e. without the AI “remembering” previous prompts). You may want to increase this number if you want to ask follow-up questions. Keep in mind, however, that a higher number will result in more tokens being used (i.e. a higher cost).

AI Service API Endpoint: This is preset to OpenAI’s endpoint at https://api.openai.com/v1/chat/completions. There is no need to change this unless you are switching AI service providers (as mentioned above).

---

Use Cases

As of version 2.6, you can use the main dialog drop-down selector to change these.

Case 1: Manipulate SketchUp (i.e. Create or Modify Geometry)

Set the System Message to “Use SketchUp Ruby“ (or similar – feel free to experiment) and the Execute Code option to Yes.

Use this to draw or manipulate something based on your prompt and what you have selected in SketchUp. This lets you control SketchUp with natural language (e.g. “Draw a box”). Note that this works better with simple and specific commands than with complex concepts (that would include multiple modeling steps, such as “Draw a house”). Rather, break up complex modeling tasks into simple steps. Also, be very careful with requests that delete something.

To give you some examples, following is a list of prompts that I have used successfully:

• “Draw a box” (these often get drawn downward from the origin because the positive pushpull operation uses a face’s white face)
• “Color the items in the selection randomly in five different blue tones” (selecting something and then asking the AI to modify those seems to work very well)
• “Scale each item in the selection randomly in the z direction” (similar behavior)
• “Replace the selected line with a construction line” (works, but only on a single line)
• “Select all vertical faces in the model” (you can even have the AI select things based on a property like this)
• “Clear the selection. Then select all the components whose name includes ‘sphere'” (works best if you clear the selection first)
• “Create a new material that is aquamarine” (color materials work well but textured materials don’t as they need access to image files)
• “Add a new layer called ‘Walls'” (because the API uses “layer” instead of “tag” you will need to use the word layer here)
• “Add a new lyaer called ‘Walls'” (it understands typos correctly!)
• “Add a new scene tab with a single line of code” (adding “with a single line of code” often prevents the AI from trying to address all the options with the response, cutting down on tokens)
• “Draw a grid of 100 12 inch boxes, spaced at 24 inches” (correctly draws a 10 x 10 grid of boxes)
• “List the center points of all the objects in the selection” (this will print the coordinates to the Ruby Console using the puts command. Make sure you have the console open before you run this command)
• “List all the unique vertices in the selection” (similar)
• “Randomly erase 50% of the selected objects” (this works, but as mentioned before: use with caution)
• “Open the SketchUp website” (opens the website in a browser, same for the 3D Warehouse)
• “Zoom extents” (you can even control the view)
• “Zeichne 10 x 10 50cm Quadrate auf den Boden mit 50cm Abstand” (you can use languages other than English and various unit systems)
• “Add a menu item that draws a parametric pyramid. Ask for the base and the height” (you can even ask for this to install as an extension)
• “Color the selected faces in 10 shades between blue and yellow showing their height (considering the selection’s minimum and maximum values)” (you can use this for analysis)

Case 2: Generate Ruby Code for Editing (i.e. as a Coding “Copilot”)

Set the System Message to “Use SketchUp Ruby“ (or similar – feel free to experiment) and the Execute Code option to No.

Use this to generate Ruby code that you can then copy and paste into your favorite code editor (e.g. my Ruby Code Editor) where you can further edit and execute it. You can use this to create and look up code snippets.

Case 3: Chat / Search and Q&A

Clear the System Message and set the Execute Code option to No.

Use this to ask the AI anything that does not necessarily generate code (e.g. “Tell me a joke”). You can this way use the AI service’s regular chatbot behavior, e.g. to explain how something works in SketchUp. If you like, you can use the system message to give instructions like: “Be a cranky assistant”.

Keep in mind, though, that AI models have a knowledge cutoff date and are not by themselves always useful as a “search replacement”. At this point, search is also not implemented in this tool.

Case 4: Chat About the Model (incl. a Screenshot)

Clear the System Message, set the Execute Code option to No, and set Submit Model View to Yes. Also make sure you are using a compatible model like gpt-4o (or you will get an error).

This will send the current model view to the AI service with your prompt. Use it to ask questions about what is visible in the model, like the example shown here.

A Word About Cost

This section is specific to OpenAI. Check with other providers for their respective costs.

There is no cost to getting and using this extension. However, in order to be able to use OpenAI’s services, you need to sign up with the OpenAI Platform here. This is also free and you get a token credit so that you can try everything out for a while. After that period ends, you will need to add funds to your OpenAI account, however.

Important:
A ChatGPT Plus subscription does not include API (Platform) access. You will need to make sure you have funds under the Platform as well. You can check that here.

In my experience, simply adding $5 to the OpenAI Platform account is plenty for a lot of use, especially if you work with the cheaper gpt-4o-mini model. I have had days where I barely used a cent’s worth. When you compare this with a $20 monthly ChatGPT Plus subscription, you will see that this is also a more cost efficient way to use their services. This is because you can test out many different models, have more control over how you use those (i.e. with the system message), and – best of all – pay by use (instead of a flat fee).

See below for current costs:

Model	Input Cost	Output Cost
gpt-4o (-08-06)	$2.50 / 1M tokens	$10.00 / 1M tokens
gpt-4o image (3000 x 2000, low-res)	$0.000213	N/A
gpt-4o image (3000 x 2000, high-res)	$0.002763	N/A
gpt-4o-mini	$0.150 / 1M tokens	$0.600 / 1M tokens
gpt-3.5-turbo	$0.50 / 1M tokens	$1.50 / 1M tokens

Please Note: I am providing this cost information as a courtesy. Of course, actual costs and your relationship with OpenAI is between you (the user) and them.

Screenshots and Videos

Extension Terms and Conditions

Downloads and Installation

You can now find this extension in SketchUcation’s PluginStore and at the download links below. Due to Trimble’s AI-specific rules, this extension cannot be installed from the Extension Warehouse.

Alternative Downloads

Linked below are the installation files (RBZ format) for the current and some previous versions of this extension. You can also find this extension on GitHub.

IMPORTANT: After updating, you may need to reset the extension’s settings (under the extension’s menu items).

• as_openaiexplorer_3-0.rbz- 3.0
• as_openaiexplorer_2-7.rbz- 2.7
• as_openaiexplorer_2-6.rbz- 2.6
• as_openaiexplorer-2_5.rbz- 2.5
• as_openaiexplorer_2-4.rbz- 2.4
• as_openaiexplorer_2-3.rbz- 2.3
• as_openaiexplorer_2-2.rbz- 2.2
• as_openaiexplorer_2-1.rbz- 2.1
• as_openaiexplorer_2-0.rbz- 2.0
• as_openaiexplorer_1-0-3.rbz- 1.0.3
• as_openaiexplorer_1-0-2.rbz- 1.0.2
• as_openaiexplorer_1-0-1.rbz- 1.0.1
• as_openaiexplorer_1-0.rbz- 1.0

Revision History

• Version 3.0 (3/13/2025):
  – Renamed to AI Explorer (Experimental)
  – Added option to use other OpenAI-compatible services (using the OpenAI API standards), e.g. Google or Anthropic via their API endpoints.
  – Minor cleanups
• Version 2.7 (11/4/2024):
  – Added feature to capture and show Ruby output (e.g. for lists)
  – Adjusted TOU and added TOU link to dialog
  – Adjusted defaults (extension starts in chat mode)
  – Minor fixes
• Version 2.6 (11/2/2024):
  – Added use case selector to main dialog (to speed things up)
  – Fix for initial API Key link error
  – Now executes code under toplevel binding in case the code includes methods
  – Fix for text rendering in output
• Version 2.5 (10/1/2024):
  – Changed default model to gpt-4o-mini
  – Fix for HTML rendering of output (b,i,h2+,code)
  – Added menu item for usage/cost check
  – Made thumbnail appearance stretchable
  – Fixed ‘no code available’ error
  – Code cleanup
• Version 2.4 (7/1/2024):
  – Now renders bold text and code visually in responses
  – Better response handling
  – Dialog now always opens centered
  – Dialog now includes thumbnail if it was submitted with the prompt.
• Version 2.3 (5/18/2024):
  – Updated some defaults
  – Updated button design
  – Added AI “memory”. User can now select how many messages get sent to OpenAI.
  – Added screenshot image upload for vision-capable models (e.g. gpt-4o)
  – Added resolution control for uploaded images
• Version 2.2 (12/9/2023):
  – Implemented correct system message handling as per OpenAI API
  – Added system message to prevent file access when code is to be executed
  – Updated default system message
  – Added menu item to reset extension settings (for troubleshooting)
• Version 2.1 (9/23/2023):
  – Better error handling
  – Fixed gpt-4 markdown extraction issue
  – Added a timer
  – Set a 30 seconds read timeout for hung requests
• Version 2.0 (7/23/2023):
  – Created dialog-based input for more chat-like experience
  – Moved error reporting into dialog to reduce pop-ups
• Version 1.0.3 (7/19/2023):
  – Added license/disclaimer file
  – Added better license display on first use
  – Added link to OpenAI TOU to menu
  – Reverted the prompt prefix to a text box (user can now enter anything)
  – Disabled code execution as the default (can be turned on in settings)
  – Changed from OpenAI completion model to chat model (OpenAI changes)
• Version 1.0.2 (3/28/2023):
  – Fixed error handling bug
  – Added a double-check for delete requests
• Version 1.0.1 (3/24/2023):
  – Added cleanup for the returned code/text
  – Added error handling for API and a one-time warning for the extension
• Version 1.0 (3/18/2023):
  – Initial release

Troubleshooting

You can reset this extension by pasting the following code into SketchUp’s Ruby Console (which can be found under the Extensions menu) and executing it (hit Enter):

Sketchup.write_default( "as_openaiexplorer" , "openai_warning" , nil )
Sketchup.write_default( "as_openaiexplorer" , "openai_explorer_settings" , nil )
Sketchup.write_default( "as_openaiexplorer" , "openai_explorer" , nil )
Sketchup.write_default( "as_openaiexplorer" , "disclaimer_acknowledged" , nil )

References

• OpenAI chat completion reference
• OpenAI chat completion API reference
• My initial post about this
• Adding AI Vision to SketchUp with OpenAI’s gpt-4o Model

Show Off Your Work

Did you create something amazing with this extension? I would love to see it! You can either upload an image in the comment section below (public) or privately send it to me.