FreeCAD macro integrating Ollama AI LLM https://spacecruft.org/deepcrayon/FreeCADLLM

addons freecad gpt gpt4 workbench

Python 100%

Find a file

Jeff Moe 8266ef491d Clarify fork/original version		2025-03-31 12:42:57 -06:00
img	Add screenshot	2025-03-31 12:41:36 -06:00
.gitignore	Ignore files	2025-03-30 18:08:38 -06:00
__init__.py	Show fewer dialogs...	2025-03-31 12:28:42 -06:00
FreeCADLLM.FCMacro	Rename macro script	2025-03-31 10:44:08 -06:00
LICENSE	Add me to MIT license	2025-03-31 10:29:03 -06:00
llm.py	s/gpt/llm	2025-03-31 10:35:14 -06:00
llm_integration.py	Default to local Ollama install	2025-03-31 11:42:26 -06:00
llmicon.png	Rename icon	2025-03-31 10:35:30 -06:00
logo.svg	Add files via upload	2023-05-02 20:24:47 -07:00
package.xml	Add branch to repo URL	2025-03-31 11:11:18 -06:00
README.md	Clarify fork/original version	2025-03-31 12:42:57 -06:00

README.md

FreeCADLLM

FreeCAD macro to use a local LLM with Ollama.

https://spacecruft.org/deepcrayon/FreeCADLLM

Screenshot

Install

Install thusly:

mkdir -p ~/.local/share/FreeCAD/AdditionalPythonPackages/py311
pip install ollama --target ~/.local/share/FreeCAD/AdditionalPythonPackages/py311
mkdir -p ~/.local/share/FreeCAD/Mod
cd ~/.local/share/FreeCAD/Mod
git clone https://spacecruft.org/deepcrayon/FreeCADLLM

Ollama

You need to have Ollama installed. It can be local or remote. It defaults to a local install. To use a remote Ollama server, edit this file: ~/.local/share/FreeCAD/Mod/FreeCADLLM/llm_integration.py and change host="http://127.0.0.1:11434" to your remote server.

LLM Models

The default model is qwen2.5-coder:32b-instruct-q8_0. This can be changed to any Ollama model. Edit ~/.local/share/FreeCAD/Mod/FreeCADLLM/llm_integration.py and change this to the desired model: model="qwen2.5-coder:32b-instruct-q8_0"

Usage

In FreeCAD, thusly:

Create a new file.
Go to Macro --> Macros.
In User macros location: select the path to the FreeCADLLM directory, such as: /home/debian/.local/share/FreeCAD/Mod/FreeCADLLM.
Select FreeCADLLM.Macro.
Click Execute.
Under Describe your part, enter what you want, for example: 30mm x 50mm x 15mm box.
Click Execute.

Bugs

For some reason, when the macro runs the first time, it may create multiple windows. Hit Escape to close the extra windows.
Sometimes it generates legitimate code, but it doesn't render in FreeCAD. If this happens, just re-run with the same description.
Sometimes it generates legitimate code, but it doesn't render. Then when another object is described, it renders the last two requests...
Need to select crufty path, since it isn't in Addons.
Many others....

Status

Alpha.

Sometimes works, but under early development.

Fork

This is a fork of Robb Sharma's (revhappy) GPT4FreeCAD. GPT4FreeCAD uses OpenAI's proprietary service. This fork uses Ollama and can be run locally.

https://github.com/revhappy/GPT4FreeCAD

License

MIT.