Welcome to the Onshape forum! Ask questions and join in the discussions about everything Onshape.

First time visiting? Here are some places to start:
  1. Looking for a certain topic? Check out the categories filter or use Search (upper right).
  2. Need support? Ask a question to our Community Support category.
  3. Please submit support tickets for bugs but you can request improvements in the Product Feedback category.
  4. Be respectful, on topic and if you see a problem, Flag it.

If you would like to contact our Community Manager personally, feel free to send a private message or an email.

How's AI for feature scripts working for you?

Kevin_Pasko
Kevin_Pasko Posts: 2 PRO
in Community Support

I am trying to use Gemini for feature script creation and it has not been awesome. I have tried up to 5 iterations where I feed OnShape errors back into Gemini to fix, and not been successful. Someone told me he had to run 40 iterations! Is this other people's experience? How many debug iterations do you run?

Answers

  • Matt_Shields
    Matt_Shields Posts: 984 PRO

    I've been using Claude and it's awesome.

  • EvanReese
    EvanReese Posts: 3,055 PRO

    @Kevin_Pasko it's been pretty good. Are you using a Gemini chat? I tried that when I was just getting into it and kept having issues where it would essentially try to re-write the code completely with each reply instead of editing it, which kept reintroducing things that had been fixed. Using one that's dedicated for code should work a lot better, like Claude Code or ChatGPT Codex. It's even better now with the Onshape MCP letting it connect directly with Onshape to update the code, and test things out.

    Evan Reese
    The Onsherpa | Reach peak Onshape productivity
    www.theonsherpa.com
  • S1mon
    S1mon Posts: 4,485 PRO

    Also, the higher level models have been working best for me. I don’t use Gemini much except when I’m just doing a regular web search and I get the “AI results”. Those have been sucking less than they used to. I’m coding with ChatGPT Sol High or Extra High.

    Simon Gatrall | Product Development, Engineering, Design, Onshape | Ex- IDEO, PCH, Unagi, Carbon | LinkedIn

  • joshtargo
    joshtargo Posts: 645 EDU

    I use either Claude chat or Claude code and it's been very good. I've made highly complex features, and improved them with generally little trouble.

    A few things that can help:

    Use a mode where it can refer to the standard library document.

    Use it to create an app spec before coding anything.

    Have it create the UI first, this is where it needs the most help from a human.

    Give it coding rules you want, such as formatting, how much it chops the code into sub functions, etc.

Categories