Introducing PyMC-Bot: An AI Assistant for PyMC Questions (Alpha)

Hello PyMC community!

I’m excited to announce the launch of PyMC-Bot, an AI assistant dedicated to helping with PyMC-related questions right here on the forum. This is currently in alpha stage as we test its capabilities and gather feedback.

How to Use PyMC-Bot

Using the bot is simple:

  1. Create a new topic or post in an existing topic
  2. In your post, tag @pymc-bot at the beginning of your post
  3. Write your PyMC-related question after the command

For example:
@pymc-bot How do I implement a hierarchical model with varying slopes in PyMC?

Current Limitations

As this is an alpha release, please note:

  • The bot is being manually triggered for now, so responses won’t be immediate
  • It works best with clear, focused questions about PyMC, PyTensor, and Bayesian modeling
  • It may not always provide perfect answers, especially for complex or cutting-edge questions
  • The bot only responds to new topics without any existing replies

Feedback Welcome

We’d love to hear your thoughts on PyMC-Bot! Please let us know if its responses are helpful, and feel free to suggest improvements. You can just leave comments and questions on this thread.

Technical Details

PyMC-Bot is powered by Onyx, a retrieval-augmented generation (RAG) system that combines large language models with PyMC documentation and resources. The bot was built by @twiecki at PyMC Labs to help make PyMC more accessible to users of all experience levels.

PyMCheers and keep on sampling!

12 Likes

Love the idea! I suppose you have already considered it and it may have some limitations (I see you are staying away from automation with the manual trigger) but the first thing that came to my mind was to use it in private messages so as not to overpopulate the forum with replies that I would imagine (but I may be totally wrong) require a little bit of fine tuning on the hands of the questioner to be precise.

Anyway, I think this approach to start experimentation is great. When I read Thomas’ post on Linkedin discussing his implementation I thought: “I want that!”, and now I have it! So thank you very much for sharing your work!!

2 Likes

Just to echo the above…and also to say a personal well done. It looks fantastic.