The lazy approach to building and modifying Monty configs

Monty Code
1

During the London retreat, we worked on simplifying and improving the Monty configs, among other projects. Our configurations are now composable, rather than being overridden through inheritance. This is a big step toward making Monty easier to use and understand. If you’re interested in more details, check out @tslominski’s post, or watch our demo presentation.

We also built a Terminal User Interface (TUI) called lazyconfigs to make it easier for us and the community to quickly put together new experiment config files or modify existing ones. I walk through a demo of this in the presentation.

lazyconfigs
lazyconfigs1860×1014 206 KB

lazyconfigs is now available in this repository and can be installed with:

brew tap thousandbrainsproject/tap
brew install --cask lazyconfigs

Alternatively, you can build from source. The tool is relatively straightforward to use and supports customizable themes and keybindings. Let us know if you run into any issues or have feedback.

3 Likes
2

Thanks, all, for helping make Monty development more approachable. As it gets closer and closer to the point of being painless, more folks may decide to jump in. Specifically, I’ve noted:

  • the team’s work on composability and framework independence
  • lazyconfigs and other promising work (e.g., Pythonx integration)

To that point, I’d like to suggest a complementary approach. Basically, it would use skills and setup prompts to assist agentic AI tools (e.g., Codex CLI) with setting up and interacting with Monty.

FWIW, I’m not totally clear on what would need to be in these, but:

  • setup prompts would include background information, links, etc.
  • skills would handle typical interaction scenarios

If anyone is using this sort of approach, I’d love to hear about it.

-r