With Braq, I'm experimenting to what extent we can interweave human-readable structured data with prose within a single document. The original ideas were first implemented in Jesth [1] which spawned two spin-offs, Paradict [2] and Braq.

So far, I have discovered the usefulness of Braq for creating flexible config files and I am witnessing a potentially interesting proposal [3] that has emerged naturally to better structure AI prompts [4].

The eponymous Python package for tinkering with Braq is available on PyPI (see the Installation section in the README). Feel free to play with examples and code snippets in the README, or clone the repository to explore the source code and the tests package. The functions and classes have been documented with well-structured docstrings (whose format is also based on Braq) which will feed a documentation generator that I am developing.

Let me know what you think of the project.

[1] https://news.ycombinator.com/item?id=35991018 (Show HN: Jesth – Next-level human-readable data serialization format)

[2] https://news.ycombinator.com/item?id=38684724 (Show HN: Paradict – Streamable multi-format serialization with schema)

[3] https://github.com/pyrustic/braq?tab=readme-ov-file#ai-promp...

[4] https://news.ycombinator.com/item?id=34988748 (OpenAI's ChatML)