LMQL is a SQL-like programming language for interacting with LMs. It takes a declarative approach to specifying the output constraints for a language model, with a SQL flavor.


Microsoft created a project called guidance which is an LLM-agnostic language to “interleave generation, prompting, and logical control into a single continuous flow matching how the language model actually processes the text”. It’s based on Handlebars templates and provides in-template notion for system and user messages.


kor is a prototype used to extract structured data from language model responses. It takes a more object oriented approach and is Python-specific.


rebuff is a multilayer prompt injection defense library that continuously improves and adapts based on previous attacks.

On this day

2026-05-16

3 years later

I keep hearing and reading the phrase "load bearing" and even though I suspect it isn't always, I can only read it as LLM generated content