LMQL is a powerful query programming language created specifically for working with large language models (LLMs). It blends natural language prompts with Python‑style scripting, giving developers fine‑grained control over how LLMs generate text, enforce constraints, and manage outputs. Unlike standard prompt templates, LMQL allows you to define LLM interactions as code, with control flow, constraints (e.g., stop conditions, length requirements), and advanced decoding strategies. It supports multiple model backends — including OpenAI, local transformers, and llama.cpp — and comes with tools like nested queries and tool augmentation to integrate Python functions or external logic directly into the generation process. This makes it ideal for building custom AI agents, structured generation workflows, chatbots, and advanced AI‑assisted applications while keeping full flexibility and control over model behavior.