Applied AI·Structured outputs and tool calling
you hand the local model a grammar and it cannot emit anything outside your format, not even a helpful preamble.
Grammar-based decoding
Draft summary, pending review
Constrained decoding driven by a formal grammar rather than a JSON schema, such as llama.cpp's GBNF format. Lets local models emit exact custom formats: SQL dialects, DSLs, or fixed-choice answers.