Programming, not prompting
Stanford's framework for programming LLM pipelines: signatures and automatic prompt optimization.
Stanford NLP's framework for programming (not prompting) LLM pipelines: declare modules and metrics, and its optimizers tune prompts/fine-tunes automatically. The research-grade answer to fragile hand-prompting, popular for RAG and classifier pipelines that must improve measurably.
Who it's for
ML teams who want to program (not prompt) language models with compile-time optimisation of prompts.
Official links
Keep exploring
Concepts in the glossary
1 in catalog
DSPy is written in
1 in catalog
DSPy features
1 in catalog