Open SourceFramework22k
DSPy
Program language models instead of prompting them, then auto-optimize.
DSPy is a framework from Stanford NLP for programming, rather than hand-prompting, language models by composing modular, declarative pipelines. Its optimizers (formerly teleprompters) automatically tune prompts and few-shot examples against a metric, so pipelines improve without manual prompt engineering. It is notable for treating prompts as learnable parameters and is widely used for building robust, self-optimizing LLM systems.
Repository
stanfordnlp/dspyLanguage
PythonWhat you'd build with it
- Building LLM pipelines whose prompts are auto-optimized against a metric
- Replacing brittle hand-tuned prompts with declarative, composable modules
- Compiling multi-stage reasoning and RAG systems that self-improve on examples
Tags
optimizationpromptingpipelinesagents