dspy

Automated prompt optimization uses data-driven teleprompters to systematically improve language model performance, replacing fragile manual prompt engineering with declarative signatures compiled into self-improving pipelines.

Build modular RAG pipelines by defining AI tasks using declarative input and output signatures rather than brittle templates, enabling portable, maintainable systems compiled by automated optimizers.

Compile multi-stage reasoning workflows across different language models by defining declarative signatures, ensuring model-agnostic pipeline portability and type-safe structured output without rewriting core logic.

dspy works with OpenAI and Anthropic models for AI agent development, allowing you to compile declarative language model calls into self-improving pipelines that maintain portability across providers.

Replace brittle prompt templates in complex AI systems by using declarative programming signatures, which compile into self-improving pipelines that automatically optimize prompt quality based on training data.

Use built-in teleprompters to systematically improve prompt quality by providing a declarative signature and a set of training examples to the optimizer, compiling data-driven self-improving pipelines.