DSPy
llm-frameworkprompt-optimizationpythonai-systems
Abstraction: Python framework replacing prompt engineering with optimizable typed signatures
Key points:
- DSPy expresses AI tasks as typed input/output signatures rather than hand-crafted prompts, producing maintainable, modular programs
- Three core primitives: Signatures (task definition), Modules (execution strategies — reasoning, ensembles, tools, REPL), and Optimizers (automatic prompt tuning given examples and a scoring function)
- Optimizers tune prompts automatically until quality converges, removing manual prompt engineering iteration
- Originated at Stanford NLP in December 2022; grown into a research-and-production community
- New optimizers and module types land in DSPy before appearing in production systems
Connections: Dspy · Stanford · Prompt Engineering · Large Language Models
Source: https://dspy.ai/