DSPy
Builds model pipelines as programs whose prompts are optimised rather than written.
Aperçu
DSPy treats a model pipeline as a program with declared inputs and outputs, and compiles the prompts by measuring which wording scores best on examples. It solves the problem that hand-tuned prompts are brittle: they encode one person's guess about phrasing and stop working when the model or the task shifts. It exists because prompt wording is a parameter that can be searched, and searching it beats guessing at it.
Graphe des technologies
Comment DSPy se rattache au reste de la pile.
Outils
Logiciels qui implémentent DSPy ou s'y intègrent.