Context Engineering with DSPy

Chapter 7. Customizing DSPy Programs

A note for Early Release readers

With Early Release ebooks, you get books in their earliest form—the author’s raw and unedited content as they write—so you can take advantage of these technologies long before the official release of these titles.

This will be the 7th chapter of the final book. The GitHub repo will be made active later on.

If you’d like to be actively involved in reviewing and commenting on this draft, please reach out to the editor at [e-mail].

In Chapter 6, you mastered DSPy’s optimizers: the algorithms that tune your prompts, select few-shot examples, and even fine-tune model weights automatically. But as we noted at the end of that chapter, if the task isn’t succeeding after a few optimization runs, it’s most likely a module problem. The optimizer can only work with the strategy you give it. A bad module with a great optimizer still produces bad results.

Modules are just inference strategies, for spending more tokens in a way that guides the model towards better results. Because DSPy defines your task with Signatures rather than string-based templates, it’s easy to swap in a different strategy without rewriting your program. Your Signature stays the same. Your dataset stays the same. Your metrics stay the same. Only the inference strategy changes.

This is one of Omar Khattab’s core bets for DSPy: inference strategies should be polymorphic modules. That sounds academic, but the practical implication is powerful. ...