•
Install DSPy, define a tiny typed module, run one predict/optimize loop—less hand-tuned prompting for after-work Python.
DSPy is a framework for programming—not just prompting—LLM pipelines: you declare signatures/modules and can optimize them with data. Home: dspy.ai.
After work, DSPy shines when you are tired of fragile prompt strings and want Python-shaped LLM apps.
Python learners who like functions, classes, and evaluation loops. If you have never written a function that returns a typed result, skim Intro first—then DSPy will click faster.
Overview: Top 10 AI agent tools for after-work Python.
python -m venv .venv
source .venv/bin/activate
pip install -U dspy
Follow dspy.ai to set dspy.LM(...) (or the current configure helper) with your provider key in the environment—not in source.
Something tiny:
import dspy
class Explain(dspy.Signature):
'''Explain a Python idea in one short paragraph for a beginner.'''
topic: str = dspy.InputField()
explanation: str = dspy.OutputField()
explain = dspy.Predict(Explain)
# after configuring LM per docs:
# print(explain(topic="list comprehensions").explanation)
Adjust field APIs to match the current docs if names differ.
Inspect what DSPy actually sent. That inspection beats memorizing “prompt engineering tips.”
If time remains, run the smallest example that uses a few labeled examples to improve the module. Stop after one compile/optimize cycle—do not build a dataset factory tonight.
Change the signature to require explanation and analogy. Call it on topic="recursion". If the analogy is empty, tighten the docstring—do not paste a 2-page prompt.
You practice LLM apps as software: interfaces, tests, iteration—skills that transfer beyond any single vendor prompt UI.
When signatures feel natural, deepen Python with Intro to Python Programming so modules and data classes stay clean as pipelines grow.