How to beginner · 3 min read

How to compile DSPy programs

Q: How to compile DSPy programs

To compile DSPy programs, define your dspy.Signature classes and use dspy.Predict to create callable modules. Then call the module with input arguments to get compiled outputs. Use dspy.configure() with an LM instance for OpenAI or other LLM providers.

Quick answer

To compile DSPy programs, define your dspy.Signature classes and use dspy.Predict to create callable modules. Then call the module with input arguments to get compiled outputs. Use dspy.configure() with an LM instance for OpenAI or other LLM providers.

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
pip install dspy openai>=1.0

Setup

Install dspy and openai packages, and set your OpenAI API key as an environment variable.

Run pip install dspy openai
Set environment variable OPENAI_API_KEY with your API key

bash

pip install dspy openai

Step by step

Define a dspy.Signature class for your program, configure the LLM, create a dspy.Predict instance, and call it with inputs to compile and run your DSPy program.

python

import os
import dspy
from openai import OpenAI

# Configure the LLM with OpenAI API key
lm = dspy.LM("openai/gpt-4o-mini", api_key=os.environ["OPENAI_API_KEY"])
dspy.configure(lm=lm)

# Define a DSPy signature for a simple QA task
class QA(dspy.Signature):
    question: str = dspy.InputField()
    answer: str = dspy.OutputField()

# Create a DSPy Predict module
qa = dspy.Predict(QA)

# Compile and run the DSPy program
result = qa(question="What is DSPy?")
print("Answer:", result.answer)

output

Answer: DSPy is a declarative Python framework for AI programming that compiles structured prompts to LLMs.

Common variations

You can use different LLM models by changing the dspy.LM model string, or switch to async calls by using await qa(question="...") inside an async function. DSPy also supports chain-of-thought with dspy.ChainOfThought.

python

import asyncio

async def async_example():
    result = await qa(question="Explain RAG.")
    print("Async answer:", result.answer)

asyncio.run(async_example())

output

Async answer: RAG stands for Retrieval-Augmented Generation, combining retrieval of documents with LLM generation.

Troubleshooting

If you get errors like API key missing, ensure OPENAI_API_KEY is set in your environment. For AttributeError, verify you installed the latest dspy and openai packages. If responses are empty, check your prompt and model usage.

✅

Key Takeaways

Use dspy.LM with your OpenAI API key to configure DSPy compilation.
Define structured input/output with dspy.Signature classes for clean program interfaces.
Call dspy.Predict instances with inputs to compile and run DSPy programs.
DSPy supports async calls and chain-of-thought for advanced workflows.
Always verify environment variables and package versions to avoid common errors.

Verified 2026-04 · openai/gpt-4o-mini

Verify ↗