How to beginner · 3 min read

DSPy signatures explained

Quick answer

In dspy, a Signature is a Pydantic-based class that defines structured input and output fields for AI model calls. Use dspy.InputField() to declare inputs and dspy.OutputField() for outputs, enabling type-safe, declarative AI interactions.

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 subclass with input and output fields, then create a dspy.Predict instance to call the AI model with structured data.

python

import os
import dspy
from openai import OpenAI

# Initialize OpenAI LLM
client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])
llm = dspy.LM("openai/gpt-4o-mini", client=client)
dspy.configure(lm=llm)

# Define a signature with input and output fields
class QA(dspy.Signature):
    question: str = dspy.InputField()
    answer: str = dspy.OutputField()

# Create a Predict instance
qa = dspy.Predict(QA)

# Call the model with structured input
result = qa(question="What is DSPy?")
print("Answer:", result.answer)

output

Answer: DSPy is a declarative Python library that simplifies AI model calls by defining structured input and output schemas using Pydantic signatures.

Common variations

You can use dspy.ChainOfThought for step-by-step reasoning, switch models by changing the LM initialization, or run async calls with asyncio.

python

import asyncio

async def async_qa():
    result = await qa.acall(question="Explain chain-of-thought in DSPy.")
    print("Async answer:", result.answer)

asyncio.run(async_qa())

output

Async answer: Chain-of-thought in DSPy allows you to define multi-step reasoning by chaining signatures for clearer AI explanations.

Troubleshooting

If you see ValidationError, check that your input types match the signature fields.
If the API key is missing or invalid, set OPENAI_API_KEY correctly in your environment.
For unexpected output, verify your model name and update dspy.LM accordingly.

✅

Key Takeaways

Use dspy.Signature to declare structured AI inputs and outputs with type safety.
Instantiate dspy.Predict with your signature to call AI models cleanly.
Configure your LLM with dspy.LM and set the API key via environment variables.
Leverage dspy.ChainOfThought for multi-step reasoning workflows.
Always validate input types to avoid runtime errors and ensure predictable AI responses.

Verified 2026-04 · openai/gpt-4o-mini

Verify ↗