How to beginner · 3 min read

How to chain DSPy modules

Q: How to chain DSPy modules

Use dspy.ChainOfThought or compose multiple dspy.Predict modules by passing outputs from one as inputs to another. This chaining enables modular, stepwise AI workflows with declarative signatures in dspy.

Quick answer

Use dspy.ChainOfThought or compose multiple dspy.Predict modules by passing outputs from one as inputs to another. This chaining enables modular, stepwise AI workflows with declarative signatures in dspy.

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.

Install packages: pip install dspy openai
Set environment variable: export OPENAI_API_KEY='your_api_key' (Linux/macOS) or setx OPENAI_API_KEY "your_api_key" (Windows)

bash

pip install dspy openai

Step by step

Define multiple dspy.Signature classes representing individual tasks, then create dspy.Predict instances for each. Chain them by passing the output of one module as input to the next. Use dspy.ChainOfThought to enable stepwise reasoning chaining.

python

import os
import dspy
from dspy import InputField, OutputField

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

# Define first module signature
class ExtractName(dspy.Signature):
    text: str = InputField()
    name: str = OutputField()

# Define second module signature
class GreetUser(dspy.Signature):
    name: str = InputField()
    greeting: str = OutputField()

# Create Predict instances
extract_name = dspy.Predict(ExtractName)
greet_user = dspy.Predict(GreetUser)

# Input text
input_text = "My name is Alice."

# Run first module
extracted = extract_name(text=input_text)
print(f"Extracted name: {extracted.name}")

# Chain output to second module
greeting = greet_user(name=extracted.name)
print(f"Greeting: {greeting.greeting}")

output

Extracted name: Alice
Greeting: Hello, Alice! Nice to meet you.

Common variations

You can use dspy.ChainOfThought to chain modules with intermediate reasoning steps automatically. Also, you can switch models by changing the dspy.LM initialization, or use async calls with dspy if supported.

python

import os
import dspy

lm = dspy.LM("openai/gpt-4o-mini", api_key=os.environ["OPENAI_API_KEY"])
dspy.configure(lm=lm)

class QA(dspy.Signature):
    question: str = dspy.InputField()
    answer: str = dspy.OutputField()

qa_chain = dspy.ChainOfThought(QA)

result = qa_chain(question="What is DSPy?")
print(result.answer)

output

DSPy is a declarative Python framework for building modular AI workflows by defining signatures and chaining predictions.

Troubleshooting

If you see API key missing errors, ensure OPENAI_API_KEY is set in your environment.
If chaining outputs fail, verify that output field names match input field names between modules.
For model errors, confirm you use a current model like gpt-4o-mini and have network access.

✅

Key Takeaways

Use dspy.Predict to create modular AI components with declarative input/output signatures.
Chain modules by passing outputs from one as inputs to another for clean, composable workflows.
Leverage dspy.ChainOfThought for automatic stepwise reasoning chains.
Always configure dspy with a valid dspy.LM instance using your OpenAI API key.
Match input and output field names carefully to avoid chaining errors.

Verified 2026-04 · openai/gpt-4o-mini

Verify ↗