Skip to content
Pydantic Docs

DSPy

See every step your DSPy program takes: each module call, the prompts it builds, the model’s responses, how many tokens it used, and any errors, as a trace (the full journey of one request, made of nested spans, where each span is one unit of work with a name, a start, and a duration) in Logfire. DSPy composes prompts and model calls into programs, and this integration shows each nested step in one trace.

What you’ll capture

  • Each DSPy module call as a span, with its duration and any exceptions
  • The prompts DSPy builds and the model’s responses
  • Nested steps of your program, shown as child spans in the trace
  • Response details, including the number of tokens used

Before you start

You’ll need a Logfire project. Open Add data in your project (top navigation) and follow the setup for your language: it signs your machine in with logfire auth (a browser sign-in, no token to copy) and, for production or other languages, creates a write token (the credential your app uses to send data). New to Logfire? Start with Getting Started.

You’ll also need an API key for whichever model provider DSPy uses (for example, OPENAI_API_KEY for the openai/... model below). DSPy reads it from the provider’s environment variable.

Installation

Install logfire with the dspy extra and the DSPy package:

Terminal
pip install 'logfire[dspy]'
Terminal
pip install dspy

Usage

Add two lines to your app: logfire.configure() to connect to your project, and logfire.instrument_dspy() to record every DSPy step.

import dspy

import logfire

logfire.configure()
logfire.instrument_dspy()

lm = dspy.LM('openai/gpt-5-mini')
dspy.configure(lm=lm)


class ExtractInfo(dspy.Signature):
    """Extract structured information from text."""

    text: str = dspy.InputField()
    title: str = dspy.OutputField()
    headings: list[str] = dspy.OutputField()
    entities: list[dict[str, str]] = dspy.OutputField(desc='a list of entities and their metadata')


module = dspy.Predict(ExtractInfo)

text = (
    'Apple Inc. announced its latest iPhone 14 today. '
    'The CEO, Tim Cook, highlighted its new features in a press release.'
)
response = module(text=text)

print(response.title)
print(response.headings)
print(response.entities)

Verify it worked

Run your program, then open your project in the Logfire web app and go to the Live view. Within a few seconds you should see a trace for the DSPy run, with a span for each module call. Click into it to see the prompts, responses, and token counts.

Troubleshooting

Not seeing your DSPy steps in Logfire? Check these first:

  • logfire.configure() runs before logfire.instrument_dspy(). Configure the connection first, then instrument.
  • You called instrument_dspy() exactly once.
  • Your Logfire write token is set. In local development, run logfire projects use <your-project>; in production, set the LOGFIRE_TOKEN environment variable. See Getting Started.

Reference