DSPy

DSPy is a framework that systematically optimizes language model prompts and weights, automating the tuning process. Muster captures DSPy runs through the OpenInference instrumentation, including Chain-of-Thought, RAG, and ReAct programs.

Setup

1. Install dependencies

%pip install langfuse dspy openinference-instrumentation-dspy -U

2. Configure credentials

import os

os.environ["LANGFUSE_PUBLIC_KEY"] = "pk-lf-..."
os.environ["LANGFUSE_SECRET_KEY"] = "sk-lf-..."
os.environ["LANGFUSE_BASE_URL"] = "https://app.getmuster.io"
os.environ["OPENAI_API_KEY"] = "sk-..."

3. Enable tracing

from openinference.instrumentation.dspy import DSPyInstrumentor

DSPyInstrumentor().instrument()

4. Configure DSPy

import dspy

lm = dspy.LM("openai/gpt-4o-mini")
dspy.configure(lm=lm)

Example use cases captured

The instrumentation captures spans for the standard DSPy idioms:

• Chain-of-Thought reasoning for mathematical problems.
• RAG pipelines combining search with question answering.
• ReAct modules using multiple tools for complex reasoning.

Each program execution becomes a trace in Muster with sub-spans for every LM call, retrieval, and tool invocation.

Add user/session attributes

from langfuse import propagate_attributes

with propagate_attributes(user_id="user_123", session_id="sess_abc", tags=["dspy"]):
    # run your DSPy program
    ...

Troubleshooting

• Enable debug logging: export LANGFUSE_DEBUG="True".
• In short-lived scripts, call langfuse.flush() before exit.
• Filter unrelated OTel spans from unrelated libraries.

See also

• Upstream Langfuse DSPy docs
• LangChain