DSPy

Serve a DSPy program as an AgentOS agent.

DSPyAgent wraps any DSPy Module so it can be served through AgentOS or used standalone.

import dspy
from agno.agents.dspy import DSPyAgent
from agno.db.sqlite import SqliteDb
from agno.os import AgentOS

dspy.configure(lm=dspy.LM("openai/gpt-5.4"))

agent = DSPyAgent(
    name="DSPy Assistant",
    program=dspy.ChainOfThought("question -> answer"),
)

agent_os = AgentOS(
    agents=[agent],
    tracing=True,
    db=SqliteDb(db_file="tmp/agentos.db"),
)
app = agent_os.get_app()

if __name__ == "__main__":
    agent_os.serve(app="dspy_agent:app", reload=True)

Install

uv pip install dspy
export OPENAI_API_KEY=sk-...

Parameters

Parameter	Type	Default	Description
name	str	None	Display name for the agent.
id	str	None	Unique identifier. Auto-generated from name if unset.
program	dspy.Module	None	A DSPy module (Predict, ChainOfThought, ReAct, or custom).
input_field	str	"question"	Name of the input field in the DSPy signature.
output_field	str	"answer"	Name of the output field on the Prediction.
lm	dspy.LM	None	Optional LM to scope this agent. Falls back to the global dspy.configure(lm=...).
program_kwargs	Dict[str, Any]	{}	Extra kwargs passed to the program on every call.
db	BaseDb	None	Database for session persistence.

Examples

AgentOS deployment

Serve a DSPy program through AgentOS.

Standalone usage

Call the agent directly with .run() and .print_response().

ReAct with tools

Use dspy.ReAct to call tools from the program.

Sessions

Resume conversations across runs with session_id.

Developer Resources

• Cookbook examples
• DSPy documentation