LangSmith can capture traces generated by Microsoft Agent Framework using its built-in OpenTelemetry instrumentation. This guide shows you how to automatically capture traces from your Microsoft Agent Framework agents and send them to LangSmith for monitoring and analysis.
Installation
Install the required packages:
pip install agent-framework opentelemetry-exporter-otlp-proto-http
Setup
Enable OpenTelemetry instrumentation of the agent and set the OpenTelemetry environment variables to point to the LangSmith OTEL endpoint:
export ENABLE_INSTRUMENTATION=true
export OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
export OTEL_EXPORTER_OTLP_ENDPOINT=https://api.smith.langchain.com/otel/v1/traces
export OTEL_EXPORTER_OTLP_HEADERS="x-api-key=<your_langsmith_api_key>,Langsmith-Project=<your_project_name>"
2. Enable OpenTelemetry in your application
In your Microsoft Agent Framework application, enable OpenTelemetry tracing using the built-in configure_otel_providers function:
from agent_framework.observability import configure_otel_providers
# Enable OpenTelemetry tracing
configure_otel_providers(enable_sensitive_data=True)
Setting enable_sensitive_data=True allows capturing input and output content in traces. Set to False if you want to exclude sensitive data from traces.
3. Create and run your agent
Once configured, your Microsoft Agent Framework agents will automatically send traces to LangSmith:
from agent_framework import ChatAgent
from agent_framework.observability import configure_otel_providers
from agent_framework.openai import OpenAIChatClient
# Enable OpenTelemetry tracing
configure_otel_providers(enable_sensitive_data=True)
agent = ChatAgent(
chat_client=OpenAIChatClient(model_id="gpt-4o"),
)
result = await agent.run("What's the the capital of Bavaria?")
print(result.text)
