docs: Add MLflow tracing example for voice agents

examples/voice_agents/README.md

@@ -87,6 +87,7 @@ session = AgentSession(
 ### 📊 Tracing & Error Handling
 
 - [`langfuse_trace.py`](./langfuse_trace.py) - LangFuse integration for conversation tracing
+- [`mlflow_trace.py`](./mlflow_trace.py) - MLflow integration for conversation tracing
 - [`error_callback.py`](./error_callback.py) - Error handling callback
 - [`session_close_callback.py`](./session_close_callback.py) - Session lifecycle management
 

examples/voice_agents/mlflow_trace.py

@@ -0,0 +1,120 @@
+import logging
+import os
+
+from dotenv import load_dotenv
+from opentelemetry.sdk.resources import SERVICE_NAME, Resource
+from opentelemetry.sdk.trace import TracerProvider
+
+from livekit.agents import (
+    Agent,
+    AgentServer,
+    AgentSession,
+    JobContext,
+    RunContext,
+    cli,
+    inference,
+    metrics,
+)
+from livekit.agents.llm import function_tool
+from livekit.agents.telemetry import set_tracer_provider
+from livekit.agents.voice import MetricsCollectedEvent
+from livekit.plugins import silero
+
+logger = logging.getLogger("mlflow-trace-example")
+
+load_dotenv()
+
+# This example shows how to use MLflow to trace the agent session.
+# MLflow accepts OpenTelemetry traces via its OTLP endpoint.
+#
+# Prerequisites:
+#   pip install mlflow
+#   mlflow server --port 5000
+#
+# Environment variables:
+#   OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:5000
+#   OTEL_EXPORTER_OTLP_HEADERS=x-mlflow-experiment-id=0
+#
+# Learn more: https://mlflow.org/docs/latest/genai/tracing/integrations/listing/livekit
+
+
+def setup_mlflow(
+    *,
+    endpoint: str | None = None,
+    experiment_id: str | None = None,
+    service_name: str | None = None,
+) -> TracerProvider:
+    """Configure OpenTelemetry to send traces to MLflow.
+
+    Args:
+        endpoint: MLflow tracking server URL (default: OTEL_EXPORTER_OTLP_ENDPOINT or http://localhost:5000)
+        experiment_id: MLflow experiment ID (default: OTEL_EXPORTER_OTLP_HEADERS or "0")
+        service_name: Service name for spans (default: OTEL_SERVICE_NAME or "livekit-agent")
+    """
+    from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
+    from opentelemetry.sdk.trace.export import BatchSpanProcessor
+
+    endpoint = endpoint or os.getenv("OTEL_EXPORTER_OTLP_ENDPOINT", "http://localhost:5000")
+    experiment_id = experiment_id or "0"
+    service_name = service_name or os.getenv("OTEL_SERVICE_NAME", "livekit-agent")
+
+    os.environ["OTEL_EXPORTER_OTLP_ENDPOINT"] = endpoint
+    os.environ["OTEL_EXPORTER_OTLP_HEADERS"] = os.getenv(
+        "OTEL_EXPORTER_OTLP_HEADERS", f"x-mlflow-experiment-id={experiment_id}"
+    )
+
+    resource = Resource.create({SERVICE_NAME: service_name})
+    trace_provider = TracerProvider(resource=resource)
+    trace_provider.add_span_processor(BatchSpanProcessor(OTLPSpanExporter()))
+    set_tracer_provider(trace_provider)
+    return trace_provider
+
+
+@function_tool
+async def lookup_weather(context: RunContext, location: str) -> str:
+    """Called when the user asks for weather related information.
+
+    Args:
+        location: The location they are asking for
+    """
+    logger.info(f"Looking up weather for {location}")
+    return "sunny with a temperature of 70 degrees."
+
+
+class MyAgent(Agent):
+    def __init__(self) -> None:
+        super().__init__(
+            instructions="You are a helpful voice assistant. Keep your responses concise and conversational.",
+            llm=inference.LLM("openai/gpt-4.1-mini"),
+            stt=inference.STT("deepgram/nova-3"),
+            tts=inference.TTS("cartesia/sonic-3"),
+            tools=[lookup_weather],
+        )
+
+    async def on_enter(self):
+        self.session.generate_reply()
+
+
+server = AgentServer()
+
+
+@server.rtc_session()
+async def entrypoint(ctx: JobContext):
+    trace_provider = setup_mlflow()
+
+    async def flush_trace():
+        trace_provider.force_flush()
+
+    ctx.add_shutdown_callback(flush_trace)
+
+    session = AgentSession(vad=silero.VAD.load())
+
+    @session.on("metrics_collected")
+    def _on_metrics_collected(ev: MetricsCollectedEvent):
+        metrics.log_metrics(ev.metrics)
+
+    await session.start(agent=MyAgent(), room=ctx.room)
+
+
+if __name__ == "__main__":
+    cli.run_app(server)