feat(major): move new telemetry architecture into new provider

llama_stack/providers/inline/telemetry/otel/README.md

@@ -0,0 +1,26 @@
+# Open Telemetry Native Instrumentation
+
+This instrumentation package is simple, and follows expected open telemetry standards. It injects middleware for distributed tracing into all ingress and egress points into the application, and can be tuned and configured with OTEL environment variables.
+
+## Set Up
+
+First, bootstrap and install all necessary libraries for open telemtry:
+
+```
+uv run opentelemetry-bootstrap -a requirements | uv pip install --requirement -
+```
+
+Then, run with automatic code injection:
+
+```
+uv run opentelemetry-instrument llama stack run --config myconfig.yaml
+```
+
+### Excluded Fast API URLs
+
+```
+export OTEL_PYTHON_FASTAPI_EXCLUDED_URLS="client/.*/info,healthcheck"
+```
+
+#### Environment Variables
+Environment Variables: https://opentelemetry.io/docs/specs/otel/configuration/sdk-environment-variables/

llama_stack/providers/inline/telemetry/otel/config.py

@@ -0,0 +1,31 @@
+from typing import Literal
+
+from pydantic import BaseModel, Field
+
+
+type BatchSpanProcessor = Literal["batch"]
+type SimpleSpanProcessor = Literal["simple"]
+
+class OTelTelemetryConfig(BaseModel):
+    """
+    The configuration for the OpenTelemetry telemetry provider.
+    Most configuration is set using environment variables.
+    See https://opentelemetry.io/docs/specs/otel/configuration/sdk-environment-variables/ for more information.
+    """
+    service_name: str = Field(
+        description="""The name of the service to be monitored. 
+        Is overridden by the OTEL_SERVICE_NAME or OTEL_RESOURCE_ATTRIBUTES environment variables.""",
+    )
+    service_version: str | None = Field(
+        description="""The version of the service to be monitored. 
+        Is overriden by the OTEL_RESOURCE_ATTRIBUTES environment variable."""
+    )
+    deployment_environment: str | None = Field(
+        description="""The name of the environment of the service to be monitored. 
+        Is overriden by the OTEL_RESOURCE_ATTRIBUTES environment variable."""
+    )
+    span_processor: BatchSpanProcessor | SimpleSpanProcessor | None = Field(
+        description="""The span processor to use. 
+        Is overriden by the OTEL_SPAN_PROCESSOR environment variable.""",
+        default="batch"
+    )

llama_stack/providers/inline/telemetry/otel/otel.py

@@ -0,0 +1,63 @@
+import os
+
+from opentelemetry import trace, metrics
+from opentelemetry.sdk.resources import Attributes, Resource
+from opentelemetry.sdk.trace import TracerProvider
+from opentelemetry.sdk.trace.export import BatchSpanProcessor, SimpleSpanProcessor
+from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
+from opentelemetry.sdk.metrics import MeterProvider
+from opentelemetry.instrumentation.fastapi import FastAPIInstrumentor
+
+from llama_stack.core.telemetry.tracing import TelemetryProvider
+from llama_stack.log import get_logger
+
+from .config import OTelTelemetryConfig
+from fastapi import FastAPI
+
+
+logger = get_logger(name=__name__, category="telemetry::otel")
+
+
+class OTelTelemetryProvider(TelemetryProvider):
+    """
+    A simple Open Telemetry native telemetry provider.
+    """
+    def __init__(self, config: OTelTelemetryConfig):
+        self.config = config
+        attributes: Attributes = {
+            key: value
+            for key, value in {
+                "service.name": self.config.service_name,
+                "service.version": self.config.service_version,
+                "deployment.environment": self.config.deployment_environment,
+            }.items()
+            if value is not None
+        }
+
+        resource = Resource.create(attributes)
+
+        # Configure the tracer provider
+        tracer_provider = TracerProvider(resource=resource)
+        trace.set_tracer_provider(tracer_provider)
+
+        otlp_span_exporter = OTLPSpanExporter()
+
+        # Configure the span processor
+        # Enable batching of spans to reduce the number of requests to the collector
+        if self.config.span_processor == "batch":
+            tracer_provider.add_span_processor(BatchSpanProcessor(otlp_span_exporter))
+        elif self.config.span_processor == "simple":
+            tracer_provider.add_span_processor(SimpleSpanProcessor(otlp_span_exporter))
+        
+        meter_provider = MeterProvider(resource=resource)
+        metrics.set_meter_provider(meter_provider)
+
+        # Do not fail the application, but warn the user if the endpoints are not set properly
+        if not os.environ.get("OTEL_EXPORTER_OTLP_ENDPOINT"):
+            if not os.environ.get("OTEL_EXPORTER_OTLP_TRACES_ENDPOINT"):
+                logger.warning("OTEL_EXPORTER_OTLP_ENDPOINT or OTEL_EXPORTER_OTLP_TRACES_ENDPOINT is not set. Traces will not be exported.")
+            if not os.environ.get("OTEL_EXPORTER_OTLP_METRICS_ENDPOINT"):
+                logger.warning("OTEL_EXPORTER_OTLP_ENDPOINT or OTEL_EXPORTER_OTLP_METRICS_ENDPOINT is not set. Metrics will not be exported.")
+
+    def fastapi_middleware(self, app: FastAPI):
+        FastAPIInstrumentor.instrument_app(app)