Otel

LlamaIndexOpenTelemetry

Bases: BaseModel

LlamaIndexOpenTelemetry is a configuration and integration class for OpenTelemetry tracing within LlamaIndex. This class manages the setup and registration of OpenTelemetry span and event handlers, configures the tracer provider, and exports trace data using the specified span exporter and processor. It supports both simple and batch span processors, and allows customization of the service name or resource, as well as the dispatcher name.

Attributes:

Name	Type	Description
span_exporter	Optional[SpanExporter]	The OpenTelemetry span exporter. Defaults to ConsoleSpanExporter.
span_processor	Literal['simple', 'batch']	The span processor type, either 'simple' or 'batch'. Defaults to 'batch'.
service_name_or_resource	Union[str, Resource]	The service name or OpenTelemetry Resource. Defaults to a Resource with service name 'llamaindex.opentelemetry'.

Source code in llama-index-integrations/observability/llama-index-observability-otel/llama_index/observability/otel/base.py

class LlamaIndexOpenTelemetry(BaseModel):
    """
    LlamaIndexOpenTelemetry is a configuration and integration class for OpenTelemetry tracing within LlamaIndex.
    This class manages the setup and registration of OpenTelemetry span and event handlers, configures the tracer provider,
    and exports trace data using the specified span exporter and processor. It supports both simple and batch span processors,
    and allows customization of the service name or resource, as well as the dispatcher name.

    Attributes:
        span_exporter (Optional[SpanExporter]): The OpenTelemetry span exporter. Defaults to ConsoleSpanExporter.
        span_processor (Literal["simple", "batch"]): The span processor type, either 'simple' or 'batch'. Defaults to 'batch'.
        service_name_or_resource (Union[str, Resource]): The service name or OpenTelemetry Resource. Defaults to a Resource with service name 'llamaindex.opentelemetry'.

    """

    model_config = ConfigDict(arbitrary_types_allowed=True)
    span_exporter: Optional[SpanExporter] = Field(
        default=ConsoleSpanExporter(),
        description="OpenTelemetry span exporter. Supports all SpanExporter-compatible interfaces, defaults to ConsoleSpanExporter.",
    )
    span_processor: Literal["simple", "batch"] = Field(
        default="batch",
        description="OpenTelemetry span processor. Can be either 'batch' (-> BatchSpanProcessor), 'simple' (-> SimpleSpanProcessor). Defaults to 'batch'",
    )
    extra_span_processors: List[SpanProcessor] = Field(
        default_factory=list,
        description="List of OpenTelemetry Span Processors to add to the tracer provider.",
    )
    tracer_provider: Optional[TracerProvider] = Field(
        default=None,
        description="Tracer Provider to inherint from the existing observability context. Defaults to None.",
    )
    service_name_or_resource: Union[str, Resource] = Field(
        default=Resource(attributes={SERVICE_NAME: "llamaindex.opentelemetry"}),
        description="Service name or resource for OpenTelemetry. Defaults to a Resource with 'llamaindex.opentelemetry' as service name.",
    )
    debug: bool = Field(
        default=False,
        description="Debug the start and end of span and the recording of events",
    )
    _tracer: Optional[trace.Tracer] = PrivateAttr(default=None)
    _tracer_provider_instance: Optional[TracerProvider] = PrivateAttr(default=None)

    def _start_otel(
        self,
    ) -> None:
        if isinstance(self.service_name_or_resource, str):
            self.service_name_or_resource = Resource(
                attributes={SERVICE_NAME: self.service_name_or_resource}
            )
        if self.tracer_provider is None:
            tracer_provider = TracerProvider(resource=self.service_name_or_resource)
        else:
            tracer_provider = self.tracer_provider
        assert self.span_exporter is not None, (
            "span_exporter has to be non-null to be used within simple or batch span processors"
        )
        if self.span_processor == "simple":
            span_processor = SimpleSpanProcessor(self.span_exporter)
        else:
            span_processor = BatchSpanProcessor(self.span_exporter)
        for extra_span_processor in self.extra_span_processors:
            tracer_provider.add_span_processor(extra_span_processor)
        tracer_provider.add_span_processor(span_processor)
        trace.set_tracer_provider(tracer_provider)
        self._tracer_provider_instance = tracer_provider
        self._tracer = trace.get_tracer("llamaindex.opentelemetry.tracer")

    def start_registering(
        self,
    ) -> None:
        """Starts LlamaIndex instrumentation."""
        self._start_otel()
        dispatcher = instrument.get_dispatcher()
        assert self._tracer is not None, (
            "The tracer has to be non-null to start observabiliy"
        )
        span_handler = OTelCompatibleSpanHandler(
            tracer=self._tracer,
            debug=self.debug,
            tracer_provider=self._tracer_provider_instance,
        )
        dispatcher.add_span_handler(span_handler)
        dispatcher.add_event_handler(
            OTelCompatibleEventHandler(span_handler=span_handler, debug=self.debug)
        )

start_registering

start_registering() -> None

Starts LlamaIndex instrumentation.

Source code in llama-index-integrations/observability/llama-index-observability-otel/llama_index/observability/otel/base.py

def start_registering(
    self,
) -> None:
    """Starts LlamaIndex instrumentation."""
    self._start_otel()
    dispatcher = instrument.get_dispatcher()
    assert self._tracer is not None, (
        "The tracer has to be non-null to start observabiliy"
    )
    span_handler = OTelCompatibleSpanHandler(
        tracer=self._tracer,
        debug=self.debug,
        tracer_provider=self._tracer_provider_instance,
    )
    dispatcher.add_span_handler(span_handler)
    dispatcher.add_event_handler(
        OTelCompatibleEventHandler(span_handler=span_handler, debug=self.debug)
    )

options: members: - LlamaIndexOpenTelemetry