`mod openinference`#

module openinference#

OpenInference subscriber support for NeMo Flow.

This crate adapts NeMo Flow lifecycle events into OpenInference trace spans:

scope/tool/LLM Start events open spans
matching End events close spans
Mark events become span events on the active parent span when possible
orphan marks fall back to zero-duration spans so they still reach OTLP

The public API is intentionally small:

OpenInferenceConfig configures the OTLP exporter and OpenInference metadata
OpenInferenceSubscriber exposes a NeMo Flow EventSubscriberFn and convenience register / deregister / force_flush / shutdown methods

Types

type Result<T>#: Result type for the OpenInference subscriber crate.

Enums

enum OpenInferenceError#

Errors produced while configuring or operating the OpenInference subscriber.

MissingTokioRuntime#: The tonic gRPC exporter requires an active Tokio runtime.

UnsupportedTransport#

The requested transport is not available on this target.

transport: &'static str#: Human-readable transport label used in the error message.

InvalidGrpcHeader#

Failed to parse a configured gRPC metadata header.

key: String#: Header name that failed to parse.

message: String#: Parser failure message.

ExporterBuild(String)#: Failed to build the OTLP exporter.

Provider(String)#: The underlying tracer provider returned an error.

Core(FlowError)#: Registration errors from the core runtime.

enum OtlpTransport#

Supported OTLP trace transports.

HttpBinary#: OTLP/HTTP protobuf, typically http://host:4318/v1/traces.

Grpc#: OTLP/gRPC, typically http://host:4317.

Structs and Unions

struct OpenInferenceConfig#

Configuration for the OpenInference subscriber.

Implementations

impl OpenInferenceConfig#

Functions

fn new() -> Self#: Creates a config with sensible defaults.

fn with_endpoint(mut self, endpoint: impl Into<String>) -> Self#: Overrides the OTLP endpoint. If unset, exporter defaults and OTEL_* env vars apply.

fn with_header(mut self, key: impl Into<String>, value: impl Into<String>) -> Self#: Adds a header/metadata entry for the exporter.

fn with_instrumentation_scope(mut self, scope: impl Into<String>) -> Self#: Sets the instrumentation scope name used for emitted spans.

fn with_resource_attribute(mut self, key: impl Into<String>, value: impl Into<String>) -> Self#: Adds a resource attribute as a string key/value pair.

fn with_service_name(mut self, service_name: impl Into<String>) -> Self#: Sets the service.name resource attribute.

fn with_service_namespace(mut self, namespace: impl Into<String>) -> Self#: Sets the service namespace resource attribute.

fn with_service_version(mut self, version: impl Into<String>) -> Self#: Sets the service version resource attribute.

fn with_timeout(mut self, timeout: Duration) -> Self#: Sets the OTLP request timeout.

fn with_transport(mut self, transport: OtlpTransport) -> Self#: Selects the OTLP transport.

Traits implemented

impl Default for OpenInferenceConfig#

struct OpenInferenceSubscriber#

OpenInference-backed NeMo Flow subscriber.

Implementations

impl OpenInferenceSubscriber#

Functions

fn deregister(&self, name: &str) -> Result<bool>#: Deregisters a previously-registered global subscriber by name.

fn force_flush(&self) -> Result<()>#: Flushes finished spans through the underlying tracer provider.

fn from_tracer_provider(provider: SdkTracerProvider, instrumentation_scope: impl Into<String>) -> Self#: Builds a subscriber from an already-configured tracer provider.

fn new(config: OpenInferenceConfig) -> Result<Self>#: Builds a subscriber backed by a new OTLP tracer provider.

fn register(&self, name: &str) -> Result<()>#: Registers this subscriber globally with the NeMo Flow runtime.

fn shutdown(&self) -> Result<()>#: Shuts down the underlying tracer provider.

Call deregister(...) first if the subscriber is still registered with NeMo Flow.

fn subscriber(&self) -> EventSubscriberFn#: Returns the raw NeMo Flow subscriber callback for custom registration flows.

mod openinference#

`mod openinference`#