feat: http disconnects (#2014)

lib/llm/src/http/service.rs

@@ -20,6 +20,7 @@
 
 mod openai;
 
+pub mod disconnect;
 pub mod error;
 pub mod health;
 pub mod metrics;

lib/llm/src/http/service/disconnect.rs

+// SPDX-FileCopyrightText: Copyright (c) 2024-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
+
+//! The `disconnect` module provides a mechanism for our axum http services to monitoring and responding
+//! to disconnects from the client.
+//!
+//! There are two potential phases in any request where we need to handle the disconnect.
+//!
+//! For unary, request-response, there is just a single phase where the primary task that axum kicks off
+//! to handle the request will be dropped if the client disconnects. In order for us to have a long running
+//! task, like an LLM request, we need to spawn our long running task in a separate task and then spawn
+//! a second task that will monitor for disconnects from the client. The primary task which spawned the
+//! two tasks will hold an "armed" [`ConnectionHandle`] which will issue a [`ConnectionStatus::ClosedUnexpectedly`]
+//! if the task is dropped before it is [`ConnectionHandle::disarm`]ed.
+//!
+//! For the streaming case, request in - stream out, we need a second [`ConnectionHandle`] which will be owned
+//! by the stream. A streaming response is when the [`axum::response::Response]] is a [axum::response::Sse] stream.
+//! This means the primary task handle will go out of scope when it returns the stream. When we create our
+//! SSE stream, we capture the second [`ConnectionHandle`] and arm it. If the stream closes gracefully, the
+//! second handle will be disarmed, otherwise, the stream was dropped and the [`Drop`] trait on the [`ConnectionHandle`]
+//! triggers a [`ConnectionStatus::ClosedUnexpectedly`] signal.
+//!
+//! The [`ConnectionHandle`] is a simple wrapper around a [`tokio::sync::oneshot::Sender`] which will send a
+//! [`ConnectionStatus`] enum to the primary task. The primary task will then use this to determine if it should
+//! cancel the request or not.
+//!
+//! The [`ConnectionHandle`] is also used to signal to the client that the request has been cancelled. This is
+//! done by sending a [`axum::response::sse::Event`] with the event type "error" and the data "[DONE]".
+//!
+
+use axum::response::sse::Event;
+use dynamo_runtime::engine::AsyncEngineContext;
+use futures::{Stream, StreamExt};
+use std::sync::Arc;
+
+use crate::http::service::metrics::InflightGuard;
+
+#[derive(Clone, Copy)]
+pub enum ConnectionStatus {
+    Disabled,
+    ClosedUnexpectedly,
+    ClosedGracefully,
+}
+
+pub struct ConnectionHandle {
+    sender: Option<tokio::sync::oneshot::Sender<ConnectionStatus>>,
+    on_drop: ConnectionStatus,
+}
+
+impl ConnectionHandle {
+    /// Handle which by default will issue a [`ConnectionStatus::ClosedGracefully`] signal when dropped.
+    pub fn create_disarmed(sender: tokio::sync::oneshot::Sender<ConnectionStatus>) -> Self {
+        Self {
+            sender: Some(sender),
+            on_drop: ConnectionStatus::ClosedGracefully,
+        }
+    }
+
+    /// Handle which will issue a [`ConnectionStatus::ClosedUnexpectedly`] signal when dropped.
+    pub fn create_armed(sender: tokio::sync::oneshot::Sender<ConnectionStatus>) -> Self {
+        Self {
+            sender: Some(sender),
+            on_drop: ConnectionStatus::ClosedUnexpectedly,
+        }
+    }
+
+    /// Handle which will not issue a signal when dropped.
+    pub fn create_disabled(sender: tokio::sync::oneshot::Sender<ConnectionStatus>) -> Self {
+        Self {
+            sender: Some(sender),
+            on_drop: ConnectionStatus::Disabled,
+        }
+    }
+
+    /// Handle which will issue a [`ConnectionStatus::ClosedGracefully`] signal when dropped.
+    pub fn disarm(&mut self) {
+        self.on_drop = ConnectionStatus::ClosedGracefully;
+    }
+
+    /// Handle which will issue a [`ConnectionStatus::ClosedUnexpectedly`] signal when dropped.
+    pub fn arm(&mut self) {
+        self.on_drop = ConnectionStatus::ClosedUnexpectedly;
+    }
+}
+
+impl Drop for ConnectionHandle {
+    fn drop(&mut self) {
+        if let Some(sender) = self.sender.take() {
+            let _ = sender.send(self.on_drop);
+        }
+    }
+}
+
+/// Creates a pair of handles which will monitor for disconnects from the client.
+///
+/// The first handle is armed and will issue a [`ConnectionStatus::ClosedUnexpectedly`] signal when dropped.
+/// The second handle is disarmed and will issue a [`ConnectionStatus::ClosedGracefully`] signal when dropped.
+///
+/// The handles are returned in the order of the first being armed and the second being disarmed.
+pub async fn create_connection_monitor(
+    engine_context: Arc<dyn AsyncEngineContext>,
+) -> (ConnectionHandle, ConnectionHandle) {
+    // these oneshot channels monitor possible disconnects from the client in two different scopes:
+    // - the local task (connection_handle)
+    // - an optionally streaming response (stream_handle)
+    let (connection_tx, connection_rx) = tokio::sync::oneshot::channel();
+    let (stream_tx, stream_rx) = tokio::sync::oneshot::channel();
+
+    // detached task that will naturally close when both handles are dropped
+    tokio::spawn(connection_monitor(
+        engine_context.clone(),
+        connection_rx,
+        stream_rx,
+    ));
+
+    // Two handles, the first is armed, the second is disarmed
+    (
+        ConnectionHandle::create_armed(connection_tx),
+        ConnectionHandle::create_disabled(stream_tx),
+    )
+}
+
+#[tracing::instrument(level = "trace", skip_all, fields(request_id = %engine_context.id()))]
+async fn connection_monitor(
+    engine_context: Arc<dyn AsyncEngineContext>,
+    connection_rx: tokio::sync::oneshot::Receiver<ConnectionStatus>,
+    stream_rx: tokio::sync::oneshot::Receiver<ConnectionStatus>,
+) {
+    match connection_rx.await {
+        Err(_) | Ok(ConnectionStatus::ClosedUnexpectedly) => {
+            // the client has disconnected, no need to gracefully cancel, just kill the context
+            tracing::trace!("Connection closed unexpectedly; issuing cancellation");
+            engine_context.kill();
+        }
+        Ok(ConnectionStatus::ClosedGracefully) => {
+            tracing::trace!("Connection closed gracefully");
+        }
+        Ok(ConnectionStatus::Disabled) => {}
+    }
+
+    match stream_rx.await {
+        Err(_) | Ok(ConnectionStatus::ClosedUnexpectedly) => {
+            tracing::trace!("Stream closed unexpectedly; issuing cancellation");
+            engine_context.kill();
+        }
+        Ok(ConnectionStatus::ClosedGracefully) => {
+            tracing::trace!("Stream closed gracefully");
+        }
+        Ok(ConnectionStatus::Disabled) => {}
+    }
+}
+
+/// This method will consume a stream of SSE events and monitor for disconnects or context cancellation.
+///
+/// Uses `tokio::select!` to choose between receiving events from the source stream or detecting when
+/// the context is stopped. If the context is stopped, we break the stream. If the source stream ends
+/// naturally, we mark the request as successful and send the final `[DONE]` event.
+pub fn monitor_for_disconnects(
+    stream: impl Stream<Item = Result<Event, axum::Error>>,
+    context: Arc<dyn AsyncEngineContext>,
+    mut inflight_guard: InflightGuard,
+    mut stream_handle: ConnectionHandle,
+) -> impl Stream<Item = Result<Event, axum::Error>> {
+    stream_handle.arm();
+    async_stream::try_stream! {
+        tokio::pin!(stream);
+        loop {
+            tokio::select! {
+                event = stream.next() => {
+                    match event {
+                        Some(Ok(event)) => {
+                            yield event;
+                        }
+                        Some(Err(err)) => {
+                            yield Event::default().event("error").comment(err.to_string());
+                        }
+                        None => {
+                            // Stream ended normally
+                            inflight_guard.mark_ok();
+                            stream_handle.disarm();
+
+                            // todo: if we yield a dynamo sentinel event, we need to do it before the done or the
+                            // async-openai client will chomp it.
+                            yield Event::default().data("[DONE]");
+                            break;
+                        }
+                    }
+                }
+                _ = context.stopped() => {
+                    tracing::trace!("Context stopped; breaking stream");
+                    break;
+                }
+            }
+        }
+    }
+}

lib/llm/src/http/service/openai.rs

@@ -3,14 +3,13 @@
 
 use std::{
     collections::HashSet,
-    pin::Pin,
     sync::Arc,
     time::{SystemTime, UNIX_EPOCH},
 };
 
 use axum::{
     extract::State,
-    http::StatusCode,
+    http::{HeaderMap, StatusCode},
     response::{
         sse::{Event, KeepAlive, Sse},
         IntoResponse, Response,
@@ -18,14 +17,17 @@ use axum::{
     routing::{get, post},
     Json, Router,
 };
-use dynamo_runtime::pipeline::{AsyncEngineContext, Context};
-use futures::{Stream, StreamExt};
+use dynamo_runtime::{
+    pipeline::{AsyncEngineContextProvider, Context},
+    protocols::annotated::AnnotationsProvider,
+};
+use futures::{stream, StreamExt};
 use serde::{Deserialize, Serialize};
-use tokio_stream::wrappers::ReceiverStream;
 
 use super::{
+    disconnect::{create_connection_monitor, monitor_for_disconnects, ConnectionHandle},
     error::HttpError,
-    metrics::{Endpoint, InflightGuard, ResponseMetricCollector},
+    metrics::{Endpoint, ResponseMetricCollector},
     service_v2, RouteDoc,
 };
 use crate::preprocessor::LLMMetricAnnotation;
@@ -38,17 +40,24 @@ use crate::protocols::openai::{
 use crate::request_template::RequestTemplate;
 use crate::types::Annotated;
 
+pub const DYNAMO_REQUEST_ID_HEADER: &str = "x-dynamo-request-id";
+
+/// Dynamo Annotation for the request ID
+pub const ANNOTATION_REQUEST_ID: &str = "request_id";
+
+pub type ErrorResponse = (StatusCode, Json<ErrorMessage>);
+
 #[derive(Serialize, Deserialize)]
-pub(crate) struct ErrorResponse {
+pub(crate) struct ErrorMessage {
     error: String,
 }
 
-impl ErrorResponse {
+impl ErrorMessage {
     /// Not Found Error
-    pub fn model_not_found() -> (StatusCode, Json<ErrorResponse>) {
+    pub fn model_not_found() -> ErrorResponse {
         (
             StatusCode::NOT_FOUND,
-            Json(ErrorResponse {
+            Json(ErrorMessage {
                 error: "Model not found".to_string(),
             }),
         )
@@ -56,10 +65,10 @@ impl ErrorResponse {
 
     /// Service Unavailable
     /// This is returned when the service is live, but not ready.
-    pub fn _service_unavailable() -> (StatusCode, Json<ErrorResponse>) {
+    pub fn _service_unavailable() -> ErrorResponse {
         (
             StatusCode::SERVICE_UNAVAILABLE,
-            Json(ErrorResponse {
+            Json(ErrorMessage {
                 error: "Service is not ready".to_string(),
             }),
         )
@@ -69,11 +78,11 @@ impl ErrorResponse {
     /// Return this error when the service encounters an internal error.
     /// We should return a generic message to the client instead of the real error.
     /// Internal Services errors are the result of misconfiguration or bugs in the service.
-    pub fn internal_server_error(msg: &str) -> (StatusCode, Json<ErrorResponse>) {
+    pub fn internal_server_error(msg: &str) -> ErrorResponse {
         tracing::error!("Internal server error: {msg}");
         (
             StatusCode::INTERNAL_SERVER_ERROR,
-            Json(ErrorResponse {
+            Json(ErrorMessage {
                 error: msg.to_string(),
             }),
         )
@@ -82,11 +91,11 @@ impl ErrorResponse {
     /// Not Implemented Error
     /// Return this error when the client requests a feature that is not yet implemented.
     /// This should be used for features that are planned but not available.
-    pub fn not_implemented_error(msg: &str) -> (StatusCode, Json<ErrorResponse>) {
+    pub fn not_implemented_error(msg: &str) -> ErrorResponse {
         tracing::error!("Not Implemented error: {msg}");
         (
             StatusCode::NOT_IMPLEMENTED,
-            Json(ErrorResponse {
+            Json(ErrorMessage {
                 error: msg.to_string(),
             }),
         )
@@ -94,31 +103,56 @@ impl ErrorResponse {
 
     /// The OAI endpoints call an [`dynamo.runtime::engine::AsyncEngine`] which are specialized to return
     /// an [`anyhow::Error`]. This method will convert the [`anyhow::Error`] into an [`HttpError`].
-    /// If successful, it will return the [`HttpError`] as an [`ErrorResponse::internal_server_error`]
+    /// If successful, it will return the [`HttpError`] as an [`ErrorMessage::internal_server_error`]
     /// with the details of the error.
-    pub fn from_anyhow(err: anyhow::Error, alt_msg: &str) -> (StatusCode, Json<ErrorResponse>) {
+    pub fn from_anyhow(err: anyhow::Error, alt_msg: &str) -> ErrorResponse {
         match err.downcast::<HttpError>() {
-            Ok(http_error) => ErrorResponse::from_http_error(http_error),
-            Err(err) => ErrorResponse::internal_server_error(&format!("{alt_msg}: {err}")),
+            Ok(http_error) => ErrorMessage::from_http_error(http_error),
+            Err(err) => ErrorMessage::internal_server_error(&format!("{alt_msg}: {err}")),
         }
     }
 
     /// Implementers should only be able to throw 400-499 errors.
-    pub fn from_http_error(err: HttpError) -> (StatusCode, Json<ErrorResponse>) {
+    pub fn from_http_error(err: HttpError) -> ErrorResponse {
         if err.code < 400 || err.code >= 500 {
-            return ErrorResponse::internal_server_error(&err.message);
+            return ErrorMessage::internal_server_error(&err.message);
         }
         match StatusCode::from_u16(err.code) {
-            Ok(code) => (code, Json(ErrorResponse { error: err.message })),
-            Err(_) => ErrorResponse::internal_server_error(&err.message),
+            Ok(code) => (code, Json(ErrorMessage { error: err.message })),
+            Err(_) => ErrorMessage::internal_server_error(&err.message),
         }
     }
 }
 
-impl From<HttpError> for ErrorResponse {
+impl From<HttpError> for ErrorMessage {
     fn from(err: HttpError) -> Self {
-        ErrorResponse { error: err.message }
+        ErrorMessage { error: err.message }
+    }
+}
+
+/// Get the request ID from a primary source, or next from the headers, or lastly create a new one if not present
+fn get_or_create_request_id(primary: Option<&str>, headers: &HeaderMap) -> String {
+    // Try to get the request ID from the primary source
+    if let Some(primary) = primary {
+        if let Ok(uuid) = uuid::Uuid::parse_str(primary) {
+            return uuid.to_string();
+        }
     }
+
+    // Try to get the request ID header as a string slice
+    let request_id_opt = headers
+        .get(DYNAMO_REQUEST_ID_HEADER)
+        .and_then(|h| h.to_str().ok());
+
+    // Try to parse the request ID as a UUID, or generate a new one if missing/invalid
+    let uuid = match request_id_opt {
+        Some(request_id) => {
+            uuid::Uuid::parse_str(request_id).unwrap_or_else(|_| uuid::Uuid::new_v4())
+        }
+        None => uuid::Uuid::new_v4(),
+    };
+
+    uuid.to_string()
 }
 
 /// OpenAI Completions Request Handler
@@ -129,11 +163,46 @@ impl From<HttpError> for ErrorResponse {
 ///
 /// Note: For all requests, streaming or non-streaming, we always call the engine with streaming enabled. For
 /// non-streaming requests, we will fold the stream into a single response as part of this handler.
-#[tracing::instrument(skip_all)]
-async fn completions(
+async fn handler_completions(
     State(state): State<Arc<service_v2::State>>,
+    headers: HeaderMap,
     Json(request): Json<NvCreateCompletionRequest>,
-) -> Result<Response, (StatusCode, Json<ErrorResponse>)> {
+) -> Result<Response, ErrorResponse> {
+    // return a 503 if the service is not ready
+    check_ready(&state)?;
+
+    // create the context for the request
+    let request_id = get_or_create_request_id(request.inner.user.as_deref(), &headers);
+    let request = Context::with_id(request, request_id);
+    let context = request.context();
+
+    // create the connection handles
+    let (mut connection_handle, stream_handle) = create_connection_monitor(context.clone()).await;
+
+    // possibly long running task
+    // if this returns a streaming response, the stream handle will be armed and captured by the response stream
+    let response = tokio::spawn(completions(state, request, stream_handle))
+        .await
+        .map_err(|e| {
+            ErrorMessage::internal_server_error(&format!(
+                "Failed to await chat completions task: {:?}",
+                e,
+            ))
+        })?;
+
+    // if we got here, then we will return a response and the potentially long running task has completed successfully
+    // without need to be cancelled.
+    connection_handle.disarm();
+
+    response
+}
+
+#[tracing::instrument(skip_all)]
+async fn completions(
+    state: Arc<service_v2::State>,
+    request: Context<NvCreateCompletionRequest>,
+    stream_handle: ConnectionHandle,
+) -> Result<Response, ErrorResponse> {
     // return a 503 if the service is not ready
     check_ready(&state)?;
 
@@ -144,15 +213,10 @@ async fn completions(
     let streaming = request.inner.stream.unwrap_or(false);
 
     // update the request to always stream
-    let inner = async_openai::types::CreateCompletionRequest {
-        stream: Some(true),
-        ..request.inner
-    };
-
-    let request = NvCreateCompletionRequest {
-        inner,
-        nvext: request.nvext,
-    };
+    let request = request.map(|mut req| {
+        req.inner.stream = Some(true);
+        req
+    });
 
     // todo - make the protocols be optional for model name
     // todo - when optional, if none, apply a default
@@ -162,7 +226,7 @@ async fn completions(
     let engine = state
         .manager()
         .get_completions_engine(model)
-        .map_err(|_| ErrorResponse::model_not_found())?;
+        .map_err(|_| ErrorMessage::model_not_found())?;
 
     let mut inflight_guard =
         state
@@ -171,27 +235,43 @@ async fn completions(
 
     let mut response_collector = state.metrics_clone().create_response_collector(model);
 
-    // setup context
-    // todo - inherit request_id from distributed trace details
-    let request = Context::with_id(request, request_id.clone());
+    // prepare to process any annotations
+    let annotations = request.annotations();
 
     // issue the generate call on the engine
     let stream = engine
         .generate(request)
         .await
-        .map_err(|e| ErrorResponse::from_anyhow(e, "Failed to generate completions"))?;
+        .map_err(|e| ErrorMessage::from_anyhow(e, "Failed to generate completions"))?;
 
     // capture the context to cancel the stream if the client disconnects
     let ctx = stream.context();
 
-    // todo - tap the stream and propagate request level metrics
-    // note - we might do this as part of the post processing set to make it more generic
+    let annotations = annotations.map_or(Vec::new(), |annotations| {
+        annotations
+            .iter()
+            .filter_map(|annotation| {
+                if annotation == ANNOTATION_REQUEST_ID {
+                    Annotated::<NvCreateCompletionResponse>::from_annotation(
+                        ANNOTATION_REQUEST_ID,
+                        &request_id,
+                    )
+                    .ok()
+                } else {
+                    None
+                }
+            })
+            .collect::<Vec<_>>()
+    });
+
+    // apply any annotations to the front of the stream
+    let stream = stream::iter(annotations).chain(stream);
 
     if streaming {
         let stream = stream.map(move |response| {
             process_event_converter(EventConverter::from(response), &mut response_collector)
         });
-        let stream = monitor_for_disconnects(stream.boxed(), ctx, inflight_guard).await;
+        let stream = monitor_for_disconnects(stream, ctx, inflight_guard, stream_handle);
 
         let mut sse_stream = Sse::new(stream);
 
@@ -202,7 +282,7 @@ async fn completions(
         Ok(sse_stream.into_response())
     } else {
         // TODO: report ISL/OSL for non-streaming requests
-        let response = NvCreateCompletionResponse::from_annotated_stream(stream.into())
+        let response = NvCreateCompletionResponse::from_annotated_stream(stream)
             .await
             .map_err(|e| {
                 tracing::error!(
@@ -210,7 +290,7 @@ async fn completions(
                     request_id,
                     e
                 );
-                ErrorResponse::internal_server_error("Failed to fold completions stream")
+                ErrorMessage::internal_server_error("Failed to fold completions stream")
             })?;
 
         inflight_guard.mark_ok();
@@ -222,7 +302,7 @@ async fn completions(
 async fn embeddings(
     State(state): State<Arc<service_v2::State>>,
     Json(request): Json<NvCreateEmbeddingRequest>,
-) -> Result<Response, (StatusCode, Json<ErrorResponse>)> {
+) -> Result<Response, ErrorResponse> {
     // return a 503 if the service is not ready
     check_ready(&state)?;
 
@@ -240,7 +320,7 @@ async fn embeddings(
     let engine = state
         .manager()
         .get_embeddings_engine(model)
-        .map_err(|_| ErrorResponse::model_not_found())?;
+        .map_err(|_| ErrorMessage::model_not_found())?;
 
     // this will increment the inflight gauge for the model
     let mut inflight =
@@ -256,11 +336,11 @@ async fn embeddings(
     let stream = engine
         .generate(request)
         .await
-        .map_err(|e| ErrorResponse::from_anyhow(e, "Failed to generate embeddings"))?;
+        .map_err(|e| ErrorMessage::from_anyhow(e, "Failed to generate embeddings"))?;
 
     // Embeddings are typically returned as a single response (non-streaming)
     // so we fold the stream into a single response
-    let response = NvCreateEmbeddingResponse::from_annotated_stream(stream.into())
+    let response = NvCreateEmbeddingResponse::from_annotated_stream(stream)
         .await
         .map_err(|e| {
             tracing::error!(
@@ -268,13 +348,45 @@ async fn embeddings(
                 request_id,
                 e
             );
-            ErrorResponse::internal_server_error("Failed to fold embeddings stream")
+            ErrorMessage::internal_server_error("Failed to fold embeddings stream")
         })?;
 
     inflight.mark_ok();
     Ok(Json(response).into_response())
 }
 
+async fn handler_chat_completions(
+    State((state, template)): State<(Arc<service_v2::State>, Option<RequestTemplate>)>,
+    headers: HeaderMap,
+    Json(request): Json<NvCreateChatCompletionRequest>,
+) -> Result<Response, ErrorResponse> {
+    // return a 503 if the service is not ready
+    check_ready(&state)?;
+
+    // create the context for the request
+    let request_id = get_or_create_request_id(request.inner.user.as_deref(), &headers);
+    let request = Context::with_id(request, request_id);
+    let context = request.context();
+
+    // create the connection handles
+    let (mut connection_handle, stream_handle) = create_connection_monitor(context.clone()).await;
+
+    let response = tokio::spawn(chat_completions(state, template, request, stream_handle))
+        .await
+        .map_err(|e| {
+            ErrorMessage::internal_server_error(&format!(
+                "Failed to await chat completions task: {:?}",
+                e,
+            ))
+        })?;
+
+    // if we got here, then we will return a response and the potentially long running task has completed successfully
+    // without need to be cancelled.
+    connection_handle.disarm();
+
+    response
+}
+
 /// OpenAI Chat Completions Request Handler
 ///
 /// This method will handle the incoming request for the /v1/chat/completions endpoint. The endpoint is a "source"
@@ -283,21 +395,23 @@ async fn embeddings(
 ///
 /// Note: For all requests, streaming or non-streaming, we always call the engine with streaming enabled. For
 /// non-streaming requests, we will fold the stream into a single response as part of this handler.
-#[tracing::instrument(skip_all)]
+#[tracing::instrument(level = "debug", skip_all, fields(request_id = %request.id()))]
 async fn chat_completions(
-    State((state, template)): State<(Arc<service_v2::State>, Option<RequestTemplate>)>,
-    Json(mut request): Json<NvCreateChatCompletionRequest>,
-) -> Result<Response, (StatusCode, Json<ErrorResponse>)> {
+    state: Arc<service_v2::State>,
+    template: Option<RequestTemplate>,
+    mut request: Context<NvCreateChatCompletionRequest>,
+    mut stream_handle: ConnectionHandle,
+) -> Result<Response, ErrorResponse> {
     // return a 503 if the service is not ready
     check_ready(&state)?;
 
+    let request_id = request.id().to_string();
+
     // Handle unsupported fields - if Some(resp) is returned by
     // validate_chat_completion_unsupported_fields,
     // then a field was used that is unsupported. We will log an error message
     // and early return a 501 NOT_IMPLEMENTED status code. Otherwise, proceeed.
-    if let Some(resp) = validate_chat_completion_unsupported_fields(&request) {
-        return Ok(resp.into_response());
-    }
+    validate_chat_completion_unsupported_fields(&request)?;
 
     // Apply template values if present
     if let Some(template) = template {
@@ -311,24 +425,16 @@ async fn chat_completions(
             request.inner.max_completion_tokens = Some(template.max_completion_tokens);
         }
     }
-    tracing::trace!("Received chat completions request: {:?}", request.inner);
-
-    // todo - extract distributed tracing id and context id from headers
-    let request_id = uuid::Uuid::new_v4().to_string();
+    tracing::trace!("Received chat completions request: {:?}", request.content());
 
     // todo - decide on default
     let streaming = request.inner.stream.unwrap_or(false);
 
     // update the request to always stream
-    let inner_request = async_openai::types::CreateChatCompletionRequest {
-        stream: Some(true),
-        ..request.inner
-    };
-
-    let request = NvCreateChatCompletionRequest {
-        inner: inner_request,
-        nvext: request.nvext,
-    };
+    let request = request.map(|mut req| {
+        req.inner.stream = Some(true);
+        req
+    });
 
     // todo - make the protocols be optional for model name
     // todo - when optional, if none, apply a default
@@ -340,7 +446,7 @@ async fn chat_completions(
     let engine = state
         .manager()
         .get_chat_completions_engine(model)
-        .map_err(|_| ErrorResponse::model_not_found())?;
+        .map_err(|_| ErrorMessage::model_not_found())?;
 
     let mut inflight_guard =
         state
@@ -349,29 +455,45 @@ async fn chat_completions(
 
     let mut response_collector = state.metrics_clone().create_response_collector(model);
 
-    // setup context
-    // todo - inherit request_id from distributed trace details
-    let request = Context::with_id(request, request_id.clone());
-
     tracing::trace!("Issuing generate call for chat completions");
+    let annotations = request.annotations();
 
     // issue the generate call on the engine
     let stream = engine
         .generate(request)
         .await
-        .map_err(|e| ErrorResponse::from_anyhow(e, "Failed to generate completions"))?;
+        .map_err(|e| ErrorMessage::from_anyhow(e, "Failed to generate completions"))?;
 
     // capture the context to cancel the stream if the client disconnects
     let ctx = stream.context();
 
+    // prepare any requested annotations
+    let annotations = annotations.map_or(Vec::new(), |annotations| {
+        annotations
+            .iter()
+            .filter_map(|annotation| {
+                if annotation == ANNOTATION_REQUEST_ID {
+                    Annotated::from_annotation(ANNOTATION_REQUEST_ID, &request_id).ok()
+                } else {
+                    None
+                }
+            })
+            .collect::<Vec<_>>()
+    });
+
+    // apply any annotations to the front of the stream
+    let stream = stream::iter(annotations).chain(stream);
+
     // todo - tap the stream and propagate request level metrics
     // note - we might do this as part of the post processing set to make it more generic
 
     if streaming {
+        stream_handle.arm();
+
         let stream = stream.map(move |response| {
             process_event_converter(EventConverter::from(response), &mut response_collector)
         });
-        let stream = monitor_for_disconnects(stream.boxed(), ctx, inflight_guard).await;
+        let stream = monitor_for_disconnects(stream, ctx, inflight_guard, stream_handle);
 
         let mut sse_stream = Sse::new(stream);
 
@@ -382,7 +504,7 @@ async fn chat_completions(
         Ok(sse_stream.into_response())
     } else {
         // TODO: report ISL/OSL for non-streaming requests
-        let response = NvCreateChatCompletionResponse::from_annotated_stream(stream.into())
+        let response = NvCreateChatCompletionResponse::from_annotated_stream(stream)
             .await
             .map_err(|e| {
                 tracing::error!(
@@ -390,7 +512,7 @@ async fn chat_completions(
                     "Failed to fold chat completions stream for: {:?}",
                     e
                 );
-                ErrorResponse::internal_server_error(&format!(
+                ErrorMessage::internal_server_error(&format!(
                     "Failed to fold chat completions stream: {}",
                     e
                 ))
@@ -406,44 +528,77 @@ async fn chat_completions(
 #[allow(deprecated)]
 pub fn validate_chat_completion_unsupported_fields(
     request: &NvCreateChatCompletionRequest,
-) -> Option<impl IntoResponse> {
+) -> Result<(), ErrorResponse> {
     let inner = &request.inner;
 
     if inner.parallel_tool_calls == Some(true) {
-        return Some(ErrorResponse::not_implemented_error(
+        return Err(ErrorMessage::not_implemented_error(
             "`parallel_tool_calls: true` is not supported.",
         ));
     }
 
     if inner.stream == Some(true) && inner.tools.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Err(ErrorMessage::not_implemented_error(
             "`stream: true` is not supported when `tools` are provided.",
         ));
     }
 
     if inner.function_call.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Err(ErrorMessage::not_implemented_error(
             "`function_call` is deprecated. Please migrate to use `tool_choice` instead.",
         ));
     }
 
     if inner.functions.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Err(ErrorMessage::not_implemented_error(
             "`functions` is deprecated. Please migrate to use `tools` instead.",
         ));
     }
 
-    None
+    Ok(())
 }
 
 /// OpenAI Responses Request Handler
 ///
 /// This method will handle the incoming request for the /v1/responses endpoint.
-#[tracing::instrument(skip_all)]
-async fn responses(
+async fn handler_responses(
     State((state, template)): State<(Arc<service_v2::State>, Option<RequestTemplate>)>,
-    Json(mut request): Json<NvCreateResponse>,
-) -> Result<Response, (StatusCode, Json<ErrorResponse>)> {
+    headers: HeaderMap,
+    Json(request): Json<NvCreateResponse>,
+) -> Result<Response, ErrorResponse> {
+    // return a 503 if the service is not ready
+    check_ready(&state)?;
+
+    // create the context for the request
+    let request_id = get_or_create_request_id(request.inner.user.as_deref(), &headers);
+    let request = Context::with_id(request, request_id);
+    let context = request.context();
+
+    // create the connection handles
+    let (mut connection_handle, _stream_handle) = create_connection_monitor(context.clone()).await;
+
+    let response = tokio::spawn(responses(state, template, request))
+        .await
+        .map_err(|e| {
+            ErrorMessage::internal_server_error(&format!(
+                "Failed to await chat completions task: {:?}",
+                e,
+            ))
+        })?;
+
+    // if we got here, then we will return a response and the potentially long running task has completed successfully
+    // without need to be cancelled.
+    connection_handle.disarm();
+
+    response
+}
+
+#[tracing::instrument(level = "debug", skip_all, fields(request_id = %request.id()))]
+async fn responses(
+    state: Arc<service_v2::State>,
+    template: Option<RequestTemplate>,
+    mut request: Context<NvCreateResponse>,
+) -> Result<Response, ErrorResponse> {
     // return a 503 if the service is not ready
     check_ready(&state)?;
 
@@ -476,21 +631,26 @@ async fn responses(
     }
     tracing::trace!("Received chat completions request: {:?}", request.inner);
 
-    let request_id = uuid::Uuid::new_v4().to_string();
+    let request_id = request.id().to_string();
+    let (request, context) = request.into_parts();
 
-    // Convert NvCreateResponse --> NvCreateChatCompletionRequest
-    let request: NvCreateChatCompletionRequest = request.try_into().map_err(|e| {
+    let mut request: NvCreateChatCompletionRequest = request.try_into().map_err(|e| {
         tracing::error!(
             request_id,
             "Failed to convert NvCreateResponse to NvCreateChatCompletionRequest: {:?}",
             e
         );
-        ErrorResponse::not_implemented_error(&format!(
+        ErrorMessage::not_implemented_error(&format!(
             "Only Input::Text(_) is currently supported: {}",
             e
         ))
     })?;
 
+    let request = context.map(|mut _req| {
+        request.inner.stream = Some(false);
+        request
+    });
+
     let model = &request.inner.model;
 
     tracing::trace!("Getting chat completions engine for model: {}", model);
@@ -498,7 +658,7 @@ async fn responses(
     let engine = state
         .manager()
         .get_chat_completions_engine(model)
-        .map_err(|_| ErrorResponse::model_not_found())?;
+        .map_err(|_| ErrorMessage::model_not_found())?;
 
     let mut inflight_guard =
         state
@@ -507,18 +667,16 @@ async fn responses(
 
     let _response_collector = state.metrics_clone().create_response_collector(model);
 
-    let request = Context::with_id(request, request_id.clone());
-
     tracing::trace!("Issuing generate call for chat completions");
 
     // issue the generate call on the engine
     let stream = engine
         .generate(request)
         .await
-        .map_err(|e| ErrorResponse::from_anyhow(e, "Failed to generate completions"))?;
+        .map_err(|e| ErrorMessage::from_anyhow(e, "Failed to generate completions"))?;
 
     // TODO: handle streaming, currently just unary
-    let response = NvCreateChatCompletionResponse::from_annotated_stream(stream.into())
+    let response = NvCreateChatCompletionResponse::from_annotated_stream(stream)
         .await
         .map_err(|e| {
             tracing::error!(
@@ -526,7 +684,7 @@ async fn responses(
                 "Failed to fold chat completions stream for: {:?}",
                 e
             );
-            ErrorResponse::internal_server_error(&format!(
+            ErrorMessage::internal_server_error(&format!(
                 "Failed to fold chat completions stream: {}",
                 e
             ))
@@ -539,7 +697,7 @@ async fn responses(
             "Failed to convert NvCreateChatCompletionResponse to NvResponse: {:?}",
             e
         );
-        ErrorResponse::internal_server_error("Failed to convert internal response")
+        ErrorMessage::internal_server_error("Failed to convert internal response")
     })?;
 
     inflight_guard.mark_ok();
@@ -552,7 +710,7 @@ pub fn validate_response_input_is_text_only(
 ) -> Option<impl IntoResponse> {
     match &request.inner.input {
         async_openai::types::responses::Input::Text(_) => None,
-        _ => Some(ErrorResponse::not_implemented_error("Only `Input::Text` is supported. Structured, multimedia, or custom input types are not yet implemented.")),
+        _ => Some(ErrorMessage::not_implemented_error("Only `Input::Text` is supported. Structured, multimedia, or custom input types are not yet implemented.")),
     }
 }
 
@@ -564,87 +722,87 @@ pub fn validate_response_unsupported_fields(
     let inner = &request.inner;
 
     if inner.background == Some(true) {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`background: true` is not supported.",
         ));
     }
     if inner.include.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`include` is not supported.",
         ));
     }
     if inner.instructions.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`instructions` is not supported.",
         ));
     }
     if inner.max_tool_calls.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`max_tool_calls` is not supported.",
         ));
     }
     if inner.metadata.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`metadata` is not supported.",
         ));
     }
     if inner.parallel_tool_calls == Some(true) {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`parallel_tool_calls: true` is not supported.",
         ));
     }
     if inner.previous_response_id.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`previous_response_id` is not supported.",
         ));
     }
     if inner.prompt.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`prompt` is not supported.",
         ));
     }
     if inner.reasoning.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`reasoning` is not supported.",
         ));
     }
     if inner.service_tier.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`service_tier` is not supported.",
         ));
     }
     if inner.store == Some(true) {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`store: true` is not supported.",
         ));
     }
     if inner.stream == Some(true) {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`stream: true` is not supported.",
         ));
     }
     if inner.text.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`text` is not supported.",
         ));
     }
     if inner.tool_choice.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`tool_choice` is not supported.",
         ));
     }
     if inner.tools.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`tools` is not supported.",
         ));
     }
     if inner.truncation.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`truncation` is not supported.",
         ));
     }
     if inner.user.is_some() {
-        return Some(ErrorResponse::not_implemented_error(
+        return Some(ErrorMessage::not_implemented_error(
             "`user` is not supported.",
         ));
     }
@@ -654,9 +812,9 @@ pub fn validate_response_unsupported_fields(
 
 // todo - abstract this to the top level lib.rs to be reused
 // todo - move the service_observer to its own state/arc
-fn check_ready(_state: &Arc<service_v2::State>) -> Result<(), (StatusCode, Json<ErrorResponse>)> {
+fn check_ready(_state: &Arc<service_v2::State>) -> Result<(), ErrorResponse> {
     // if state.service_observer.stage() != ServiceStage::Ready {
-    //     return Err(ErrorResponse::service_unavailable());
+    //     return Err(ErrorMessage::service_unavailable());
     // }
     Ok(())
 }
@@ -676,7 +834,7 @@ fn check_ready(_state: &Arc<service_v2::State>) -> Result<(), (StatusCode, Json<
 /// }
 async fn list_models_openai(
     State(state): State<Arc<service_v2::State>>,
-) -> Result<Response, (StatusCode, Json<ErrorResponse>)> {
+) -> Result<Response, ErrorResponse> {
     check_ready(&state)?;
 
     let created = SystemTime::now()
@@ -716,45 +874,6 @@ struct ModelListing {
     owned_by: String,
 }
 
-/// This method will consume a stream of SSE events and forward them to a new stream defined by a tokio channel.
-/// In this way, if the downstream is dropped, then the upstream will be unable to send any more events. This is
-/// how we can monitor for disconnects and stop the generation of completions.
-///
-/// If a disconnect is detected, then the context will issue a `stop_generating` call to the context which will
-/// propagate the cancellation signal to the backend.
-async fn monitor_for_disconnects(
-    stream: Pin<
-        Box<dyn Stream<Item = Result<axum::response::sse::Event, axum::Error>> + std::marker::Send>,
-    >,
-    context: Arc<dyn AsyncEngineContext>,
-    mut inflight_guard: InflightGuard,
-) -> ReceiverStream<Result<Event, axum::Error>> {
-    let (tx, rx) = tokio::sync::mpsc::channel(8);
-
-    tokio::spawn(async move {
-        let mut stream = stream;
-        while let Some(event) = stream.next().await {
-            let event = match event {
-                Ok(event) => Ok(event),
-                Err(err) => Ok(Event::default().event("error").comment(err.to_string())),
-            };
-
-            if (tx.send(event).await).is_err() {
-                tracing::trace!("Forwarding SSE stream was dropped; breaking loop");
-                context.stop_generating();
-                break;
-            }
-        }
-
-        // Stream completed successfully - mark as ok
-        if tx.send(Ok(Event::default().data("[DONE]"))).await.is_ok() {
-            inflight_guard.mark_ok();
-        }
-    });
-
-    ReceiverStream::new(rx)
-}
-
 struct EventConverter<T>(Annotated<T>);
 
 impl<T> From<Annotated<T>> for EventConverter<T> {
@@ -816,7 +935,7 @@ pub fn completions_router(
     let path = path.unwrap_or("/v1/completions".to_string());
     let doc = RouteDoc::new(axum::http::Method::POST, &path);
     let router = Router::new()
-        .route(&path, post(completions))
+        .route(&path, post(handler_completions))
         .with_state(state);
     (vec![doc], router)
 }
@@ -831,7 +950,7 @@ pub fn chat_completions_router(
     let path = path.unwrap_or("/v1/chat/completions".to_string());
     let doc = RouteDoc::new(axum::http::Method::POST, &path);
     let router = Router::new()
-        .route(&path, post(chat_completions))
+        .route(&path, post(handler_chat_completions))
         .with_state((state, template));
     (vec![doc], router)
 }
@@ -876,7 +995,7 @@ pub fn responses_router(
     let path = path.unwrap_or("/v1/responses".to_string());
     let doc = RouteDoc::new(axum::http::Method::POST, &path);
     let router = Router::new()
-        .route(&path, post(responses))
+        .route(&path, post(handler_responses))
         .with_state((state, template));
     (vec![doc], router)
 }
@@ -942,7 +1061,7 @@ mod tests {
     #[test]
     fn test_http_error_response_from_anyhow() {
         let err = http_error_from_engine(400).unwrap_err();
-        let (status, response) = ErrorResponse::from_anyhow(err, BACKUP_ERROR_MESSAGE);
+        let (status, response) = ErrorMessage::from_anyhow(err, BACKUP_ERROR_MESSAGE);
         assert_eq!(status, StatusCode::BAD_REQUEST);
         assert_eq!(response.error, "custom error message");
     }
@@ -950,17 +1069,17 @@ mod tests {
     #[test]
     fn test_error_response_from_anyhow_out_of_range() {
         let err = http_error_from_engine(399).unwrap_err();
-        let (status, response) = ErrorResponse::from_anyhow(err, BACKUP_ERROR_MESSAGE);
+        let (status, response) = ErrorMessage::from_anyhow(err, BACKUP_ERROR_MESSAGE);
         assert_eq!(status, StatusCode::INTERNAL_SERVER_ERROR);
         assert_eq!(response.error, "custom error message");
 
         let err = http_error_from_engine(500).unwrap_err();
-        let (status, response) = ErrorResponse::from_anyhow(err, BACKUP_ERROR_MESSAGE);
+        let (status, response) = ErrorMessage::from_anyhow(err, BACKUP_ERROR_MESSAGE);
         assert_eq!(status, StatusCode::INTERNAL_SERVER_ERROR);
         assert_eq!(response.error, "custom error message");
 
         let err = http_error_from_engine(501).unwrap_err();
-        let (status, response) = ErrorResponse::from_anyhow(err, BACKUP_ERROR_MESSAGE);
+        let (status, response) = ErrorMessage::from_anyhow(err, BACKUP_ERROR_MESSAGE);
         assert_eq!(status, StatusCode::INTERNAL_SERVER_ERROR);
         assert_eq!(response.error, "custom error message");
     }
@@ -968,7 +1087,7 @@ mod tests {
     #[test]
     fn test_other_error_response_from_anyhow() {
         let err = other_error_from_engine().unwrap_err();
-        let (status, response) = ErrorResponse::from_anyhow(err, BACKUP_ERROR_MESSAGE);
+        let (status, response) = ErrorMessage::from_anyhow(err, BACKUP_ERROR_MESSAGE);
         assert_eq!(status, StatusCode::INTERNAL_SERVER_ERROR);
         assert_eq!(
             response.error,

lib/llm/src/protocols.rs

@@ -19,7 +19,7 @@
 //! both publicly via the HTTP API and internally between Dynamo components.
 //!
 
-use futures::StreamExt;
+use futures::{Stream, StreamExt};
 use serde::{Deserialize, Serialize};
 
 pub mod codec;
@@ -49,12 +49,12 @@ pub trait ContentProvider {
 
 /// Converts of a stream of [codec::Message]s into a stream of [Annotated]s.
 pub fn convert_sse_stream<R>(
-    stream: DataStream<Result<codec::Message, codec::SseCodecError>>,
-) -> DataStream<Annotated<R>>
+    stream: impl Stream<Item = Result<codec::Message, codec::SseCodecError>>,
+) -> impl Stream<Item = Annotated<R>>
 where
     R: for<'de> Deserialize<'de> + Serialize,
 {
-    let stream = stream.map(|message| match message {
+    stream.map(|message| match message {
         Ok(message) => {
             let delta = Annotated::<R>::try_from(message);
             match delta {
@@ -63,6 +63,5 @@ where
             }
         }
         Err(e) => Annotated::from_error(e.to_string()),
-    });
-    Box::pin(stream)
+    })
 }

lib/llm/src/protocols/openai/chat_completions/aggregator.rs

@@ -13,7 +13,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use futures::StreamExt;
+use futures::{Stream, StreamExt};
 use std::collections::HashMap;
 
 use super::{NvCreateChatCompletionResponse, NvCreateChatCompletionStreamResponse};
@@ -22,7 +22,6 @@ use crate::protocols::{
     convert_sse_stream, Annotated,
 };
 
-/// A type alias for a pinned, dynamically-dispatched stream that is `Send` and `Sync`.
 use dynamo_runtime::engine::DataStream;
 
 /// Aggregates a stream of [`NvCreateChatCompletionStreamResponse`]s into a single
@@ -95,7 +94,7 @@ impl DeltaAggregator {
     /// * `Ok(NvCreateChatCompletionResponse)` if aggregation is successful.
     /// * `Err(String)` if an error occurs during processing.
     pub async fn apply(
-        stream: DataStream<Annotated<NvCreateChatCompletionStreamResponse>>,
+        stream: impl Stream<Item = Annotated<NvCreateChatCompletionStreamResponse>>,
     ) -> Result<NvCreateChatCompletionResponse, String> {
         let aggregator = stream
             .fold(DeltaAggregator::new(), |mut aggregator, delta| async move {
@@ -260,7 +259,7 @@ impl NvCreateChatCompletionResponse {
     /// * `Ok(NvCreateChatCompletionResponse)` if aggregation succeeds.
     /// * `Err(String)` if an error occurs.
     pub async fn from_annotated_stream(
-        stream: DataStream<Annotated<NvCreateChatCompletionStreamResponse>>,
+        stream: impl Stream<Item = Annotated<NvCreateChatCompletionStreamResponse>>,
     ) -> Result<NvCreateChatCompletionResponse, String> {
         DeltaAggregator::apply(stream).await
     }

lib/llm/src/protocols/openai/completions/aggregator.rs

@@ -16,7 +16,7 @@
 use std::collections::HashMap;
 
 use anyhow::Result;
-use futures::StreamExt;
+use futures::{Stream, StreamExt};
 
 use super::NvCreateCompletionResponse;
 use crate::protocols::{
@@ -64,7 +64,7 @@ impl DeltaAggregator {
 
     /// Aggregates a stream of [`Annotated<CompletionResponse>`]s into a single [`CompletionResponse`].
     pub async fn apply(
-        stream: DataStream<Annotated<NvCreateCompletionResponse>>,
+        stream: impl Stream<Item = Annotated<NvCreateCompletionResponse>>,
     ) -> Result<NvCreateCompletionResponse> {
         let aggregator = stream
             .fold(DeltaAggregator::new(), |mut aggregator, delta| async move {
@@ -183,7 +183,7 @@ impl NvCreateCompletionResponse {
     }
 
     pub async fn from_annotated_stream(
-        stream: DataStream<Annotated<NvCreateCompletionResponse>>,
+        stream: impl Stream<Item = Annotated<NvCreateCompletionResponse>>,
     ) -> Result<NvCreateCompletionResponse> {
         DeltaAggregator::apply(stream).await
     }

lib/llm/src/protocols/openai/embeddings/aggregator.rs

@@ -20,7 +20,7 @@ use crate::protocols::{
 };
 
 use dynamo_runtime::engine::DataStream;
-use futures::StreamExt;
+use futures::{Stream, StreamExt};
 
 /// Aggregates a stream of [`NvCreateEmbeddingResponse`]s into a single
 /// [`NvCreateEmbeddingResponse`]. For embeddings, this is typically simpler
@@ -58,7 +58,7 @@ impl DeltaAggregator {
     /// * `Ok(NvCreateEmbeddingResponse)` if aggregation is successful.
     /// * `Err(String)` if an error occurs during processing.
     pub async fn apply(
-        stream: DataStream<Annotated<NvCreateEmbeddingResponse>>,
+        stream: impl Stream<Item = Annotated<NvCreateEmbeddingResponse>>,
     ) -> Result<NvCreateEmbeddingResponse, String> {
         let aggregator = stream
             .fold(DeltaAggregator::new(), |mut aggregator, delta| async move {
@@ -133,7 +133,7 @@ impl NvCreateEmbeddingResponse {
     /// * `Ok(NvCreateEmbeddingResponse)` if aggregation succeeds.
     /// * `Err(String)` if an error occurs.
     pub async fn from_annotated_stream(
-        stream: DataStream<Annotated<NvCreateEmbeddingResponse>>,
+        stream: impl Stream<Item = Annotated<NvCreateEmbeddingResponse>>,
     ) -> Result<NvCreateEmbeddingResponse, String> {
         DeltaAggregator::apply(stream).await
     }

lib/llm/tests/http-service.rs

@@ -28,6 +28,8 @@ use dynamo_llm::http::{
     },
 };
 use dynamo_llm::protocols::{
+    codec::SseLineCodec,
+    convert_sse_stream,
     openai::{
         chat_completions::{NvCreateChatCompletionRequest, NvCreateChatCompletionStreamResponse},
         completions::{NvCreateCompletionRequest, NvCreateCompletionResponse},
@@ -45,11 +47,31 @@ use futures::StreamExt;
 use prometheus::{proto::MetricType, Registry};
 use reqwest::StatusCode;
 use rstest::*;
-use std::sync::Arc;
+use std::{io::Cursor, sync::Arc};
+use tokio::time::timeout;
+use tokio_util::codec::FramedRead;
 
 struct CounterEngine {}
 
-#[allow(deprecated)]
+// Add a new long-running test engine
+struct LongRunningEngine {
+    delay_ms: u64,
+    cancelled: Arc<std::sync::atomic::AtomicBool>,
+}
+
+impl LongRunningEngine {
+    fn new(delay_ms: u64) -> Self {
+        Self {
+            delay_ms,
+            cancelled: Arc::new(std::sync::atomic::AtomicBool::new(false)),
+        }
+    }
+
+    fn was_cancelled(&self) -> bool {
+        self.cancelled.load(std::sync::atomic::Ordering::Acquire)
+    }
+}
+
 #[async_trait]
 impl
     AsyncEngine<
@@ -66,6 +88,7 @@ impl
         let ctx = context.context();
 
         // ALLOW: max_tokens is deprecated in favor of completion_usage_tokens
+        #[allow(deprecated)]
         let max_tokens = request.inner.max_tokens.unwrap_or(0) as u64;
 
         // let generator = NvCreateChatCompletionStreamResponse::generator(request.model.clone());
@@ -88,6 +111,54 @@ impl
     }
 }
 
+#[async_trait]
+impl
+    AsyncEngine<
+        SingleIn<NvCreateChatCompletionRequest>,
+        ManyOut<Annotated<NvCreateChatCompletionStreamResponse>>,
+        Error,
+    > for LongRunningEngine
+{
+    async fn generate(
+        &self,
+        request: SingleIn<NvCreateChatCompletionRequest>,
+    ) -> Result<ManyOut<Annotated<NvCreateChatCompletionStreamResponse>>, Error> {
+        let (_request, context) = request.transfer(());
+        let ctx = context.context();
+
+        tracing::info!(
+            "LongRunningEngine: Starting generation with {}ms delay",
+            self.delay_ms
+        );
+
+        let cancelled_flag = self.cancelled.clone();
+        let delay_ms = self.delay_ms;
+
+        let ctx_clone = ctx.clone();
+        let stream = async_stream::stream! {
+
+            // the stream can be dropped or it can be cancelled
+            // either way we consider this a cancellation
+            cancelled_flag.store(true, std::sync::atomic::Ordering::SeqCst);
+
+            tokio::select! {
+                _ = tokio::time::sleep(std::time::Duration::from_millis(delay_ms)) => {
+                    // the stream went to completion
+                    cancelled_flag.store(false, std::sync::atomic::Ordering::SeqCst);
+
+                }
+                _ = ctx_clone.stopped() => {
+                    cancelled_flag.store(true, std::sync::atomic::Ordering::SeqCst);
+                }
+            }
+
+            yield Annotated::<NvCreateChatCompletionStreamResponse>::from_annotation("event.dynamo.test.sentinel", &"DONE".to_string()).expect("Failed to create annotated response");
+        };
+
+        Ok(ResponseStream::new(Box::pin(stream), ctx))
+    }
+}
+
 struct AlwaysFailEngine {}
 
 #[async_trait]
@@ -880,3 +951,311 @@ async fn test_generic_byot_client(
     cancel_token.cancel();
     task.await.unwrap().unwrap();
 }
+
+#[rstest]
+#[tokio::test]
+async fn test_client_disconnect_cancellation_unary() {
+    let service = HttpService::builder().port(8993).build().unwrap();
+    let state = service.state_clone();
+    let manager = state.manager();
+
+    let token = CancellationToken::new();
+    let cancel_token = token.clone();
+
+    // Start the service
+    let task = tokio::spawn(async move { service.run(token).await });
+
+    // Wait for service to be ready
+    wait_for_service_ready(8993).await;
+
+    // Create a long-running engine (10 seconds)
+    let long_running_engine = Arc::new(LongRunningEngine::new(10_000));
+    manager
+        .add_chat_completions_model("slow-model", long_running_engine.clone())
+        .unwrap();
+
+    let client = reqwest::Client::new();
+
+    let message = async_openai::types::ChatCompletionRequestMessage::User(
+        async_openai::types::ChatCompletionRequestUserMessage {
+            content: async_openai::types::ChatCompletionRequestUserMessageContent::Text(
+                "This will take a long time".to_string(),
+            ),
+            name: None,
+        },
+    );
+
+    let request = async_openai::types::CreateChatCompletionRequestArgs::default()
+        .model("slow-model")
+        .messages(vec![message])
+        .stream(false) // Test unary response
+        .build()
+        .expect("Failed to build request");
+
+    // Start the request and cancel it after 1 second
+    let start_time = std::time::Instant::now();
+
+    let request_future = async {
+        client
+            .post("http://localhost:8993/v1/chat/completions")
+            .json(&request)
+            .send()
+            .await
+    };
+
+    // Use timeout to simulate client disconnect after 1 second
+    let result = timeout(std::time::Duration::from_millis(1000), request_future).await;
+
+    let elapsed = start_time.elapsed();
+
+    // The request should timeout (simulating client disconnect)
+    assert!(result.is_err(), "Request should have timed out");
+
+    // Give the service a moment to detect the disconnect and propagate cancellation
+    tokio::time::sleep(std::time::Duration::from_millis(500)).await;
+
+    // Verify the engine was cancelled
+    assert!(
+        long_running_engine.was_cancelled(),
+        "Engine should have been cancelled due to client disconnect"
+    );
+
+    // Verify cancellation happened quickly (within 2 seconds, not the full 10 seconds)
+    assert!(
+        elapsed < std::time::Duration::from_secs(2),
+        "Cancellation should have propagated quickly, took {:?}",
+        elapsed
+    );
+
+    tracing::info!(
+        "✅ Client disconnect test passed! Request cancelled in {:?}, engine detected cancellation",
+        elapsed
+    );
+
+    cancel_token.cancel();
+    task.await.unwrap().unwrap();
+}
+
+#[rstest]
+#[tokio::test]
+async fn test_client_disconnect_cancellation_streaming() {
+    dynamo_runtime::logging::init();
+
+    let service = HttpService::builder().port(8994).build().unwrap();
+    let state = service.state_clone();
+    let manager = state.manager();
+
+    let token = CancellationToken::new();
+    let cancel_token = token.clone();
+
+    // Start the service
+    let task = tokio::spawn(async move { service.run(token).await });
+
+    // Wait for service to be ready
+    wait_for_service_ready(8994).await;
+
+    // Create a long-running engine (10 seconds)
+    let long_running_engine = Arc::new(LongRunningEngine::new(10_000));
+    manager
+        .add_chat_completions_model("slow-stream-model", long_running_engine.clone())
+        .unwrap();
+
+    let client = reqwest::Client::new();
+
+    let message = async_openai::types::ChatCompletionRequestMessage::User(
+        async_openai::types::ChatCompletionRequestUserMessage {
+            content: async_openai::types::ChatCompletionRequestUserMessageContent::Text(
+                "This will stream for a long time".to_string(),
+            ),
+            name: None,
+        },
+    );
+
+    let request = async_openai::types::CreateChatCompletionRequestArgs::default()
+        .model("slow-stream-model")
+        .messages(vec![message])
+        .stream(true) // Test streaming response
+        .build()
+        .expect("Failed to build request");
+
+    // Start the request and cancel it after 1 second
+    let start_time = std::time::Instant::now();
+
+    let request_future = async {
+        let response = client
+            .post("http://localhost:8994/v1/chat/completions")
+            .json(&request)
+            .send()
+            .await
+            .unwrap();
+
+        // Start reading the stream, then drop it to simulate client disconnect
+        let mut stream = response.bytes_stream();
+        tokio::time::sleep(std::time::Duration::from_millis(500)).await;
+
+        // Read one chunk then drop the stream (simulating client disconnect)
+        let _ = StreamExt::next(&mut stream).await;
+        // Stream gets dropped here when function exits
+    };
+
+    // Use timeout to simulate the streaming request timing out
+    let _result = timeout(std::time::Duration::from_millis(1500), request_future).await;
+
+    let elapsed = start_time.elapsed();
+
+    // Give the service time to detect the disconnect
+    tokio::time::sleep(std::time::Duration::from_millis(1000)).await;
+
+    // Verify the engine was cancelled
+    assert!(
+        long_running_engine.was_cancelled(),
+        "Engine should have been cancelled due to streaming client disconnect"
+    );
+
+    // Verify cancellation happened reasonably quickly
+    assert!(
+        elapsed < std::time::Duration::from_secs(3),
+        "Stream cancellation should have propagated reasonably quickly, took {:?}",
+        elapsed
+    );
+
+    tracing::info!(
+        "✅ Streaming client disconnect test passed! Stream cancelled in {:?}, engine detected cancellation",
+        elapsed
+    );
+
+    cancel_token.cancel();
+    task.await.unwrap().unwrap();
+}
+
+#[rstest]
+#[tokio::test]
+async fn test_request_id_annotation() {
+    // TODO(ryan): make better fixtures, this is too much to test sometime so simple
+    dynamo_runtime::logging::init();
+
+    let service = HttpService::builder().port(8995).build().unwrap();
+    let state = service.state_clone();
+    let manager = state.manager();
+
+    let token = CancellationToken::new();
+    let cancel_token = token.clone();
+
+    // Start the service
+    let task = tokio::spawn(async move { service.run(token).await });
+
+    // Wait for service to be ready
+    wait_for_service_ready(8995).await;
+
+    // Add a counter engine for this test
+    let counter_engine = Arc::new(CounterEngine {});
+    manager
+        .add_chat_completions_model("test-model", counter_engine)
+        .unwrap();
+
+    // Create reqwest client directly
+    let client = reqwest::Client::new();
+
+    // Generate a UUID for the request ID
+    let request_uuid = uuid::Uuid::new_v4();
+
+    // Create the request JSON directly
+    let request_json = serde_json::json!({
+        "model": "test-model",
+        "messages": [
+            {
+                "role": "user",
+                "content": "Test request with annotation"
+            }
+        ],
+        "stream": true,
+        "max_tokens": 50,
+        "nvext": {
+            "annotations": ["request_id"]
+        }
+    });
+
+    // Make the streaming request with custom header
+    let response = client
+        .post("http://localhost:8995/v1/chat/completions")
+        .header("x-dynamo-request-id", request_uuid.to_string())
+        .json(&request_json)
+        .send()
+        .await
+        .expect("Request should succeed");
+
+    assert!(
+        response.status().is_success(),
+        "Response should be successful"
+    );
+
+    // Collect the entire response body as bytes first
+    let body_bytes = response
+        .bytes()
+        .await
+        .expect("Failed to read response body");
+    let body_text = String::from_utf8_lossy(&body_bytes);
+
+    // Create a cursor from the text and use SseLineCodec to parse it
+    let cursor = Cursor::new(body_text.to_string());
+    let framed = FramedRead::new(cursor, SseLineCodec::new());
+    let annotated_stream = convert_sse_stream::<NvCreateChatCompletionStreamResponse>(framed);
+
+    // Look for the annotation in the stream
+    let mut found_request_id_annotation = false;
+    let mut received_request_id = None;
+
+    // Process the annotated stream and look for the request_id annotation
+    let mut annotated_stream = std::pin::pin!(annotated_stream);
+    while let Some(annotated_response) = annotated_stream.next().await {
+        // Check if this is a request_id annotation
+        if let Some(event) = &annotated_response.event {
+            if event == "request_id" {
+                found_request_id_annotation = true;
+                // Extract the request ID from the annotation
+                if let Some(comments) = &annotated_response.comment {
+                    if let Some(comment) = comments.first() {
+                        // The comment contains a JSON-encoded string, so we need to parse it
+                        if let Ok(parsed_value) = serde_json::from_str::<String>(comment) {
+                            received_request_id = Some(parsed_value);
+                        } else {
+                            // Fallback: remove quotes manually if JSON parsing fails
+                            received_request_id = Some(comment.trim_matches('"').to_string());
+                        }
+                    }
+                }
+                break;
+            }
+        }
+    }
+
+    // Verify we found the annotation
+    assert!(
+        found_request_id_annotation,
+        "Should have received request_id annotation in the stream"
+    );
+
+    // Verify the request ID matches what we sent
+    assert!(
+        received_request_id.is_some(),
+        "Should have received the request ID in the annotation"
+    );
+
+    let received_uuid_str = received_request_id.unwrap();
+    assert_eq!(
+        received_uuid_str,
+        request_uuid.to_string(),
+        "Received request ID should match the one we sent: expected {}, got {}",
+        request_uuid,
+        received_uuid_str
+    );
+
+    tracing::info!(
+        "✅ Request ID annotation test passed! Sent UUID: {}, Received: {}",
+        request_uuid,
+        received_uuid_str
+    );
+
+    cancel_token.cancel();
+    task.await.unwrap().unwrap();
+}

lib/runtime/src/pipeline/context.rs

@@ -75,10 +75,16 @@ impl<T: Send + Sync + 'static> Context<T> {
         }
     }
 
+    /// Get the id of the context
     pub fn id(&self) -> &str {
         self.controller.id()
     }
 
+    /// Get the content of the context
+    pub fn content(&self) -> &T {
+        &self.current
+    }
+
     pub fn controller(&self) -> &Controller {
         &self.controller
     }