Nexus: worker, workflow-backed operations, and workflow caller

README.md

@@ -94,6 +94,7 @@ informal introduction to the features and their implementation.
         - [Heartbeating and Cancellation](#heartbeating-and-cancellation)
         - [Worker Shutdown](#worker-shutdown)
       - [Testing](#testing-1)
+    - [Nexus](#nexus)
     - [Workflow Replay](#workflow-replay)
     - [Observability](#observability)
       - [Metrics](#metrics)
@@ -1308,6 +1309,113 @@ affect calls activity code might make to functions on the `temporalio.activity`
 * `cancel()` can be invoked to simulate a cancellation of the activity
 * `worker_shutdown()` can be invoked to simulate a worker shutdown during execution of the activity
 
+
+### Nexus
+
+⚠️  **Nexus support is currently at an experimental release stage. Backwards-incompatible changes are anticipated until a stable release is announced.** ⚠️
+
+[Nexus](https://github.com/nexus-rpc/) is a synchronous RPC protocol. Arbitrary duration operations that can respond
+asynchronously are modeled on top of a set of pre-defined synchronous RPCs.
+
+Temporal supports calling Nexus operations **from a workflow**. See https://docs.temporal.io/nexus. There is no support
+currently for calling a Nexus operation from non-workflow code.
+
+To get started quickly using Nexus with Temporal, see the Python Nexus sample:
+https://github.com/temporalio/samples-python/tree/nexus/hello_nexus.
+
+
+Two types of Nexus operation are supported, each using a decorator:
+
+- `@temporalio.nexus.workflow_run_operation`: a Nexus operation that is backed by a Temporal workflow. The operation
+  handler you write will start the handler workflow and then respond with a token indicating that the handler workflow
+  is in progress. When the handler workflow completes, Temporal server will automatically deliver the result (success or
+  failure) to the caller workflow.
+- `@nexusrpc.handler.sync_operation`: an operation that responds synchronously. It may be `def` or `async def` and it
+may do network I/O, but it must respond within 10 seconds.
+
+The following steps are an overview of the [Python Nexus sample](
+https://github.com/temporalio/samples-python/tree/nexus/hello_nexus).
+
+1. Create the caller and handler namespaces, and the Nexus endpoint. For example,
+    ```
+    temporal operator namespace create --namespace my-handler-namespace
+    temporal operator namespace create --namespace my-caller-namespace
+
+    temporal operator nexus endpoint create \
+      --name my-nexus-endpoint \
+      --target-namespace my-handler-namespace \
+      --target-task-queue my-handler-task-queue
+    ```
+
+2. Define your service contract. This specifies the names and input/output types of your operations. You will use this
+   to refer to the operations when calling them from a workflow.
+    ```python
+    @nexusrpc.service
+    class MyNexusService:
+        my_sync_operation: nexusrpc.Operation[MyInput, MyOutput]
+        my_workflow_run_operation: nexusrpc.Operation[MyInput, MyOutput]
+    ```
+
+3. Implement your operation handlers in a service handler:
+    ```python
+    @service_handler(service=MyNexusService)
+    class MyNexusServiceHandler:
+        @sync_operation
+        async def my_sync_operation(
+            self, ctx: StartOperationContext, input: MyInput
+        ) -> MyOutput:
+            return MyOutput(message=f"Hello {input.name} from sync operation!")
+
+        @workflow_run_operation
+        async def my_workflow_run_operation(
+            self, ctx: WorkflowRunOperationContext, input: MyInput
+        ) -> nexus.WorkflowHandle[MyOutput]:
+            return await ctx.start_workflow(
+                WorkflowStartedByNexusOperation.run,
+                input,
+                id=str(uuid.uuid4()),
+            )
+    ```
+
+4. Register your service handler with a Temporal worker.
+    ```python
+    client = await Client.connect("localhost:7233", namespace="my-handler-namespace")
+    worker = Worker(
+        client,
+        task_queue="my-handler-task-queue",
+        workflows=[WorkflowStartedByNexusOperation],
+        nexus_service_handlers=[MyNexusServiceHandler()],
+    )
+    await worker.run()
+    ```
+
+5. Call your Nexus operations from your caller workflow.
+    ```python
+    @workflow.defn
+    class CallerWorkflow:
+        def __init__(self):
+            self.nexus_client = workflow.create_nexus_client(
+                service=MyNexusService, endpoint="my-nexus-endpoint"
+            )
+
+        @workflow.run
+        async def run(self, name: str) -> tuple[MyOutput, MyOutput]:
+            # Start the Nexus operation and wait for the result in one go, using execute_operation.
+            wf_result = await self.nexus_client.execute_operation(
+                MyNexusService.my_workflow_run_operation,
+                MyInput(name),
+            )
+            # Or alternatively, obtain the operation handle using start_operation,
+            # and then use it to get the result:
+            sync_operation_handle = await self.nexus_client.start_operation(
+                MyNexusService.my_sync_operation,
+                MyInput(name),
+            )
+            sync_result = await sync_operation_handle
+            return sync_result, wf_result
+    ```
+
+
 ### Workflow Replay
 
 Given a workflow's history, it can be replayed locally to check for things like non-determinism errors. For example,

pyproject.toml

@@ -11,6 +11,7 @@ keywords = [
     "workflow",
 ]
 dependencies = [
+    "nexus-rpc>=1.1.0",
     "protobuf>=3.20,<6",
     "python-dateutil>=2.8.2,<3 ; python_version < '3.11'",
     "types-protobuf>=3.20",
@@ -44,7 +45,7 @@ dev = [
     "psutil>=5.9.3,<6",
     "pydocstyle>=6.3.0,<7",
     "pydoctor>=24.11.1,<25",
-    "pyright==1.1.377",
+    "pyright==1.1.402",
     "pytest~=7.4",
     "pytest-asyncio>=0.21,<0.22",
     "pytest-timeout~=2.2",
@@ -53,6 +54,8 @@ dev = [
     "twine>=4.0.1,<5",
     "ruff>=0.5.0,<0.6",
     "maturin>=1.8.2",
+    "pytest-cov>=6.1.1",
+    "httpx>=0.28.1",
     "pytest-pretty>=1.3.0",
 ]
 
@@ -162,6 +165,7 @@ exclude = [
   "tests/worker/workflow_sandbox/testmodules/proto",
   "temporalio/bridge/worker.py",
   "temporalio/contrib/opentelemetry.py",
+  "temporalio/contrib/pydantic.py",
   "temporalio/converter.py",
   "temporalio/testing/_workflow.py",
   "temporalio/worker/_activity.py",
@@ -173,6 +177,10 @@ exclude = [
   "tests/api/test_grpc_stub.py",
   "tests/conftest.py",
   "tests/contrib/test_opentelemetry.py",
+  "tests/contrib/pydantic/models.py",
+  "tests/contrib/pydantic/models_2.py",
+  "tests/contrib/pydantic/test_pydantic.py",
+  "tests/contrib/pydantic/workflows.py",
   "tests/test_converter.py",
   "tests/test_service.py",
   "tests/test_workflow.py",

temporalio/bridge/src/worker.rs

@@ -20,7 +20,7 @@ use temporal_sdk_core_api::worker::{
 };
 use temporal_sdk_core_api::Worker;
 use temporal_sdk_core_protos::coresdk::workflow_completion::WorkflowActivationCompletion;
-use temporal_sdk_core_protos::coresdk::{ActivityHeartbeat, ActivityTaskCompletion};
+use temporal_sdk_core_protos::coresdk::{ActivityHeartbeat, ActivityTaskCompletion, nexus::NexusTaskCompletion};
 use temporal_sdk_core_protos::temporal::api::history::v1::History;
 use tokio::sync::mpsc::{channel, Sender};
 use tokio_stream::wrappers::ReceiverStream;
@@ -60,6 +60,7 @@ pub struct WorkerConfig {
     graceful_shutdown_period_millis: u64,
     nondeterminism_as_workflow_fail: bool,
     nondeterminism_as_workflow_fail_for_types: HashSet<String>,
+    nexus_task_poller_behavior: PollerBehavior,
 }
 
 #[derive(FromPyObject)]
@@ -565,6 +566,18 @@ impl WorkerRef {
         })
     }
 
+    fn poll_nexus_task<'p>(&self, py: Python<'p>) -> PyResult<Bound<'p, PyAny>> {
+        let worker = self.worker.as_ref().unwrap().clone();
+        self.runtime.future_into_py(py, async move {
+            let bytes = match worker.poll_nexus_task().await {
+                Ok(task) => task.encode_to_vec(),
+                Err(PollError::ShutDown) => return Err(PollShutdownError::new_err(())),
+                Err(err) => return Err(PyRuntimeError::new_err(format!("Poll failure: {err}"))),
+            };
+            Ok(bytes)
+        })
+    }
+
     fn complete_workflow_activation<'p>(
         &self,
         py: Python<'p>,
@@ -599,6 +612,22 @@ impl WorkerRef {
         })
     }
 
+    fn complete_nexus_task<'p>(&self,
+        py: Python<'p>,
+        proto: &Bound<'_, PyBytes>,
+) -> PyResult<Bound<'p, PyAny>> {
+        let worker = self.worker.as_ref().unwrap().clone();
+        let completion = NexusTaskCompletion::decode(proto.as_bytes())
+            .map_err(|err| PyValueError::new_err(format!("Invalid proto: {err}")))?;
+        self.runtime.future_into_py(py, async move {
+            worker
+                .complete_nexus_task(completion)
+                .await
+                .context("Completion failure")
+                .map_err(Into::into)
+        })
+    }
+
     fn record_activity_heartbeat(&self, proto: &Bound<'_, PyBytes>) -> PyResult<()> {
         enter_sync!(self.runtime);
         let heartbeat = ActivityHeartbeat::decode(proto.as_bytes())
@@ -696,6 +725,7 @@ fn convert_worker_config(
                 })
                 .collect::<HashMap<String, HashSet<WorkflowErrorType>>>(),
         )
+        .nexus_task_poller_behavior(conf.nexus_task_poller_behavior)
         .build()
         .map_err(|err| PyValueError::new_err(format!("Invalid worker config: {err}")))
 }

temporalio/bridge/worker.py

@@ -26,6 +26,7 @@
 import temporalio.bridge.client
 import temporalio.bridge.proto
 import temporalio.bridge.proto.activity_task
+import temporalio.bridge.proto.nexus
 import temporalio.bridge.proto.workflow_activation
 import temporalio.bridge.proto.workflow_completion
 import temporalio.bridge.runtime
@@ -35,7 +36,7 @@
 from temporalio.bridge.temporal_sdk_bridge import (
     CustomSlotSupplier as BridgeCustomSlotSupplier,
 )
-from temporalio.bridge.temporal_sdk_bridge import PollShutdownError
+from temporalio.bridge.temporal_sdk_bridge import PollShutdownError  # type: ignore
 
 
 @dataclass
@@ -60,6 +61,7 @@ class WorkerConfig:
     graceful_shutdown_period_millis: int
     nondeterminism_as_workflow_fail: bool
     nondeterminism_as_workflow_fail_for_types: Set[str]
+    nexus_task_poller_behavior: PollerBehavior
 
 
 @dataclass
@@ -216,6 +218,14 @@ async def poll_activity_task(
             await self._ref.poll_activity_task()
         )
 
+    async def poll_nexus_task(
+        self,
+    ) -> temporalio.bridge.proto.nexus.NexusTask:
+        """Poll for a nexus task."""
+        return temporalio.bridge.proto.nexus.NexusTask.FromString(
+            await self._ref.poll_nexus_task()
+        )
+
     async def complete_workflow_activation(
         self,
         comp: temporalio.bridge.proto.workflow_completion.WorkflowActivationCompletion,
@@ -229,6 +239,12 @@ async def complete_activity_task(
         """Complete an activity task."""
         await self._ref.complete_activity_task(comp.SerializeToString())
 
+    async def complete_nexus_task(
+        self, comp: temporalio.bridge.proto.nexus.NexusTaskCompletion
+    ) -> None:
+        """Complete a nexus task."""
+        await self._ref.complete_nexus_task(comp.SerializeToString())
+
     def record_activity_heartbeat(
         self, comp: temporalio.bridge.proto.ActivityHeartbeat
     ) -> None:

temporalio/client.py

@@ -464,9 +464,17 @@ async def start_workflow(
         rpc_metadata: Mapping[str, str] = {},
         rpc_timeout: Optional[timedelta] = None,
         request_eager_start: bool = False,
-        stack_level: int = 2,
         priority: temporalio.common.Priority = temporalio.common.Priority.default,
         versioning_override: Optional[temporalio.common.VersioningOverride] = None,
+        # The following options should not be considered part of the public API. They
+        # are deliberately not exposed in overloads, and are not subject to any
+        # backwards compatibility guarantees.
+        callbacks: Sequence[Callback] = [],
+        workflow_event_links: Sequence[
+            temporalio.api.common.v1.Link.WorkflowEvent
+        ] = [],
+        request_id: Optional[str] = None,
+        stack_level: int = 2,
     ) -> WorkflowHandle[Any, Any]:
         """Start a workflow and return its handle.
 
@@ -529,7 +537,6 @@ async def start_workflow(
         name, result_type_from_type_hint = (
             temporalio.workflow._Definition.get_name_and_result_type(workflow)
         )
-
         return await self._impl.start_workflow(
             StartWorkflowInput(
                 workflow=name,
@@ -557,6 +564,9 @@ async def start_workflow(
                 rpc_timeout=rpc_timeout,
                 request_eager_start=request_eager_start,
                 priority=priority,
+                callbacks=callbacks,
+                workflow_event_links=workflow_event_links,
+                request_id=request_id,
             )
         )
 
@@ -5193,6 +5203,10 @@ class StartWorkflowInput:
     rpc_timeout: Optional[timedelta]
     request_eager_start: bool
     priority: temporalio.common.Priority
+    # The following options are experimental and unstable.
+    callbacks: Sequence[Callback]
+    workflow_event_links: Sequence[temporalio.api.common.v1.Link.WorkflowEvent]
+    request_id: Optional[str]
     versioning_override: Optional[temporalio.common.VersioningOverride] = None
 
 
@@ -5807,8 +5821,30 @@ async def _build_start_workflow_execution_request(
         self, input: StartWorkflowInput
     ) -> temporalio.api.workflowservice.v1.StartWorkflowExecutionRequest:
         req = temporalio.api.workflowservice.v1.StartWorkflowExecutionRequest()
-        req.request_eager_execution = input.request_eager_start
         await self._populate_start_workflow_execution_request(req, input)
+        # _populate_start_workflow_execution_request is used for both StartWorkflowInput
+        # and UpdateWithStartStartWorkflowInput. UpdateWithStartStartWorkflowInput does
+        # not have the following two fields so they are handled here.
+        req.request_eager_execution = input.request_eager_start
+        if input.request_id:
+            req.request_id = input.request_id
+
+        links = [
+            temporalio.api.common.v1.Link(workflow_event=link)
+            for link in input.workflow_event_links
+        ]
+        req.completion_callbacks.extend(
+            temporalio.api.common.v1.Callback(
+                nexus=temporalio.api.common.v1.Callback.Nexus(
+                    url=callback.url,
+                    header=callback.headers,
+                ),
+                links=links,
+            )
+            for callback in input.callbacks
+        )
+        # Links are duplicated on request for compatibility with older server versions.
+        req.links.extend(links)
         return req
 
     async def _build_signal_with_start_workflow_execution_request(
@@ -7231,6 +7267,25 @@ def api_key(self, value: Optional[str]) -> None:
         self.service_client.update_api_key(value)
 
 
+@dataclass(frozen=True)
+class NexusCallback:
+    """Nexus callback to attach to events such as workflow completion.
+
+    .. warning::
+        This API is experimental and unstable.
+    """
+
+    url: str
+    """Callback URL."""
+
+    headers: Mapping[str, str]
+    """Header to attach to callback request."""
+
+
+# Intended to become a union of callback types
+Callback = NexusCallback
+
+
 async def _encode_user_metadata(
     converter: temporalio.converter.DataConverter,
     summary: Optional[Union[str, temporalio.api.common.v1.Payload]],