if (!status->status.ok()) {
    delete lib_handle;
    return nullptr;
  }
  return lib_handle;
}

TF_Buffer TF_GetOpList(TF_Library* lib_handle) { return lib_handle->op_list; }

void TF_DeleteLibraryHandle(TF_Library* lib_handle) {
  if (lib_handle == nullptr) return;
  tensorflow::port::Free(const_cast<void*>(lib_handle->op_list.data));
  delete lib_handle;
}

TF_Buffer* TF_GetAllOpList() {

  TFE_TensorHandle* value_handle = nullptr;
  TFE_Execute(op, &value_handle, &num_retvals, status);
  TFE_DeleteOp(op);

  ASSERT_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);
  ASSERT_EQ(1, num_retvals);
  EXPECT_EQ(TF_FLOAT, TFE_TensorHandleDataType(value_handle));
  EXPECT_EQ(0, TFE_TensorHandleNumDims(value_handle, status));
  ASSERT_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);

To do this, use `yield` instead of `return`, and write the extra steps (code) after.

!!! tip
    Make sure to use `yield` one single time.

!!! note "Technical Details"
    Any function that is valid to use with:

Andreas Auernhammer <******@****.***> 1658195667 +0200

Until a connection is established, all outgoing requests will return `ErrDisconnected`.

# Usage

## Single Payload Requests

Single payload requests are requests and responses that are sent in a single message.
In essence, they are `[]byte` -> `[]byte, error` functions.

It is not possible to return *both* an error and a response.

	// There is flow control in both directions.
	StreamHandler struct {
		// Handle an incoming request. Initial payload is sent.
		// Additional input packets (if any) are streamed to request.
		// Upstream will block when request channel is full.
		// Response packets can be sent at any time.
		// Any non-nil error sent as response means no more responses are sent.
		Handle StreamHandlerFn

		// Subroute for handler.

* **Cloud services** that handle this for you
    * The cloud service will probably **handle replication for you**. It would possibly let you define **a process to run**, or a **container image** to use, in any case, it would most probably be **a single Uvicorn process**, and the cloud service would be in charge of replicating it.

!!! tip

#### Docker Compose

You could be deploying to a **single server** (not a cluster) with **Docker Compose**, so you wouldn't have an easy way to manage replication of containers (with Docker Compose) while preserving the shared network and **load balancing**.

  // indication of the dtype of the variable's value.
  //
  // Note that creating this resource-dtype handle can fail, so `Create` is a
  // separate static method which returns a status.
  Variable(TFE_TensorHandle* handle, TF_DataType type)
      : handle_(handle), type_(type) {}

  // Helper for constructing a resource handle and wrapping it in a `Variable`
  // object.
  static Variable* Create(TFE_Context* context, TF_DataType type,

You can probably right-click each link and select an option similar to `Save link as...`.

**Swagger UI** uses the files:

* <a href="https://cdn.jsdelivr.net/npm/swagger-ui-dist@5.9.0/swagger-ui-bundle.js" class="external-link" target="_blank">`swagger-ui-bundle.js`</a>
* <a href="https://cdn.jsdelivr.net/npm/swagger-ui-dist@5.9.0/swagger-ui.css" class="external-link" target="_blank">`swagger-ui.css`</a>

And **ReDoc** uses the file: