Add Executors#

A Flow orchestrates its Executors as a graph and sends requests to all Executors in the order specified by add() or listed in a YAML file.

When you start a Flow, Executors always run in separate processes. Multiple Executors run in different processes. Multiprocessing is the lowest level of separation when you run a Flow locally. When running a Flow on Kubernetes, Docker Swarm, Jina AI Cloud Hosting, different Executors run in different containers, pods or instances.

Add Executors#

Executors can be added into a Flow with add().

from jina import Flow

f = Flow().add()

This adds an “empty” Executor called BaseExecutor to the Flow. This Executor (without any parameters) performs no actions.


To more easily identify an Executor, you can change its name by passing the name parameter:

from jina import Flow

f = Flow().add(name='myVeryFirstExecutor').add(name='secondIsBest')

You can also define the above Flow in YAML:

jtype: Flow
  - name: myVeryFirstExecutor
  - name: secondIsBest

Save it as flow.yml and run it:

jina flow --uses flow.yml

More Flow YAML specifications can be found in Flow YAML Specification.

Define Executor types with uses#

An Executor’s type is defined by the uses keyword. Note that some usages are not supported on JCloud due to security reasons and the nature of facilitating local debugging.

Local Dev





Use ExecutorClass from the inline context.


Use ExecutorClass from my.py_modules.


Use an Executor from a YAML file defined by Executor YAML interface.


Use an Executor as Python source from Executor Hub.


Use an Executor as a Docker container from Executor Hub.


Use a Sandbox Executor hosted on Executor Hub. The Executor runs remotely on Executor Hub.


Use a pre-built Executor as a Docker container.

Hint: Load multiple Executors from the same directory

You don’t need to specify the parent directory for each Executor. Instead, you can configure a common search path for all Executors:

├── app
│   └── ▶
└── executor
    ├── config1.yml
    ├── config2.yml
f = Flow(extra_search_paths=['../executor']).add(uses='config1.yml').add(uses='config2.yml')

External Executors#

How-To chapter

For a more detailed look at this feature, see our how-to on external Executors. The also covers launching an Executor that can then be used as an External Executor in a Flow.

Usually a Flow starts and stops all of its own Executors. However, external Executors are started and stopped by other Flows, meaning they can reside on any machine.

This is useful for sharing expensive Executors (like stateless, GPU-based encoders) between Flows.

Both served and shared Executors can be used as external Executors.

When you add an external Executor to a Flow, you have to provide a host and port, and enable the external flag:

from jina import Flow

Flow().add(host='', port=12345, external=True)

The Flow doesn’t start or stop this Executor and assumes that it is externally managed and available at

You can also use external Executors with tls:

from jina import Flow

Flow().add(host='', port=443, external=True, tls=True)


Using tls to connect to the External Executor is especially needed to use an external Executor deployed with JCloud. See the JCloud documentation for further details

You can also pass the grpc_metadata parameter to the Executor. grpc_metadata is a dictionary of key-value pairs to be passed along with the gRPC request. An example is the authorization metadata, which is used to authenticate the request to the external Executor:

from jina import Flow

    grpc_metadata={'authorization': '<TOKEN>'},


The grpc_metadata parameter here is the same as the metadata concept in gRPC. See the gRPC documentation for further details.

Floating Executors#

Some Executors in your Flow can be used for asynchronous background tasks that take time and don’t generate a required output. For instance, logging specific information in external services, storing partial results, etc.

You can unblock your Flow from such tasks by using floating Executors.

Normally, all Executors form a pipeline that handles and transforms a given request until it is finally returned to the Client.

However, floating Executors do not feed their outputs back into the pipeline. Therefore, the Executor’s output does not affect the response for the Client, and the response can be returned without waiting for the floating Executor to complete its task.

Those Executors are marked with the floating keyword when added to a Flow:

import time
from jina import Flow, Executor, requests, DocumentArray

class FastChangingExecutor(Executor):
    def foo(self, docs, **kwargs):
        for doc in docs:
            doc.text = 'Hello World'

class SlowChangingExecutor(Executor):
    def foo(self, docs, **kwargs):
        print(f' Received {docs.texts}')
        for doc in docs:
            doc.text = 'Change the document but will not affect response'

f = (
    .add(name='executor0', uses=FastChangingExecutor)
with f:'/endpoint', inputs=DocumentArray.empty(1))  # we need to send a first
    start_time = time.time()
    response ='/endpoint', inputs=DocumentArray.empty(2))
    end_time = time.time()
    print(f' Response time took {end_time - start_time}s')
    print(f' {response.texts}')
 Response time took 0.011997222900390625s
 ['Hello World', 'Hello World']
 Received ['Hello World', 'Hello World']

In this example the response is returned without waiting for the floating Executor to complete. However, the Flow is not closed until the floating Executor has handled the request.

You can plot the Flow and see the Executor is floating disconnected from the Gateway.


A floating Executor can never come before a non-floating Executor in your Flow’s topology.

This leads to the following behaviors:

  • Implicit reordering: When you add a non-floating Executor after a floating Executor without specifying its needs parameter, the non-floating Executor is chained after the previous non-floating one.

from jina import Flow

f = Flow().add().add(name='middle', floating=True).add()
  • Chaining floating Executors: To chain more than one floating Executor, you need to add all of them with the floating flag, and explicitly specify the needs argument.

from jina import Flow

f = Flow().add().add(name='middle', floating=True).add(needs=['middle'], floating=True)
  • Overriding the floating flag: If you add a floating Executor as part of needs parameter of a non-floating Executor, then the floating Executor is no longer considered floating.

from jina import Flow

f = Flow().add().add(name='middle', floating=True).add(needs=['middle'])

Configure Executors#

You can set and override Executor configuration when adding them to a Flow.

This example shows how to start a Flow with an Executor using the Python API:

from jina import Flow

with Flow().add(
    uses_with={"parameter_1": "foo", "parameter_2": "bar"},
        "name": "MyExecutor",
        "description": "MyExecutor does a thing to the stuff in your Documents",
    uses_requests={"/index": "my_index", "/search": "my_search", "/random": "foo"},
) as f:
  • uses_with is a key-value map that defines the arguments of the Executor’ __init__ method.

  • uses_requests is a key-value map that defines the mapping from endpoint to class method. This is useful to overwrite the default endpoint-to-method mapping defined in the Executor python implementation.

  • workspace is a string that defines the workspace.

  • py_modules is a list of strings that defines the Executor’s Python dependencies;

  • uses_metas is a key-value map that defines some of the Executor’s internal attributes. It contains the following fields:

    • name is a string that defines the name of the Executor;

    • description is a string that defines the description of this Executor. It is used in the automatic docs UI;

Set with via uses_with#

To set/override an Executor’s with configuration, use uses_with. The with configuration refers to user-defined constructor kwargs.

from jina import Executor, requests, Flow

class MyExecutor(Executor):
    def __init__(self, param1=1, param2=2, param3=3, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self.param1 = param1
        self.param2 = param2
        self.param3 = param3

    def foo(self, docs, **kwargs):
        print('param1:', self.param1)
        print('param2:', self.param2)
        print('param3:', self.param3)

flow = Flow().add(uses=MyExecutor, uses_with={'param1': 10, 'param3': 30})
with flow as f:'/')
      [email protected][L]:ready and listening
        [email protected][L]:ready and listening
           [email protected][I]:🎉 Flow is ready to use!
	🔗 Protocol: 		GRPC
	🏠 Local access:
	🔒 Private network:
	🌐 Public address:
param1: 10
param2: 2
param3: 30

Set requests via uses_requests#

You can set/override an Executor’s requests configuration and bind methods to custom endpoints. In the following code:

  • We replace the endpoint /foo bound to the foo() function with both /non_foo and /alias_foo.

  • We add a new endpoint /bar for binding bar().

Note the all_req() function is bound to all endpoints except those explicitly bound to other functions, i.e. /non_foo, /alias_foo and /bar.

from jina import Executor, requests, Flow

class MyExecutor(Executor):
    def all_req(self, parameters, **kwargs):
        print(f'all req {parameters.get("recipient")}')

    def foo(self, parameters, **kwargs):
        print(f'foo {parameters.get("recipient")}')

    def bar(self, parameters, **kwargs):
        print(f'bar {parameters.get("recipient")}')

flow = Flow().add(
        '/bar': 'bar',
        '/non_foo': 'foo',
        '/alias_foo': 'foo',
with flow as f:'/bar', parameters={'recipient': 'bar()'})'/non_foo', parameters={'recipient': 'foo()'})'/foo', parameters={'recipient': 'all_req()'})'/alias_foo', parameters={'recipient': 'foo()'})
      [email protected][L]:ready and listening
        [email protected][L]:ready and listening
           [email protected][I]:🎉 Flow is ready to use!
	🔗 Protocol: 		GRPC
	🏠 Local access:
	🔒 Private network:
	🌐 Public address:
bar bar()
foo foo()
all req all_req()
foo foo()

Set metas via uses_metas#

To set/override an Executor’s metas configuration, use uses_metas:

from jina import Executor, requests, Flow

class MyExecutor(Executor):
    def foo(self, docs, **kwargs):

flow = Flow().add(
    uses_metas={'name': 'different_name'},
with flow as f:'/')
      [email protected][L]:ready and listening
        [email protected][L]:ready and listening
           [email protected][I]:🎉 Flow is ready to use!
	🔗 Protocol: 		GRPC
	🏠 Local access:
	🔒 Private network:

Unify output ndarray types#

Different Executors in a Flow may depend on different types for array-like data such as doc.tensor and doc.embedding, often because they were written with different machine learning frameworks. As the builder of a Flow you don’t always have control over this, for example when using Executors from Executor Hub.

To ease the integration of different Executors, a Flow allows you to convert tensor and embedding by using the f.add(..., output_array_type=..):

from jina import Flow

f = Flow().add(uses=MyExecutor, output_array_type='numpy').add(uses=NeedsNumpyExecutor)

This converts the .tensor and .embedding fields of all output Documents of MyExecutor to numpy.ndarray, making the data usable by NeedsNumpyExecutor. This works whether MyExecutor populates these fields with arrays/tensors from PyTorch, TensorFlow, or any other popular ML framework.

Output types

output_array_type= supports more types than 'numpy'. For the full specification and further details, check the protobuf serialization docs.