While PyPIM can be used by an application developer, it can also be used behind the scene by other PyAnsys libraries to manage remote instances of the products that they interact with. This enables an application developer to write code that works both in an environment configured with PyPIM and an environment without such a configuration.

For example, an application developer can write the following code:

from ansys.mapdl.core import launch_mapdl

mapdl = launch_mapdl()

Instead of:

import ansys.platform.instancemanagement as pypim
from ansys.mapdl.core import launch_mapdl

if pypim.is_configured():
    pim = pypim.connect()
    instance = pypim.create_instance(product_name="mapdl")
    channel = instance.build_grpc_channel(
        options=[("grpc.max_receive_message_length", 8 * 1024**2)]
    mapdl = Mapdl(channel=channel)
    mapdl = launch_mapdl()

This topic provides guidelines for implementing the launch_* method that takes PyPIM into account. Just like the entire PIM API, this topic is targeted only toward products that are stateful and require explicit lifecycle management.



PyPIM uses semantic versioning. To depend on PyPIM, a library must include the following require string:


Condition for PyPIM Usage#

The condition for using PyPIM transparently is that the user must be able to launch the product in an environment configured with PyPIM without specifying launch information. In other words, PyPIM must be the default startup method in an environment that is configured with PyPIM.

To integrate PyPIM correctly, you should use either a generic method such as launch_my_product() or a constructor that usually starts a local process.

For example, with PyMAPDL in an environment configured with PyPIM, the following code uses PyPIM:

from ansys.mapdl.core import launch_mapdl

mapdl = launch_mapdl()

However, this code does not use PyPIM:

from ansys.mapdl.core import launch_mapdl

mapdl = launch_mapdl(exec_file="/usr/bin/mapdl")

Starting a gRPC Product#

To use PyPIM, the code flow should first check if PyPIM is configured with is_configured() and then check to ensure that the user has not specified how to launch it.

If both conditions are met:

  1. Connect to PyPIM with connect()

  2. Create an instance with Client.create_instance()

  3. Wait for the instance to be ready with Instance.wait_for_ready()

  4. Build a gRPC channel with Instance.build_grpc_channel()

Typically, the resulting code looks like this:

import ansys.platform.instancemanagement as pypim

def launch_my_product(self, ...):
    if pypim.is_configured() and not user_has_specified_how_to_launch_the_product:
        pim = pypim.connect()
        self.instance = pypim.create_instance("my_product_name")
        channel = self.instance.build_grpc_channel()
        # usual start-up
        self.process =
        channel = grpc.insecure_chanel(...)

When stopping the product, ensure that the remote instance is deleted:

def stop(self):
    if self.instance is not None:


While it is PyPIM’s responsibility to clean up any resource and process associated with the product, relevant product-specific cleanup can still be performed.

Starting a Non-gRPC Product#

While the code flow for a non-gRPC product is the same, connection information is more specific.

  • For a REST-ful product, the base uniform resource identifier (URI) must be found under["http"].uri and all requests must include the headers in["http"].headers.

  • For other protocols, an agreement between the PIM implementation and the client code determines how to pass the required information in a dedicated entry in .services.


When testing the PyPIM integration, you should not rely on an actual PIM implementation. Instead, you should mock the interaction with PyPIM. Verifying that a specific PIM implementation is able to start and provide an endpoint to the product in a specific environment is the responsibility of the team managing this environment.

This test approach mocks PyPIM behavior, resulting in a verbose test with no additional dependencies. It is also not subject to bugs in PyPIM or in a PIM implementation.

The initial setup of such a mock can look like this:

from unittest.mock import create_autospec
import grpc
import ansys.platform.instancemanagement as pypim

def test_pim(monkeypatch):
    # Start the product
    product = launch_my_product(port=50052)

    # Create a mock PyPIM instance object representing the running product
    mock_instance = pypim.Instance(
        services={"grpc": pypim.Service(uri="localhost:50052", headers={})},

    # Create a working gRPC channel to this product
    pim_channel = grpc.insecure_channel("localhost:50052")

    # Mock the wait_for_ready method so that it immediately returns
    mock_instance.wait_for_ready = create_autospec(mock_instance.wait_for_ready)

    # Mock the `build_grpc_channel` to return the working channel
    mock_instance.build_grpc_channel = create_autospec(
        mock_instance.build_grpc_channel, return_value=pim_channel

    # Mock the deletion method
    mock_instance.delete = create_autospec(mock_instance.delete)

    # Mock the PyPIM client so that on the "create_instance" call it returns the mock instance
    # Note: the host and port here will not be used.
    mock_client = pypim.Client(channel=grpc.insecure_channel("localhost:12345"))
    mock_client.create_instance = create_autospec(
        mock_client.create_instance, return_value=mock_instance

    # Mock the general PyPIM connection and configuration check method to expose the mock client.
    mock_connect = create_autospec(pypim.connect, return_value=mock_client)
    mock_is_configured = create_autospec(pypim.is_configured, return_value=True)
    monkeypatch.setattr(pypim, "connect", mock_connect)
    monkeypatch.setattr(pypim, "is_configured", mock_is_configured)

This initial setup is faking all the necessary parts of PyPIM. From here, calling launch_my_product() with no parameter is expected to call only the mocks, which the test should now do:

my_product = launch_my_product()

After this call, the test is ready to make all the assertions verifying that the PyPIM workflow was applied:

# The launch method checked if it was in a PyPIM environment
assert mock_is_configured.called

# It connected to PyPIM
assert mock_connect.called

# It created a remote instance through PyPIM
    product_name="my_product_name", product_version=None

# It waited for this instance to be ready
assert mock_instance.wait_for_ready.called

# It created an gRPC channel from this instance
assert mock_instance.build_grpc_channel.called

# It connected using the channel created by PyPIM
assert my_product._channel == pim_channel

When stopping the product, the test should also verify that the remote instance is deleted:

# Stop the product

assert mock_instance.delete.called


An example of such an integration can be seen in PyMAPDL.