For added flexibility, it's possible to use PyVmMonitor programatically.

Some use-cases include connecting programatically to PyVmMonitor (instead of using the attach to running process) and making a profile session and sending the results to be shown in PyVmMonitor.

The main window has links with the available code-samples:

As a reference, the public API is designed around the pyvmmonitor package which has the following functions:

          
def connect(host='127.0.0.1', port=None):
    '''
    The host and port are optional. By default the host is 127.0.0.1 and the
    port is obtained by querying the current settings of the application.
    '''
          

          
def profile_method(*args, **kwargs):
    '''
    Decorator for profiling a method.

    It can be used as a decorator without arguments (which will show the results
    in the currently active pyvmmonitor ui):

    @profile_method
    def my_call():
        ...

    Or can be customized by receiving kwargs:

    @profile_method(filename='output.pstats', show_text=True, show_in_pyvmmonitor=False)
    def my_call():
        ...


    :param show_in_pyvmmonitor:
        Determines if we should send the results to pyvmmonitor. If still
        not connected, it'll connect to pyvmmonitor and then send the pstats
        for visual inspection.

    :param host:
        Defaults to 127.0.0.1

    :param port:
        Defaults to the value which is currently saved in the application
        settings.

    :param filename:
        Optional attribute. If given we'll also save the pstats results to the
        given filename.

    :param show_text:
        If given we'll also print the profile results when the profile finishes.

        Should be a str or list(str) saying how to show the results.

        I.e.: 'cumulative' or ('cumulative', 'time')

    :param show_text_rows:
        Only used if show_text is also passed (determines the number of rows to
        print).

    :param condition:
        A callable which will receive *args and **kwargs to the wrapped function and will determine
        whether the function should be profiled (if it returns True it'll be profiled, if it returns
        False it won't).
    '''