Simple and flexible pipeline system for Python
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


This small library implements a basic Kahn process network system. By wrapping functions in Kahn objects, you can link the return of one into the parameters of another you form the network.

The linked processes can be called, flowing data through the network in a buffered way (a queue is put between the output and input). This allows the network to process data by simple pipelining with parallel implementations.


# Declare the processes.
a = lambda: 4
b = lambda: 5
add = lambda x, y: x + y

# Link the processes and start the network.
connect((a, add), (b, add), (add, print))

Why use a Kahn process network?

  • Simplifies the individual processes, using regular parameters and return values, especially helping the tests.

  • The network state is completely defined by the items in the queues, making debugging much easier.

  • There are more levels of parallelism available, both between processes and process instances.

  • Determinism guaranteed, regardless of concurrency, as long as the processes are pure.