Skip to content

Latest commit

 

History

History
80 lines (59 loc) · 3.29 KB

README.md

File metadata and controls

80 lines (59 loc) · 3.29 KB

Nextmv OR-Tools Python template

This template demonstrates how to solve a minimum cost flow problem using the open source software suite OR-Tools using the integer optimzation interface.

In the minimum cost flow problem, the goal is to find the flow with the smallest total costs on a network of nodes and edges between those nodes.

There are nodes with a demand of a certain material. In order to fulfill these demands, other nodes offer a supply of this material. The edges in the network can be used to transport materials from supply nodes to demand nodes. Transporting on an edge incurs some cost. While solving the given minimum cost flow problem, the solver finds the flows of materials between nodes that leads to the least total cost.

The input defines the nodes of the network with their supply (negative supply equals a demand) as well as the edges between nodes. The graph doesn't have to be a complete graph so not every node is connected to every other node. Furthermore, the input contains information about the capacities of edges as well as the unit costs to utilize an edge for transportation.

In this case, we're assigning workers to projects. Projects have a demand for time units and require certain skills. Workers can offer time units and their skills. Skills between workers and projects have to match in order to being able to assign a worker to a project. Furthermore, projects have a value associated with it (e.g. the contract value of that project).

The most important files created are main.py and input.json.

  • main.py implements a minimum cost flow solver.
  • input.json is a sample input file.

Follow these steps to run locally.

  1. The packages listed in the requirements.txt will get bundled with the app as defined in the app.yaml manifest. When working locally, make sure that these are installed as well:

    pip3 install -r requirements.txt

  2. Run the command below to check that everything works as expected:

    python3 main.py -input input.json -output output.json

  3. A file output.json should have been created with the min cost flow.

Mirror running on Nextmv Cloud locally

Pre-requisites: Docker needs to be installed.

To run the application locally in the same docker image as the one used on the Nextmv Cloud, you can use the following command:

cat input.json | docker run -i --rm \
-v $(pwd):/app ghcr.io/nextmv-io/runtime/python:3.11 \
sh -c 'pip install -r requirements.txt > /dev/null && python3 /app/main.py'

You can also debug the application by running it in a Dev Container. This workspace recommends to install the Dev Container extension for VSCode. If you have the extension installed, you can open the workspace in a container by using the command Dev Containers: Reopen in Container.

Next steps

  • Open main.py and read through the comments to understand the model.
  • Further documentation, guides, and API references about custom modeling and deployment can also be found on our blog and on our documentation site.
  • Need more assistance? Send us an email!