cornflow
cornflow is an open source multi-solver optimization server with a REST API built using flask, airflow and pulp.
While most deployment servers are based on the solving technique (MIP, CP, NLP, etc.), cornflow focuses on the optimization problems themselves. However, it does not impose any constraint on the type of problem and solution method to use.
With cornflow you can deploy a Traveling Salesman Problem solver next to a Knapsack solver or a Nurse Rostering Problem solver. As long as you describe the input and output data, you can upload any solution method for any problem and then use it with any data you want.
cornflow helps you formalize your problem by proposing development guidelines. It also provides a range of functionalities around your deployed solution method, namely:
storage of users, instances, solutions and solution logs.
deployment and maintenance of models, solvers and algorithms.
scheduling of executions in remote machines.
management of said executions: start, monitor, interrupt.
centralizing of commercial licenses.
scenario storage and comparison.
user management, roles and groups.
Installation instructions
cornflow is tested with Ubuntu 20.04, python >= 3.8 and git.
Download the cornflow project and install requirements:
python3 -m venv venv venv/bin/pip3 install cornflow
initialize the sqlite database:
source venv/bin/activate export FLASK_APP=cornflow.app export DATABASE_URL=sqlite:///cornflow.db flask db upgrade flask access_init flask create_service_user -u airflow -e airflow_test@admin.com -p airflow_test_password flask create_admin_user -u cornflow -e cornflow_admin@admin.com -p cornflow_admin_password
activate the virtual environment and run cornflow:
source venv/bin/activate export FLASK_APP=cornflow.app export SECRET_KEY=THISNEEDSTOBECHANGED export DATABASE_URL=sqlite:///cornflow.db export AIRFLOW_URL=http://127.0.0.1:8080/ export AIRFLOW_USER=airflow_user export AIRFLOW_PWD=airflow_pwd flask run
cornflow needs a running installation of Airflow to operate and more configuration. Check the installation docs for more details on installing airflow, configuring the application and initializing the database.
Using cornflow to solve a PuLP model
We’re going to test the cornflow server by using the cornflow-client and the pulp python package:
pip install cornflow-client pulp
Initialize the api client:
from cornflow_client import CornFlow email = 'some_email@gmail.com' pwd = 'Some_password1' username = 'some_name' client = CornFlow(url="http://127.0.0.1:5000")
Create a user:
config = dict(username=username, email=email, pwd=pwd) client.sign_up(**config)
Log in:
client.login(username=username, pwd=pwd)
Prepare an instance:
import pulp
prob = pulp.LpProblem("test_export_dict_MIP", pulp.LpMinimize)
x = pulp.LpVariable("x", 0, 4)
y = pulp.LpVariable("y", -1, 1)
z = pulp.LpVariable("z", 0, None, pulp.LpInteger)
prob += x + 4 * y + 9 * z, "obj"
prob += x + y <= 5, "c1"
prob += x + z >= 10, "c2"
prob += -y + z == 7.5, "c3"
data = prob.to_dict()
insName = 'test_export_dict_MIP'
description = 'very small example'
Send instance:
instance = client.create_instance(data, name=insName, description=description, schema="solve_model_dag",)
Solve an instance:
config = dict(
solver = "PULP_CBC_CMD",
timeLimit = 10
)
execution = client.create_execution(
instance['id'], config, name='execution1', description='execution of a very small instance',
schema="solve_model_dag",
)
Check the status of an execution:
status = client.get_status(execution["id"]) print(status['state']) # 1 means "finished correctly"
Retrieve a solution:
results = client.get_solution(execution['id']) print(results['data']) # returns a json with the solved pulp object _vars, prob = pulp.LpProblem.from_dict(results['data'])
Retrieve the log of the solver:
log = client.get_log(execution['id']) print(log['log']) # json format of the solver log
Using cornflow to deploy a solution method
To deploy a cornflow solution method, the following tasks need to be accomplished:
Create an Application for the new problem
Do a PR to a compatible repo linked to a server instance (e.g., like this one).
For more details on each part, check the deployment guide.
Using cornflow to solve a problem
For this example we only need the cornflow_client package. We will test the graph-coloring demo defined here. We will use the test server to solve it.
Initialize the api client:
from cornflow_client import CornFlow email = 'readme@gmail.com' pwd = 'some_password' username = 'some_name' client = CornFlow(url="https://devsm.cornflow.baobabsoluciones.app/") client.login(username=username, pwd=pwd)
solve a graph coloring problem and get the solution:
data = dict(pairs=[dict(n1=0, n2=1), dict(n1=1, n2=2), dict(n1=1, n2=3)])
instance = client.create_instance(data, name='gc_4_1', description='very small gc problem', schema="graph_coloring")
config = dict()
execution = client.create_execution(
instance['id'], config, name='gc_4_1_exec', description='execution of very small gc problem',
schema="graph_coloring",
)
status = client.get_status(execution["id"])
print(status['state'])
solution = client.get_solution(execution["id"])
print(solution['data']['assignment'])
Running tests and coverage
Then you have to run the following commands:
export FLASK_ENV=testing
Finally you can run all the tests with the following command:
python -m unittest discover -s cornflow.tests
If you want to only run the unit tests (without a local airflow webserver):
python -m unittest discover -s cornflow.tests.unit
If you want to only run the integration test with a local airflow webserver:
python -m unittest discover -s cornflow.tests.integration
After if you want to check the coverage report you need to run:
coverage run --source=./cornflow/ -m unittest discover -s=./cornflow/tests/ coverage report -m
or to get the html reports:
coverage html
Release files for cornflow 1.3.8
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| cornflow-1.3.8.tar.gz | 177.7 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| cornflow-1.3.8-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 432.1 kB
Release files / cornflow-1.3.8.tar.gz
| Download URL | cornflow-1.3.8.tar.gz |
|---|---|
| Size | 177.7 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
bc02fd7e5122736b45c7222c882d9f8421663aaa6734bee5715f935fbbec3bc0
|
|
BLAKE2b-256 checksum How to use checksums |
b068108771f320ef67865fc20a888e02a78b6d6313b09fed76351bc182eee872
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Release files / cornflow-1.3.8-py3-none-any.whl
| Download URL | cornflow-1.3.8-py3-none-any.whl |
|---|---|
| Size | 254.4 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
144b324fc9a1205789d86915926af4115f26fa7d2e24a048863284a15f980e0b
|
|
BLAKE2b-256 checksum How to use checksums |
4d9eb645b15ee8faf2b018df1d5b09a431659032e37b24e9f75de0d6876b5aa6
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
