ml-tooling / contaxy

MIT License
10 stars 8 forks source link

Contaxy

Collaborative, extensible platform to create projects, share files, and deploy services & jobs.

Getting StartedFeaturesExamplesSupportReport a BugContributionChangelog

Contaxy is an extensible collaborative platform for teams to share files and deploy services and jobs easily using Docker or Kubernetes.The platform is secured so only members of a project can access its resources (files, services, jobs). Its strong extension system allows to add further functionality to the backend and web app and also to override the behavior of Contaxy's core endpoints. By adding project members or generating fine-granular access tokens, resources can easily be shared with others. Deploying a service and securely sharing it with a stakeholder? No problem.

Highlights

Getting Started

Installation

Note: Currently, not all Docker images are pushed to DockerHub. So, the project has to be built locally, see the build section.

For the Docker deployment, have a look at the docker-compose.yaml file; you can start it via docker compose up. For Kubernetes, have a look at the deploy script.

Important: The configurations are not meant to be used for production as the JWT secret is the default one and the ports of all services instead of only the core service are published; change the configuration accordingly.

For a list of all configurable environment variables, have a look at the config file. All fields of the Settings class represent an environment variable that can be set.

Deploy Behind a Proxy

If you deploy Contaxy behind a proxy, you can define a base url. It must not end with a slash. For example, if the web app should be accessible behind /test/app instead of /app, the CONTAXY_BASE_URL would be defined as /test.

Use external OAuth2/OIDC Identity Provider

The application needs to be registered at the desired identity provider as a prerequisite. The IDP requires a redirect URI which is https://{CONTAXY_HOST}/api/auth/oauth/callback. For further details on how to registering the application at your IDP you should refer to it's documentation (e.g. Google). Moreover, the following env variables need to be added to contaxy:

If you want to prevent user registration via the default username / password flow, then you can set USER_REGISTRATION_ENABLED=0. You still need to initially create an admin user.

Initialize and register admin user

In order to initialze the system and create the admin user, call the endpoint /api/system/admin-form on the started instance and submit the form. This can only be done once.

Usage

After initializing, visit /app/.

Development

Start the backend locally

In the backend directory, run uvicorn --app-dir=src contaxy.main:app --reload. Set the environment variables POSTGRES_CONNECTION_URI, S3_ENDPOINT, S3_ACCESS_KEY, S3_SECRET_KEY so that the locally-started Python instance can access them.

Build

Using Act

Note: We recommend to use Docker and Act to execute the containerized build process. If this is not an option, run python build.py --make and install the dependencies based on probable errors.

To simplify the process of building this project from scratch, we provide build-scripts - based on universal-build - that run all necessary steps (build, check, test, and release) within a containerized environment. To build and test your changes, execute the following command in the project root folder:

act -b -j build

Refer to our contribution guides for more detailed information on our build scripts and development process.

Testing

The project contains unit tests, integration tests, and stress tests. For the unit and integration tests pytest is used, for the stress tests locust.

To run the unit/integration tests, you have the following options:

  1. in the root folder, execute act -j build -s BUILD_ARGS="--test"
  2. in the root folder, execute python build.py --test
  3. cd into the backend/ folder and run PYTHONPATH=$(pwd) pytest

By default, unit tests are executed (marked via pytest.mark.unit). If you want to execute the integration tests, pass --test-marker integration to options 1 or 2 or -m "integration" to option 3. This can be a valid pytest markexpression, hence it would also be possible to execute unit and integration tests by passing "unit and integration" to one of the flags. If neither unit nor integration tests should be executed, pass an unknown marker such as notests. By default, the deployment tests are executed for the DockerDeploymentManager. If you want to execute them for the KubernetesDeploymentManager, set the environment variables KUBERNETES_INTEGRATION_TESTS=True and DEPLOYMENT_MANAGER=kubernetes (for example in an .env file in the backend directory). Using option 1 (act) will create a Kubernetes cluster. If you don't use act, make sure that a Kubernetes cluster is accessible.

❗ To point to a running instance for Contaxy and execute the RemoteEndpoint tests, set the REMOTE_BACKEND_TESTS=True and REMOTE_BACKEND_ENDPOINT=$endpoint variables. For executing LocalEndpoint tests, make sure that the Postgres database and S3 storage are reachable from the locally spawned instance via the environment variables POSTGRES_CONNECTION_URI, S3_ENDPOINT, S3_ACCESS_KEY, S3_SECRET_KEY.

To run the stress tests, you have the following options:

  1. cd into the backend/ folder and execute, for example, locust -f tests/endpoint_tests/locustfile.py --host=http://localhost:8000 --headless -t1m --csv=./tests/results/locust

❗ For the stress tests, the --host flag must be set and pointing to a running Contaxy instance.