The Parma-Light Platform

This git repository contains an implementation of the Parma-Light platform. Parma-Light is a simple, experimental implementation of the platform published in

Pintz, M., Becker, D., Mock, M. (2024). PARMA: a Platform Architecture to enable Automated, Reproducible, and Multi-party Assessments of AI Trustworthiness. In Proceedings of the 2nd International Workshop on Responsible AI Engineering (pp. 20–27).

Licence

Copyright 2025 Fraunhofer-Institut für Intelligente Analyse- und Informationssysteme, Schloss Birlinghoven 1, 53757 Sankt Augustin

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

The SBOM togehther with potential vulnerabilities and license information are stored in files

• audit.json (currently one vulnerability of pip:25.2 is known, that will not affect this software)
• license.md and license.json

Introduction

Clone this repository from github, if you want to experiment with the platform and its sources.

• the branch master contains the current published version
• the branch develop contains a stable version, which will become master soon.
• the commands of the command language interpreter (CLI), which is a front-end for the platform, are described in resources/APIDOC.md
• the REST-endpoints of the platform backend are described in the same document.

Parma_light can be run on a local system (see next section) and packaged into a docker image and run in a container (see overnext section)

Using the Platform on a local system from the git repository

• a recent Python3, uv and docker installation are required. You need a shell that can run bash scripts.
• start the docker demon
• uv sync - install dependencies
• activate the virtual environment generated by uv (this is done for you, if you use Vscode and start a terminal):
  • Windows: source .venv/Scripts/activate
  • Linux/MacOS: source .venv/bin/activate
• pytest test - check that the setup is fine
• run the admin script ./admin.sh --help. If you get a "permission denied" error, you need to make the admin script executable: chmod +x admin.sh

The setup is complete now. You can start to work with the platform

• ./admin.sh --images sklearn - you need Docker images to work with. Generate them
• ./admin.sh --init - initialize the data base
• ./admin.sh --backend
• ./admin.sh --frontend - start the CLI in another shell and activate the virtual environment before the start
• in the CLI execute commands as:
  • test_data example:sklearn; - to load test data. This takes some time :-). The command executed can be found in the file example/sklearn/demo.txt. Some nodes and workflows are defined. Two workflows are run.
  • login sklearn; - you must be logged in to execute commands
  • view workflow; or view run; to see results of loading the test data.
  • view data_of run_predict_and_metric latest; or view log_of run_predict_and_metric latest; or cat result_file latest; to see generated data or logging
• in directories test/test_cmds/** a lot of node and workflow definitions are available to experiment with. As already said, CLI commands are described in resources/APIDOC.md.
• Terminate the frontend with the command quit;. Then terminate the backend by hitting CTRL-C.

Further Notes

The implementation is IP of the department AAA of the Fraunhofer Institute IAIS and the Fraunhofer Institute IAIS.

For ideas, proposals and errors: contact reinhard.budde AT iais.fraunhofer.de