AI cockpit aims at building a human control interface to AI systems. It is supposed to cover requirements from EU's AI act. More details can be found at https://www.kicockpit.eu/. All shared components necessary for this repo can be found here. There you also can find more info about all other activities & results of this project.
This project was partly funded by the government of the federal republic of Germany. It is part of a research project aiming to keep humans in command and is organized by the Federal Ministry of Labour and Social Affairs.
Cockpit is a part of a bundle of services. For deployment options see https://github.com/starwit/ai-cockpit-deployment
Software in this repository is licensed under the AGPL-3.0 license. See license agreement for more details.
This section describes how to run this software on your machine.
The following components and tools are necessary to build & run AICockpit in your dev environment.
- Java JDK 21 or later
- Maven 3
- NodeJs (16.9.1) and NPM (8.3.2) - NodeJS Install
- Postgres (available for development via docker-compose scripts)
- using Keycloak is optional
Project needs a library from an AI cockpit Github repo and thus you need to adapt your Maven settings (usually located here ~/.m2/settings.xml). Following example of a complete settings.xml shows you necessary additions. Please note, that you need username/Github token replaced by values, from your user.
<settings xmlns="http://maven.apache.org/SETTINGS/1.0.0"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://maven.apache.org/SETTINGS/1.0.0
http://maven.apache.org/xsd/settings-1.0.0.xsd">
<activeProfiles>
<activeProfile>github</activeProfile>
</activeProfiles>
<profiles>
<profile>
<id>github</id>
<repositories>
<repository>
<id>central</id>
<url>https://repo1.maven.org/maven2</url>
</repository>
<repository>
<id>github</id>
<url>https://maven.pkg.github.com/starwit/vision-api</url>
<snapshots>
<enabled>true</enabled>
</snapshots>
</repository>
</repositories>
</profile>
</profiles>
<servers>
<server>
<id>github</id>
<username>username</username>
<password>TOKEN</password>
</server>
</servers>
</settings>In order to run a local development environment execute the following steps. Please note: all steps need to be executed from base folder of repositories.
-
go to the deployment folder and start the environment (database and keycloak) via docker-compose:
cd deployment docker compose up -
go to
webclient/appand install the frontend applications dependenciescd webclient/app npm install -
build the project
mvn clean install -P frontend
-
start project
java -jar application/target/application-0.0.1-SNAPSHOT.jar
Once all steps ran successfully application will be reachable with the following coordinates:
- application can be reached under http://localhost:8081/ai-cockpit/
- swagger under http://localhost:8081/ai-cockpit/swagger-ui/
- If you are using keycloak:
- default user/password is admin/admin
- keycloak can be reached under http://localost:8081/auth
If you want to start the Application with Minio then you need to run:
cd deployment
docker compose -f docker-compose-minio upThe Minio WebApp is available under localhost:9001 and the shell Application is available under localhost:9000. You need to upload a videos a bucket named anomalies link video files in decision data accordingly.
AI cockpit is aimed to serve in as many application domains as possible. Thus cockpit can be started such that decision and action types can be imported. This is controlled by two fields in application properties:
# setup scenario & demo data
scenario.setup=true
scenario.importFolder=experiment-setup/traffic/In folder experiment-setup you will find a growing number of demo data sets. Each folder can contain three files
- miticationtypes.json - Definition of possible action types
- decisiontypes.json - Definition of possible analysis results
- demodata.json - actual samples for observed/analyzed situtions and proposed actions
File with demo data can contain marker DATETIME as value for decision creation time. This will then on import be replaced by a date time in the past relative to import time.
AI Cockpit inserts upon first start a set of decision and action types. If you want to insert actual decisions you can use a script in folder experiment-setup. The following code snippet shows example usage:
cd experiment-setup
./fillDatabaseScript.sh http://localhost:8081/ai-cockpit/api/decision demo_setup.json 10Params are dev instance location (no auth), file with data to import and duration between inserts. Please note, that this will work only in a local development setup without authentication.
This script also allows you, to test how decisions are flowing into AI Cockpit. Note: you will have to prepare video data on a Minio instance.
For debugging, you can start the frontend separately.
cd webclient/app
npm startNPM server starts under localhost:3000/starwit/ by default
! If you are using the installation with keycloak, make sure you are logged in before first usage - just go to localhost:8081/starwit in your browser.
You can start the spring boot application in debug mode. See Spring Boot documentation for further details. The easiest way is, to use debug functionality integrated with your IDE like VS Code.
The database is available under localhost:3006
Username:starwit
Database:starwit
Password:starwitPGAdmin is recommended to access database for development purpose. It can be deployed via docker-compose file.
If you want to start your application without keycloak, you need to change spring boot profile to dev in application\src\main\resources\application.properties.
spring.profiles.active=devor define env-variable
SPRING_PROFILES_ACTIVE=devStart the database without keycloak:
cd deployment
docker compose upAI Cockpit is composed by a number of components. The following diagram shows, how those components are working together.
AI Cockpit needs a data model to collect system decisions and their resulting actions. The following class diagram shows the core data model.
