project-anomaly-detection

build and run

1. create a .env inside this projects root folder with following content:

AWS_ACCESS_KEY_ID=minio
AWS_ACCESS_KEY=minio_password
MYSQL_DATABASE=mlflow_database
MYSQL_USER=mlflow
MYSQL_PASSWORD=mlflow_password
MYSQL_ROOT_PASSWORD=mysql_root_password

2. run docker compose up -d in root folder to start the project as a new docker container
   • if docker compose errors out with error getting credentials - err: exec: "docker-credential-desktop": executable file not found in %PATH%, out: '' -> change credsStore to credStore in %USERPROFILE%/.docker/config.json on windows or $HOME/.docker/config.json on linux
   • depending on OS or distribution docker compose is not a known command and/or aliased to docker-compose
3. services available
   • MLflow on port 5000
   • minIO on port 9000/9001 -> user=minio, password=minio_password
   • MYSQL database on port 3306 -> user=mysql, password=mysql_password, db=mlflow_database, root_password=mysql_root_password
4. log model into MLflow and generate Docker image for inference server
   • cd into ./src/model
   • install dependencies with: pip install -r requirements.txt
   • log model into MLflow and build inference server: python ./model.py
   • run docker run -p 5001:8080 -e DISABLE_NGINX=true "anomaly-detection-model" to create a service running on port 5001 with the built inference server
5. test inference server
   • cd into ./src/model
   • run the test-service: python ./simulate.py