How to start and configure server

Setup local development workspace

1. Modify development.env that meets your requirements
2. By default, the application points to a keycloak instance of a test stage; not yet mocked
3. Update variables in file /config/ui.js if you want to change default GUI variables file right menu's width, number of images for page, ...

Environment variable explanations

Variable	default value	Description
ENV_SILENT	true	Keep default value, details see AdonisJS reference
HOST	0.0.0.0	Configures AdonisJS api backend, should be default
PORT	3333	Configures AdonisJS api backend, should be default
NODE_ENV	development	Keep default value, production can be used for simulation purposes
SESSION_DRIVER	cookie	See AdonisJS for reference
KEYCLOAK_URI_*	see env file	Keycloak parameters point to test instance; should be mocked in future
STORAGE_PATH	./development/storages	Contains backups
ROOT_PATH	./development/mlexplorer-root	Files in workspaces are processed there, e.g. labeling, training, etc.
COMMAND_FILES_PATH	./development/script	Location of scripts
STATIC_SERVER_PORT	2929	Keep default value
STATIC_SERVER_HOST	0.0.0.0	Keep default value
SCRIPT_*	see env file	Location of script files
PATH_LOG_*	see env file	Location of respective log files
OUT_*	see env file	Configure location of training output
FOLDER_*	see env file	Subfolder configuration
DB_CONNECTION	sqlite	Until now keep default value, but sqlite not yet used

Start workspace with Docker

1. Make sure that your Docker environment is fully operational, e.g. install Docker Desktop
2. Start your local development environment: docker-compose up -d. This will provide you:
   1. A nginx that acts as a reverse proxy in order to serve backend and frontend dev environments at the same time
   2. Executes node --inspect=0.0.0.0 server.js of api component (backend part based on AdonisJS)
   3. Executes yarn run serve --host 0.0.0.0 of ui component (frontend part of application based on VueJS)
   4. Both ui and api components are started with auto refresh with live code changes and debugging
   5. Node modules being created a locally mounted (means persisted between docker restarts), but be aware that node modules being built inside the docker are not compatible with your host OS which is no problem as these are separated

Optional: Install npm packages

If you require node_modules to be in your local environment (e.g. your IDE requires it for code lookup), please execute yarn install command. Be aware that we do utilize the lock file here, so running yarn install --frozen-lockfile will make sure that exactly the versions that have been selected and bundled are being installed. If you don't use frozen lockfiles, be aware that you will change yarn.lock as well which is fine, if you intend to upgrade versions. Otherwise please don't commit changes.

Open file explorer

Create new browser window and type http://localhost:3333. That is the port nginx is listening on your local while reverse proxying to api and ui components.

Logging

Check user activities in file "SVTrain.log"

Docker

Pulling already built image

The CI is building docker images with each push for master branch and for each pull request.

The images are stored in GitHub Docker registry. To pull it you first need to find the right tag.

Each pushed image is tagged with it's git commit short hash in a form sha-<short-hash> e.g. sha-bb2b6e93.

Each pull request image is additionally tagged with pr-<pull request id>, e.g. for pull request #10 it will be pr-10. Each push to master is additionally tagged with edge tag.

List of all available images (tags) can be found under: https://github.com/com2u?repo_name=SVTrain&tab=packages.

Be carefull when using edge, if you already have that pulled on your computer, docker will not pull newer version, so it's recommended to use sha-<hash> image versions instead.

When you already obtain information about which tag do you want to run use, prepare your local configuration:

1. Adapt development.env accordingly, esp. ROOT_PATH and COMMAND_FILES_PATH. It is recommended to use /data/root and /data, respectiveley. Make sure to change the docker run commend below, in case you use different values.
2. Copy api/roles.json.example to api/roles.json. Adapt if needed.
3. Generate a user.json file by running node setPassword in directory api
4. Prepare and empty data dir by running mkdir -p ./data/root/

Now you can run the docker image as follows:

docker run \
    -p 3333:3333 \
    -v "$(pwd)/data:/data" \
    -v "$(pwd)/.env:/app/.env" \
    -v "$(pwd)/roles.json:/app/roles.json" \
    -v "$(pwd)/users.json:/app/users.json" \
    -it --rm ghcr.io/com2u/svtrain:<version-tag>

This command will automatically pull and run the specified image.

In case you just want to pull the image (without running it) you can use

docker pull ghcr.io/com2u/svtrain:<version-tag>