/markup_server

Primary LanguageHTML

This is a markup server, which may be connected to an ML core.

You can run it by a docker with nginx as a web-server. To start you will need a MONGODB and an S3 bucket, where all input and result annotation are stored.

Pay attention to YOUR_SITE in confd_nginx.conf before building a docker image - this should correspond to your site name.

To build and run:

sudo docker build -t markup .

sudo docker create --name markup_server -p 27017:27017 -p 443:443 \
            -e ....
            markup:latest
sudo docker cp ssl_and_bundle.crt markup_server:/app/ssl.crt
sudo docker cp key.pem markup_server:/app/ssl.pem
sudo docker start markup_server

Environmental variables:

  • SECRET_KEY - random key for FLASK
  • MONGO_DB_ADDRESS - like mongodb:// .... , that contain everything to connect
  • MONGO_COLLECTION - just a name of the collection for this service
  • AWS_KEY
  • AWS_KEY_ID

You can exclude the SSL certificate by changing protocol in nginx config (change 443 port to 80 and remove lines about ssl certificates).

To add a new user, run add_user.sh role token and a token. Then, you can register a new user passing him the token and the following link: your_site.com/register/role/token.

Or, being an admin, you can add user through the address line: /add_pre_user/role/token (token is a random key-digit sequence)

Possible roles for users:

  • admin
  • operator

All html templates are placed in the folder templates with corresponding names, described in config.json

Input data

There are two ways to upload new data to markup.

  1. Just choose files in "upload" on the site your_site.com/markup/<application_name>
  2. Or your can upload files through POST requests. ('/upload/string:application'). Using API you also can upload results from an ML machine throug '/upload_result/' attaching 2 files: image + json. Then, you'll see pre-marked up images in b>your_site.com/markup/<application_name>

Debug

Also, to debug, you can run python3 main.py from app folder.

You'll also need two more env variables DEBUG_ADMIN_EMAIL and DEBUG_ADMIN_PASSWORD to create a new admin user without running add_user.sh script.