Dolores - Frontend Expert System

This is a demo of a frontend expert system with the Personal Data Protection Act (PDPA) example.

Requirements

• NodeJS (versions 16 and above)

Setup and Usage

Configuring Environment Variables with .env

Before proceeding with the steps below, copy the .env.example to .env. Inside the .env file, configure the following settings according to your application:

BASE_URL=/
VUE_APP_BROWSER_NAME=CCLAW
VUE_APP_NAME="PDPA DBNO PoC - Draft"

Installing NVM

NVM is a version manager for node. Follow the instructions on the nvm github.

Installing Packages

First, install nvm from https://nvm.sh/

Then run the following set of install scripts, from inside the vue-pure-pdpa directory:

$ nvm install
$ nvm use
$ nvm exec
$ npm install -g node-gyp@latest
$ npm install
$ npm run deps

Running in Development

Start the application in development mode by running:

$ npm run serve

By default, that brings up the application at localhost:8080.

If you've been assigned a different port 8xxx, use

$ npm run serve -- --port=8xxx

If you have a working web app at the interface, you should be able to proceed to use this repo as part of the L4 backend.

Building for Production

To build the application for production, run:

$ npm run build

A dist/ directory should be produced. Deploy the contents in the dist/ directory to your application of choice (e.g. Github pages).

An example can be found at https://smucclaw.github.io/vue-pure-pdpa/ or https://smucclaw.github.io/mengwong/pdpa/.

Basically, mv the relevant files from dist/ into the smucclaw/smucclaw.github.io repository.

Using Docker

Alternatively, you may run this application on a Docker container by running:

$ docker-compose up --build

Using v8k

v8k is a little server-management script that allocates new development servers to ports in the 8000 range.

Once you're able to run npm run serve you can think about spawning new servers in response to updates from the Python/Flask/natural4-exe end of things.

Setup

You only have to do this once:

mkdir ~/v8kworkdir
cd ~/v8kworkdir
export V8K_WORKDIR=$HOME/v8kworkdir
git clone git://github.com/smucclaw/vue-pure-pdpa
rsync -va vue-pure-pdpa/ vue-big/
(cd vue-big; npm i; npm run deps)
rsync -va --exclude={.spago,.git,node_modules} vue-big/ vue-small/
cd vue-small
ln -s ../vue-big/.git .
ln -s ../vue-big/node_modules .
ln -s ../vue-big/.spago .

Remember to update your gunicorn conf file to set V8K_WORKDIR to the path as above.

This sets you up with vue-small, which is a copy of vue-big, which is a copy of the vue-pure-pdpa repo.

vue-small takes up a bit less space, by taking advantage of symlinks to reuse existing files that don't change across copies.

This becomes valuable because v8k later rsyncs vue-small to vue-01, vue-02, and so on, at runtime. This rsync happens when we're trying to hurry: the end-user could click on the "vue web app" link at any time, so we want to bring up the vue web app as fast as possible.

If the vue-YY directory doesn't already exist, then the rsync is liable to be slow, so we want to minimize the size of the vue-small.

We could pre-rsync the vue-YY directories at this point, too, with something like

for yy in 01 02 03 04 05 06 07 08 09; do rsync -va vue-small/ vue-$yy/; done

Spawning a new server

Every time the Python Flask subsystem runs natural4-exe to refresh the workdir output, it will want to tell Vue "here is a new .purs file that contains a Rule Library".

It should run:

~/src/smucclaw/vue-pure-pdpa/bin/v8k up --uuid=000 --ssid=111 --sheetid=222 ~/src/smucclaw/dsl/lib/haskell/natural4/workdir/000/111/222/purs/LATEST.purs

The STDOUT of this command will be a port number and path. Append that to the server public ip or domain name to get something like

http://18.139.62.80:8001/000/111/222/

And a few seconds later that link should serve an instance of the Vue app configured with the LATEST.purs that you gave it.

The working name for this expert-system part of the web app is called Orwell, because it is good at evaluating And/Or trees in the context of a mechanized public service interface with the public that computes truth values.

The above up call is idempotent: the Flask system can just re-run it when the workdir changes and serve the constructed URL back to the sidebar. Most of the time the URL will not change but sometimes it will, so do read it each time.

Bringing Down A Server

Usually you can just leave the servers running. The v8k script has the notion of a pool, whose default size is 10; after 10 slots are allocated, new up commands will overwrite the oldest server.

If you have the notion of "ending a session" in the Google Sheets "IDE" frontend, the Python Flask subsystem is invited to call

~/src/smucclaw/vue-pure-pdpa/bin/v8k down --uuid=000 --ssid=111 --sheetid=222

This will deallocate the running instance and make the slot available.

Bringing Down A Server With Great Prejudice

If you know for sure you want to bring down slot 03,

~/src/smucclaw/vue-pure-pdpa/bin/v8k downdir 03

Checking to see what's running using v8k

This command will walk the V8K_WORKDIR to see what's going on -- in theory.

~/src/smucclaw/vue-pure-pdpa/bin/v8k list

Checking to see what's running

If you don't trust it, probe the OS:

ps wwaux | grep port=80

to see what's on port 80something.