webapp

Web app for amazon recommender system

Architecture

web -> nginx reverse proxy -> flask app server <-> postgres/redis DB

I image this acting like an REST API where you can get rankings like:

GET /users/[userid]/rankings?limit=10

which would return the top ten items in json:

{
  [
    {asin: "xxxxxx", image_url:"http://...jpg", rank: r.r},
    {asin: "xxxxxx", image_url:"http://...jpg", rank: r.r},
    {asin: "xxxxxx", image_url:"http://...jpg", rank: r.r},
    {asin: "xxxxxx", image_url:"http://...jpg", rank: r.r},
    {asin: "xxxxxx", image_url:"http://...jpg", rank: r.r},
    {asin: "xxxxxx", image_url:"http://...jpg", rank: r.r},
    {asin: "xxxxxx", image_url:"http://...jpg", rank: r.r},
    {asin: "xxxxxx", image_url:"http://...jpg", rank: r.r},
    {asin: "xxxxxx", image_url:"http://...jpg", rank: r.r},
    {asin: "xxxxxx", image_url:"http://...jpg", rank: r.r},
    {asin: "xxxxxx", image_url:"http://...jpg", rank: r.r}
  ]
}

Setup

docker-compose build
docker-compose up

Then it should be running on port 3000

Misc

Shell:

docker-compose run --entrypoint /bin/ash web

Query to get users w/ most reviews

select reviews.aid, users.id, count(*) cnt
from reviews
join users on users.aid = reviews.aid
group by reviews.aid, users.id
order by cnt desc

K8s shell

kubectl exec -ti $POD --namespace sharknado-recsys -- bash

Export local DB pg_dump -U postgres -h localhost -c sharknado-web > database.sql

To k8s remote:

cat database.sql | kubectl exec -i $POD --namespace sharknado-recsys -- psql -U postgres -d sharknado-web