Zero-K API ?

Hi,

I'am searching an API for zero-k to search for example the grade of a user.

Thanks.

I'd be interested in building something simple and secure like this on top of the ZK database if there's interest. Probably just a flask server.

I'd love that.

We can host it for free on https://fly.io/ (3x 256mb RAM VMs for free monthly)

Looks like the server is MSSQL 2019. So, PostgREST won't work but hasura will work basically straight out of the box (Just, please no graphql..). Hasura is basically just a dbms client with tooling attached, so we don't necessarily need to figure out the schema beforehand.

Once connected to the prod db, we'd then just expose API-relevant data over API.

The big challenge, depending on the ZK Dev's threat model, is that hasura will need a read-only user that has certain sensitive tables blacklisted, just so they're not exposed through Hasura just so that me nor Killingtime or whoever works on this, can steal those tables.

Aquanim -- is that something you'd be comfortable with? AFAIK, we'd just need a read-only user, preferably with credential/moderator tables blacklisted.

If you are interested in scrapped stuff you can find a sqlite database here: http://zerok-local-analysis.s3-website.eu-west-3.amazonaws.com/data/zk.db.gz (the one that backs this stats website http://zerok-local-analysis.s3-website.eu-west-3.amazonaws.com/release2022-07-07-01/, see http://zero-k.info/Forum/Thread/34840?page=1&Search=&User=&grorder=&grdesc=False&grpage=1)

I did not yet save the rating, I could do that as well if there is interest. If you want them you can easily do a get on the page "http://zero-k.info/Charts/Ratings?RatingCategory=1&UserId=<user of interest>" and look for the data.

For me, most interesting statistics will overwhelm the server and see no easy way to restrict that. We could get a database export regularly (and compute the statistics in our own environments), but the database export would have same info as the scrapping above, without any server change. That's why I did a javascript based solution in which you do all computation locally (downside: you need to download the data, but I have data split per year, so if you want just last year will be smaller, plus browser should cache it so you do it only once).

I think some computations (ex: recomputing WHR/trying other rating algorithms/big summarizations over large periods of time) should not be done on the server. A regular database export would be nice and more "scalable" (downloading once a week a large file might be easier than installing an application on the ZK server)