>>526866159
I finished my scrape of vrcw.net's database of vrchat world metadata. It has 236,593 worlds, with upload date distribution in picrel. It also has world name, author name, and tags (both in original japanese, and translated back to english through PLaMo, which I glanced over and they seem reasonable). Compresses to 11MB of parquet. You can query the data yourself with duckdb:

create table vrcw as select * from 'https://file.vrg.party/vrcw-scrape-2025-06-11.parquet';

or download it and query with pyarrow or whatever. I also scraped most of the world thumbnails which you can download at https://file.vrg.party/vrcw-scrape-images-2025-06-11.zip (3.5 GB).

This isn't that useful as is since it's missing other metadata like player limit, assetbundle size, view and favorite count, etc. However, it might still be useful as is if you want to, say, take screenshots with world ids in the title and resolve them to an actual world name, or do some basic queries with tags and world names. The important part for me is the world ids themselves, which I can now use to scrape the actual vrchat API for the rest of the metadata. I dunno what vrchat rate limits requests to, but assuming a conservative 1rps, it'll take another day or so.