HackPrinceton Matterport: models_index.csv returns HTML (200 text/html) or 403 on static.matterport.com: the real URL is under hackathon/<build>/assets/quickstart/
Symptom: pd.read_csv('https://beta-app.matterport.com/2026-princeton-hackathon/models_index.csv') either fails with a tokenizing error or gives a frame of HTML lines, and curl -I shows HTTP/2 200, content-type: text/html;charset=UTF-8. Variants like /2026-princeton-hackathon/data/models_index.csv, /models.json, /splat_models.csv and https://beta-app.matterport.com/resources/models_index.csv behave the same way. Guessing https://static.matterport.com/2026-princeton-hackathon/models_index.csv gives 403 Forbidden with an XML body. That bucket most likely returns 403 for any missing object, so it looks as if the data is locked.
Cause: the site is a client-side SPA that answers every unknown non-API path with its index.html and HTTP 200. The quickstart reference page links models_index.csv relatively, and that link resolves against the static asset folder, not the site root.
Fix: the file is public at https://static.matterport.com/hackathon/<build>/assets/quickstart/models_index.csv (content-type: text/csv, about 2.6 MB). <build> is the versioned folder in the site's main script URL. Get it with:
BUILD=$(curl -s https://beta-app.matterport.com/2026-princeton-hackathon/ | grep -oE 'static.matterport.com/hackathon/[^/]+' | head -1)
curl -sO https://$BUILD/assets/quickstart/models_index.csvAt the time of writing the folder was hackathon/26.15.35-main-35-g24e0220. Don't hardcode it, because it changes on redeploy. <site>/assets/quickstart/models_index.csv answers 307 to the same file, so curl -L on that also works. The same folder serves matterport_quickstart.py / .ipynb and reference.html.
Verify: head -1 models_index.csv should print the header row starting with sid,name,segment,lat,lon. There's no need to generate synthetic sample models for a geo pipeline: this CSV has the real sids with lat/lon for Houston and Los Angeles, and .../2026-princeton-hackathon/api/models lists the subset with splats and drone photos.