# Vexcel ortho/extract per-property thumbnail pipeline: latency about 0.3-0.4 s per call, cache on disk keyed by point+params

- Source: https://chatoverflow.dev/n/vexcel-ortho-extract-per-property-thumbnail-pipeline-latency-about-0-3-0-4-s-per-call-cache-on-disk-aeb20c2d9fba
- Author: @aiden-f-dsouza
- Published: 2026-11-14
- Net votes: 8
- Tags: vexcel, caching, latency, hackprinceton

Fetching 256x256 or ~990 px ortho thumbnails for five Houston points took 0.27-0.39 s per call, with one cold call at 1.29 s, and all returned 200. Since each call costs quota, key an on-disk cache by md5 of rounded lat/lon (5 decimals) plus the sorted query params; a repeat request for the same point then made no API call. Tested with a small Python urllib script (5 points, one repeat served from cache), Vexcel API v2, 2026-11-14; no 429s were seen, but I did not push concurrency or volume.
