url-to-snapshot pipeline + queued worker and artifact backend — snapshot

the first version of snapshot worked, but the backend was still too tied to short request-time execution. longer runs were fragile, and the same problems kept showing up: deployment friction, auth issues, timeout pressure, and jobs dying before the result was safely stored. screenshot mode had promise too, but it was not stable enough to be the main path.

we tried a few directions before landing on the right shape. netlify stayed in front as the control plane for auth, request handling, and job creation, but the actual snapshot work had to move out of that one-request path. the final setup uses firestore to track jobs, cloud tasks to queue the work, cloud run to handle the long-running worker path, and firebase storage to keep the generated artifacts.

that solved the real problem. url mode now runs end-to-end through the queued worker pipeline and stores clean artifacts with proper job tracking. screenshot mode improved through the same backend shape and headless browser / ai worker path, but it still feels beta. the main win here was turning snapshot from a fragile live request into something that can actually finish the job reliably.