ScrapingEvals
Operational evidenceNot a benchmark

Field notes from scrapes that were already happening.

Scrape Gateway now sends privacy-reduced attempt traces into a private review inbox. We inspect them, publish only the useful slice, and keep every claim tied to its source events.

Reviewed runs
11
Provider attempts
13
Public domains
5
Failed attempts
2
Recorded units
43

Route escalation · one request · path withheld

One G2 request consumed two failed tiers before Hyperdrive returned validated content

In one ordinary Scrape Gateway run, ScrapeDrive Standard and Advanced each returned HTTP 500. Hyperdrive then returned HTTP 200 with content validation passing. The route recorded 31 estimated cost units in total.

Attempt trace

ordered by ledger event
  1. Tier 1Standard
    scrapedrive:standardevent 5
    714 ms1 unit · estimated
    HTTP 500
  2. Tier 2Advanced
    scrapedrive:advancedevent 6
    1.01 s5 units · estimated
    HTTP 500
  3. Tier 3Hyperdrive
    scrapedrive:hyperdriveevent 7
    11.59 s25 units · estimated
    HTTP 200
End-to-end
13.79 s
Route cost
31 est. units
Final content
Validated

Private artifacts, public hashes

final html1,220,313 bytes
sha256:5959d189b06d…65eb5046
final markdown85,337 bytes
sha256:0730690f1f73…34c59eaa

What this does not prove

  • This is one request to one redacted G2 path from one network position.
  • It does not compare ScrapeDrive with another provider or estimate general success rates.
  • All 31 cost units are provider-route estimates, not provider-reported billing.
  • No screenshot was requested, and artifact bodies remain private.
Audit statusAdvisory pass from google/gemini-3.1-flash-lite; uncalibrated audit. The model verdict is advisory. Publication relies on the recorded route outcomes and deterministic content validation.
ProducerScrape Gateway v0.23.0 · 3 source-scoped ledger events