HAULWRIGHT | FRESH-EXTRACTION VERIFICATION REPORT
Date: 27 October 2026
Scope: Existing package, with zero edits to the application source files.

VERIFIED ON ISOLATED TRAINING DATABASE
PASS  ZIP archive consistency and SHA-256 checks for application files.
PASS  Embedded Haulwright logo/truck assets decode as valid WebP images.
PASS  Python backend imports, starts, returns health JSON and serves assets.
PASS  All 13 management tabs render without JavaScript exceptions using
      an automated Chromium browser with a Python-to-API bridge.
PASS  Desktop and mobile layouts render; no horizontal overflow on 390px phone.
PASS  Create customer, driver, truck and uniquely referenced dispatch load.
PASS  Reject conflicting truck/driver scheduling.
PASS  Require a delivery photograph before marking a load delivered.
PASS  Upload JPEG/PNG photo, confirm quantity, receiver and ticket number.
PASS  Automatically generate linked numbered draft invoice on completion.
PASS  Prevent duplicate delivery completion; update invoice to Issued/Paid.
PASS  Record fuel, driver daily logs, maintenance and truck-specific expenses.
PASS  Fail an inspection and place the associated truck out of service.
PASS  Retain audit activity and download load CSV export.
PASS  Back up database + uploaded photos and restore them to a clean folder;
      restored database passes SQLite integrity_check.

Sample scenario: HW-2026-00004 -> INV-2026-00002, delivered quantity 17.5
with unit rate BSD 55 produced a BSD 962.50 draft invoice in the isolated QA.
Identifiers in any deployment will depend on its database history.

NOT VERIFIED
- External hosting provider configuration, public HTTPS/DNS or /training/
  proxy settings; these must be tested by the webmaster after installation.
- Multi-user security or verified audit attribution; no login is provided.
- Tax/accounting compliance, regulated vehicle inspection requirements,
  or synchronization with third-party accounting platforms.

QA sample test data is NOT bundled in this release. The server creates its
own fictional sample records when started in training mode.
