MehZ Solution
My agency's live website: services, pricing and case studies, with service requests, a client dashboard, an owner admin and an AI assistant.

Live on mehzsolution.com · own Rust API, client dashboard and AI assistant
Independent build
The website of my own agency, MehZ Solution: designed, built and run by me on the shared server.
THE ENGINEERING
The live website of MehZ Solution, the agency I founded. An Astro static site with a Rust (Axum) API and SQLite: visitors request a service instead of buying blind, get an account by email link and follow their request in a client dashboard; I set status, quotes, payment links and replies from an owner admin. Partner applications, legal pages, every portfolio project as a case study and an AI assistant that answers service, pricing and process questions under a one-dollar daily spend cap. The scroll-scrubbed hero film was made from generated keyframes with no people in it. It runs as one hardened container (non-root, read-only, loopback port) behind the shared server's Caddy and Cloudflare, with daily SQLite backups and a tested restore. 19 Rust tests, a 38-check API walk-through and fresh-context reviews before release; the live container matched its release manifest file for file (166/166). On November 12 the site moved to its own domain, mehzsolution.com, with strict HTTPS; old and www links redirect permanently with their paths kept, the 40 public pages are in a new sitemap, and six new domain mailboxes joined my existing mail hub. That release matched 198 of 198 files on the server, and its API walk-through passed 38 of 38. Backups of the site database are not yet copied off the server automatically.
BEFORE → AFTER
Real problems from this build, what I changed and the verified result.
- CASE 01Before
Sign-in answered faster for emails without an account, so timing revealed who was a client.
AfterUnit, integration and API walk-through tests assert the same error for both cases, and the code always runs one password-hash check.
How I fixed itEvery sign-in now runs exactly one Argon2 check on the blocking pool, whether or not the account exists.
- CASE 02Before
Through the CDN the hero film could not be seeked: a byte-range request returned the whole file instead of a part.
AfterRange requests return 206 through Cloudflare on cache miss and hit; the film scrubs to its full 18 seconds.
How I fixed itStopped compressing video, added a regression test that sends a CDN-style request, and versioned the video URL.
- CASE 03Before
After rebuilds, browsers kept showing old pages because one cache rule covered HTML and assets alike.
AfterRebuilt pages appear on reload, and an integration test checks the API is never cached.
How I fixed itCache by file type: HTML revalidates, hashed assets are immutable, API responses are never stored.
Evidence reviewed: Project dossier and release records, October 31 – November 12, 2026