Technical Details
Two production systems built on edge computing, with zero plugin dependency
An end-to-end infrastructure and admin-panel inventory for motperde.com and bursaucuznakliye.net.tr, pulled straight from the codebase. Unlike off-the-shelf, plugin-based content management systems, both are custom software written from scratch that needs no database server and runs on an edge network.
Why it's different: compared to off-the-shelf systems
Five dimensions — compared, without naming names, against traditional plugin-based CMS architecture.
| Dimension | This system | Traditional plugin-based CMS |
|---|---|---|
| Hosting model | No single server — requests are answered from the nearest edge point worldwide, so a "server went down" incident is structurally impossible | One central server (or limited CDN caching), a bottleneck under traffic spikes |
| Database | Motperde: a low-latency key-value data store, no SQL server to manage. Nakliye: no database at all — content is compiled directly into the code at build time | A separate SQL database process — requires backups, version compatibility, query optimization |
| Attack surface | No plugins → no plugin-sourced vulnerabilities. Admin login uses custom token/session logic, so standard login-page scanners don't find anything to target | Thousands of third-party plugins, regular vulnerability disclosures, a known target for automated scanner bots |
| Page speed | Server-side rendering + static assets served from the edge; measured PageSpeed score of 100/100 (moving-company site, measured) | Slows down as the plugin stack grows, patched over with caching plugins |
| Maintenance load | Automated build + single-command deployment, no version/compatibility conflicts | Core + theme + every plugin must be updated separately; sites commonly break after an update |
| Customization | Every admin module was written from scratch for the actual need of the business (e.g. 7 different product calculator types, full integration with 5 payment providers) | Limited to whatever the off-the-shelf plugin offers; custom business logic needs extra plugins/code injection |
11
7
5
2
3
Infrastructure & Deployment
- Serverless edge architecture — every request passes through a single entry point, where payment-provider callbacks, redirects, and security headers are managed centrally.
- Two-stage deployment configuration — development and production are isolated with separate config files; image delivery to production is separately verified.
- Automatic security headers — HSTS, clickjacking protection, content-type sniffing protection, and referrer policy are added to every HTML response.
- www → root domain redirect — case inconsistencies and multiple address variants are collapsed into a single address for search engines.
- Branded error page — on a server-side rendering error, visitors see a branded page instead of a raw technical error.
Authentication & Sessions
- Custom admin authentication layer — the password is stored irreversibly, a single-use session key is generated at each login, and it expires on its own after 7 days.
- Minimal exposure surface — client-side token logic instead of a standard cookie/JWT session; automated bots scanning for generic CMS login screens don't recognize this system.
- In-panel password change — requires verification of the old password, plus a minimum-length check on the new one.
- Panel closed to search engines — the admin route is marked with a "noindex" meta tag.
Pricing & Currency Engine
- Centralized exchange-rate management — all product prices are kept in USD; a single rate change updates the entire catalog instantly.
- Markup/margin by product group — a percentage markup per category, dynamically supported even for new categories added from the panel.
- Price-history protection — even if a category is removed from the panel, its markup rate is preserved without data loss.
- "Ask for price" mode — on selected products, the price can be hidden and replaced with a contact prompt instead of an add-to-cart button.
Page & Menu Management
- Drag-and-drop menu ordering — the top menu, homepage band/category order, and footer order all sync from a single panel with one save.
- Page visibility toggle — any category/page can be hidden from the menu while its data stays archived, not deleted.
- Custom page creation — unlimited new product categories/pages can be added from the panel: title, cover image, intro text, search-engine description, parent/child page hierarchy.
- Old page → new page redirects — a removed page is automatically linked to its new address with a search-engine-friendly permanent redirect.
- Bulk price editing — all products on a page can be repriced at once from a single screen.
Product & Calculator System
- 7 different pricing engines in one panel — square-meter based, width×fullness×panel based, quantity-based campaign, model/table based, discrete measurement-list based, part/code based, and type-table based calculators — each written specifically for its own product group's real sales logic.
- Add/edit/delete products from the panel — new products, models, colors, and size options can be defined without developer involvement.
- Calculator default-value management — the starting value plus the min/max/step range can be overridden from the panel for each product/group.
Media & File Management
- In-browser image compression — every uploaded image is automatically converted into three resolutions in a modern compressed format and stored.
- Album/folder organization — thousands of images can be kept organized in tag-based groups.
- In-use image protection — an image used on any product/page is protected from accidental deletion.
- Orphaned image cleanup — unused images can be cleared both automatically and manually from the panel — storage cost doesn't balloon.
- Separate file/catalog storage — downloadable files like PDFs are served from their own public address, independent of the image store.
Payment Infrastructure
- 5 payment providers fully integrated — iyzico, PayTR, Craftgate, Moka, and PayU — each connected from scratch using its official signature/verification algorithm.
- Provider selection from the panel — the active provider, test/live mode, and on/off status are all managed from a single screen; the default is test mode and off — no risk of accidentally taking a live payment.
- Masked credentials — API keys entered in the panel are never shown in full again — only the last 4 digits are visible for verification.
- Payment data excluded from backups — given its sensitivity, payment credentials are deliberately excluded from the general backup system.
- Provider-specific callback endpoints — each payment provider's callback request is handled separately at the raw HTTP level, with signature verification.
Backup & Disaster Recovery
- Daily automatic backup — taken every night via a scheduled task, with the last 14 days retained.
- Instant backup before saving — the current state is automatically backed up right before any "save" action in the panel.
- One-click restore — you can roll back to any backup point; the current state is separately backed up before a restore, too.
- Export — the entire site's data can be downloaded as JSON.
AI Image Studio
- Text-to-image room previews — when a customer picks a curtain group, color, and room type, a matching scene description is assembled automatically and an image is generated instantly by an image-generation model running at the edge.
- Prompt engineering per product group — separate scene rules are defined for 9 different curtain groups.
- Quota-honest error handling — when the daily generation limit is hit, the user sees a clear message stating the real status — no silent failure.
SEO & AI Visibility (GEO)
- Sitemap generated from live data — every change made in the panel is reflected in the sitemap instantly, and removed pages are automatically dropped.
- Structured-data automation — rich-result data for the store, products, and website is generated automatically on every page; fake ratings/review data is deliberately never added.
- Summary file for AI search engines — a separate, live-data-generated summary endpoint is provided so tools like ChatGPT/Perplexity can accurately summarize the site.
- Search console integration — real click/impression data is regularly exported and archived.
A standout engineering decision — something that looks like a single action, like reordering the menu, consistently updates 3 separate data sets behind the scenes; the user only ever sees one "save" button.
bursaucuznakliye.net.tr
corporate / service site · data-driven, panel-free architecture100/100
45+
6
0
Infrastructure & Deployment
- Fully data-file-based content architecture — services, districts, neighborhoods, routes, and blog posts are all kept in type-safe code files; there's no separate database queried at runtime — one less point of failure.
- Public test subdomain deliberately disabled — prevents the risk of a search engine seeing the same content at two different addresses and splitting its ranking value.
- One-year immutable cache on static assets — repeat visits load almost instantly.
- Build-time-fixed timestamps — the "last updated" info in the sitemap is fixed at the actual build moment — no false "changes every day" signal is sent to search engines.
Security
- Zero-attack-surface architecture — with no login form, user account, or database, classic attack classes like brute-force, SQL injection, and session hijacking are technically not applicable at all.
- Security headers at both the server and static-asset layers — HSTS, clickjacking protection, and content-type sniffing protection are present on every response.
- The single external endpoint is protected by a shared secret key — the one endpoint used by the automation agent is verified with a key that isn't publicly known.
Autonomous AI Content Agent
- Daily automatic FAQ generation — a scheduled external trigger calls the secure endpoint once a day, pulls draft content from two different AI models, and adds it to the site content.
- Quota-aware error handling — a limit being hit is clearly detected and reported, never a silent failure.
- Human-reviewed publishing flow — generated content doesn't go live directly — it passes through the normal build/deploy process first.
SEO & AI Visibility (GEO)
- Sitemap, robots, and AI-summary file all generated from data — service/district/neighborhood/route/blog listings are derived automatically from a single data source, never edited by hand.
- Doorway-page discipline — neighborhood pages are deliberately kept to a limited number — a safeguard against being penalized as low-quality/templated pages.
- Structured data with real pricing — service pages use real, customer-confirmed price tiers instead of invented prices or ratings.
- Search console verification and fast indexing of new content — new/updated pages are reported to the search engine without waiting for the standard crawl.
Current-status note — this site currently runs entirely on code/data files with no panel; unlike motperde, this repository doesn't have an admin panel. That's actually an advantage — "maintenance load close to zero" — and a modular panel architecture could be brought here too if requested.
Bottom Line
Standout engineering decisions
Details that give a buyer code-level proof that this was actually taken seriously.
Honest data discipline
No page carries fake ratings, fabricated reviews, or made-up stock information — never sending Google a false signal was a deliberate design choice.
Measurable SEO progress
Search console data is archived regularly; any claim that "traffic is growing" rests on recorded measurement, not assumption.
Self-protecting backups
An automatic snapshot is taken before every admin save — the risk of an unrecoverable wrong click is practically eliminated.
Zero plugins, zero license risk
Not dependent on any third-party plugin/theme license — no future risk of "the plugin developer abandoned the project."
Provider credentials are never shown in full
Keys entered for payment integration are masked even inside the panel — the insider-threat surface is narrowed.