Skerter

Case study / 01

DemoPlast

A lead survives even when integrations fail.

Role
Backend engineering
Context
Commercial project showcase
Status
Live
Core
Django / PostgreSQL / HTMX

Context

DemoPlast is an anonymized showcase of a real commercial project: a six-page product catalog and lead-generation site for a manufacturer. Client data is replaced, while the architecture and engineering decisions remain intact.

The product needed an indexable catalog, fast forms, editable business details and reliable delivery to email, Telegram and amoCRM. It also had to handle personal data deliberately instead of treating compliance as a document added after launch.

Lead flow

  1. Validate at the boundary

    Django validates consent, rate limits and the honeypot before the application accepts a submission.

  2. Persist before integration

    PostgreSQL receives the lead first and becomes the source of truth. No external service sits on the critical request path.

  3. Fan out independent work

    The dispatcher queues email, Telegram and amoCRM as separate django-q2 tasks rather than one fragile notification chain.

  4. Retry one channel at a time

    A CRM outage or expired OAuth token can be retried without losing the submission or blocking the other delivery channels.

Decisions

  1. Monolith over microservices

    Six pages, a catalog and a form do not justify distributed deployment complexity. One Django application keeps ownership and operations explicit.

  2. Server rendering over a SPA

    Django templates keep the catalog indexable. HTMX adds fragment updates and validation feedback without turning the form into a separate client application.

  3. Privacy inside the system

    Consent is validated and timestamped, analytics wait for cookie consent, and Sentry strips names, phone numbers and form bodies before an event leaves the process.

What shipped

The public repository preserves the production architecture while remaining safe to run without real credentials.

  • Functional demo mode with generated catalog, images and leads.
  • Tests for persistence, channel failures, services, consent-gated analytics and the Sentry scrubber.
  • Automated image build and deployment through GitHub Actions, GHCR, Docker Compose and Traefik.
  • Bilingual technical documentation with runbooks and architecture decision records.