Nex Demo Factory Generator

What it does

• Generates pipeline.py, an async 6-step runner (scrape, places, classify, brand_extract, generate, compile) tracked in a generation_jobs Postgres table with pause checkpoints and per-step timing.
• Ships the per-sector archetype config system: one JSON preset per sector decides sections, tone, palette and imagery for the generated site.
• Includes the AI content generator with a Claude-then-Gemini-then-Qwen fallback chain, so one rate-limited API never stops a batch run.
• Compiles the final site with Jinja2 section templates into a single deployable HTML file per prospect.
• Defines the client config JSON contract, the one file that carries a prospect through every step.

Why use this

Demo-first selling works when the demo exists before the pitch. This factory automates the path from prospect URL to a branded demo site, built from the pipeline patterns Nex AI uses for its own outreach. You feed it a URL; it scrapes, classifies the sector, extracts brand colors, generates copy and compiles a site.

See it in action

Prompt:

Set up a pipeline that turns a company website URL into a finished demo site, sector-aware, with AI copy.

Result:

POST /jobs {url: prospect.be}
[1/6] scrape        -> texts, images, contact data
[2/6] places        -> reviews + hours via Places
[3/6] classify      -> sector: bakkerij (archetype: food-retail)
[4/6] brand_extract -> #8B4513 / #F5DEB3, serif headings
[5/6] generate      -> sections copy via Claude (fallback: Gemini, Qwen)
[6/6] compile       -> dist/prospect-be/index.html (single file)

Use cases

• Auto-build prospect demos before pitching, at scale.
• A sector archetype engine for a section-based site generator.
• A resilient bulk content pipeline with model fallback.
• An agency demo factory behind a FastAPI dashboard.

Known limitations

• A scaffold for your factory, not a hosted SaaS; you deploy and operate it.
• Needs Postgres and at least one LLM API key; Places data requires a Google API key.
• Scraping respects robots.txt and rate limits by design; some sites block scraping entirely.

FAQ

What exactly is in the download?

SKILL.md, README.md, the pipeline runner, archetype config system, AI generator with fallback chain, Jinja2 compiler wiring, client config contract, reference docs, manifest, changelog and license.

Which agents does it work with?

Claude Code, Claude Cowork, Cursor, Codex CLI, and any SKILL.md agent.

Do I need anything else?

Python 3.11+, Postgres, and an Anthropic or Google or Qwen API key. A server or Pi to run it on.

Is this the same as the single-file template skills?

They compose: the factory generates and compiles into the single-file architecture; the sector template skills are hand-tuned starting points for one-off builds.