AI-Stack-Fit

Which AI-stack is really yours?

A browser-side screening tool for DACH-region companies that figures out, in 90 seconds, which 12 of 52 catalogued AI and business systems actually fit a given setup — including an EU-sovereignty score, friction check, trade-off profile and PDF export.

Live: braum.org

What it does

• Wizard in 4-5 steps (mode, size, industry, revenue, existing stack, strategy) collects the profile entirely client-side.
• Scoring ranks 52 systems across 12 categories (Office, AI, Comm, Knowledge, CRM, ERP, Data, Automation, Dev, Marketing, Support, Cloud) and picks one top match plus three alternatives per category.
• Insights include five KPIs (synergy, AI readiness, dominant family, EU sovereignty, license cost), a trade-off bar profile, industry-specific peer notes and automatic friction detection in the resulting stack.
• Worth Watching highlights one underdog or rising star per category (e.g. Aleph Alpha, Mattermost, Klaviyo, Cursor) as inspiration beyond the mainstream picks.
• Dice button generates a random valid profile straight from the landing — useful for demo walks.
• PDF export is built locally in the browser via jsPDF (forest-gradient background, around 80 KB, A4) behind a light lead-gate (name, company, email, consent). Lead data is sent only as a custom event to Umami — no mailing list, no backend.

Stack

Framework	Astro 5 SSR (@astrojs/node, standalone)
UI	React 18 (one island with client:load)
Styles	Tailwind 4 + CSS variables, Forest/Mint/Cream tokens
Animations	GSAP (inside the tool island bundle), CSS keyframes for polish
PDF	jsPDF (lazy-imported on click)
Analytics	Self-hosted Umami, custom events
Lint/Format	Biome
Runtime	Node 22-alpine, multi-stage Dockerfile, non-root

Data pipeline

The tool catalogue is CSV-driven:

data/data.csv  ──┐
                 │  pnpm gen:data
                 ▼
src/components/labs/ai-stack-fit/systems.generated.ts

scripts/gen-data.mjs validates and generates a typed SYSTEMS array along with a refresh-date label. Scoring logic, frictions and industry insights live hand-curated in data.ts. Refresh workflow is documented in docs/research-prompts/.

Quick start

pnpm install
cp .env.example .env   # leave defaults for local dev
pnpm dev               # http://localhost:4322

Scripts

Command	Action
pnpm dev	Astro dev server with HMR
pnpm build	astro check && astro build
pnpm preview	Serve the built output
pnpm check	astro check && tsc --noEmit
pnpm gen:data	Rebuild systems.generated.ts from data/data.csv
pnpm check:data	Validate the CSV without writing
pnpm lint	Biome lint
pnpm format	Biome write-formatter

Repository layout

.
├── data/data.csv                     # System catalogue (52 tools, single source)
├── docs/research-prompts/            # LLM prompts for quarterly catalogue refresh
├── public/
│   ├── og/default.png                # 1200x630 social card
│   ├── sitemap.xml
│   └── robots.txt
├── scripts/gen-data.mjs              # CSV → generated TypeScript
├── src/
│   ├── components/labs/ai-stack-fit/ # Tool React island + scoped styles
│   ├── layouts/BaseLayout.astro      # Head, OG, JSON-LD, atmosphere
│   ├── lib/
│   │   ├── env.ts                    # SSR / Vite env shim
│   │   ├── pdf.client.ts             # jsPDF builder
│   │   └── track.client.ts           # Umami helper + funnel registry
│   ├── pages/
│   │   ├── index.astro               # Tool lives at /
│   │   ├── 404.astro
│   │   ├── api/health.ts             # Container probe
│   │   └── labs/[...rest].astro      # 301-redirects for legacy /labs/* paths
│   └── styles/global.css             # Tokens, atmosphere layer, eyebrow utility
└── Dockerfile                        # Multi-stage, runs `node ./dist/server/entry.mjs`

Container

docker build -t ai-stack-fit .
docker run --rm -p 4322:4322 \
  -e PUBLIC_SITE_URL=https://your.domain \
  ai-stack-fit

The HEALTHCHECK hits GET /api/health every 30 s.

License

MIT — see LICENSE.