nt-ai — AI for TYPO3

AI-powered tools for TYPO3 backend — accessibility audit, accessibility statement, alt-text generation, content quality scoring and Lighthouse monitoring — plus a frontend accessibility widget with read-aloud.

• Text generation

  ---

  Generate copy, summaries, translations directly in the backend — wherever you see the "✨ KI" button.

• Alt-text generation

  ---

  Alt texts for images (vision-based) and for PDF documents (from the PDF's text), configurable for your brand's tone and style. Run individually or in bulk via CLI.

• Accessibility audit

  ---

  WCAG 2.2 checks with AI-generated fix suggestions, 26 built-in rules, on demand or scheduled.

• Accessibility statement

  ---

  Turns the audit data into a draft accessibility statement (BITV model / BFSG § 14) and publishes it after editorial approval — multilingual, versioned.

• Content scoring

  ---

  Five LLM-driven quality scores (GEO/SEO, Performance, Semantics, Keywords, Accessibility) per page, with sparkline history.

• Lighthouse monitoring

  ---

  Google PageSpeed Insights scores and Core Web Vitals, tracked over time via the PSI v5 API.

• Score alerts

  ---

  Email notifications when pages fall below configurable thresholds for AI content scores or Lighthouse metrics.

• Accessibility widget

  ---

  Frontend personalisation (not an overlay): text size, contrast/saturation, blue-light filter, dark mode, colour deficiency, motion, bigger cursor and more — backend-configurable, selectable icon.

• Read-aloud (TTS)

  ---

  Self-hosted text-to-speech replacing ReadSpeaker — hybrid (Web Speech or cloud, cached), in the widget and freely placeable in the template, rate-limited.

• PDF accessibility

  ---

  Checks all PDFs against PDF/UA/Matterhorn (tagged, text layer, language, alt texts …) — backend module with traffic light & recommendations, upload gate (warn/block), CLI for CI.

Supported AI providers

Provider	Default model	Vision
Anthropic Claude	claude-sonnet-4-5	All Claude 3+ / 4+ models
OpenAI	gpt-4o-mini	gpt-4o, gpt-4o-mini
Google Gemini	gemini-2.0-flash	All listed Gemini models
DeepSeek	deepseek-chat	No
Mistral AI	mistral-small-latest	pixtral-large-latest
Groq	llama-3.3-70b-versatile	No
Ollama (local)	llama3.2	Multimodal models (e.g. llama3.2-vision)

Requirements

• TYPO3 12.4 LTS, 13.4 LTS, or 14 LTS
• PHP 8.1 – 8.4
• At least one AI provider API key configured (or a local Ollama instance)

Get started Configuration reference

License

GPL-2.0-or-later