Configuration

Extension configuration

Open System > Settings > Extension Configuration and select t3_content_quality. The settings are stored in config/system/settings.php under $GLOBALS['TYPO3_CONF_VARS']['EXTENSIONS']['t3_content_quality'].

Name	Type	Default
aiProvider	options	`ollama`
enableAiAnalysis	boolean	`1`
maxTokens	positive integer	`16000`
anthropicApiKey	string	(empty)
anthropicModel	string	`claude-sonnet-5-5`
anthropicEffort	options	`low`
anthropicFallback	boolean	`1`
openaiApiKey	string	(empty)
openaiModel	string	`gpt-4o-mini`
ollamaModel	string	`llama3.2`
ollamaUrl	string	`http://ollama:11434`

General

aiProvider

Type: options
Default: ollama

The AI provider used by all AI features. Possible values:

anthropic: Anthropic Claude. Requires anthropicApiKey.
openai: OpenAI. Requires openaiApiKey.
ollama: A local Ollama server, no API key needed. See Ollama.

enableAiAnalysis

Type: boolean
Default: 1

Uses the AI during page analysis for improvement suggestions and internal link suggestions, and enables AI schema detection.

When disabled, the page analysis is rule-based only. The on-demand AI features (Fix buttons, text generator, image metadata) are not affected by this switch; they only need a configured provider.

maxTokens

Type: positive integer
Default: 16000

Upper limit of tokens per AI response for Anthropic and OpenAI. For current Claude models the limit also covers the model's thinking, so do not set it too low; you only pay for tokens actually generated. If the limit is reached before an answer is written, the log shows "Anthropic response hit maxTokens before producing text". Ollama ignores this setting.

Anthropic

anthropicApiKey

Type: string
Default: (empty)

API key from the Anthropic Console. Required when aiProvider is anthropic.

anthropicModel

Type: string
Default: claude-sonnet-5-5

Model ID sent to the Anthropic Messages API, for example claude-sonnet-5-5 (default), claude-opus-5-5 (higher quality, higher cost) or claude-haiku-4-5 (lower cost). All three accept images, which the file list metadata button needs. See the list of Anthropic models for current model IDs, as older models are retired over time.

anthropicEffort

Type: options
Default: low

How much the model thinks before answering: low, medium, high or off. The extension's tasks are short, so low keeps cost and response time down. The value is sent as output_config.effort. It is not sent for claude-haiku-* and claude-3* models, which do not support it; off never sends it.

anthropicFallback

Type: boolean
Default: 1

If the model declines a request (stop_reason: refusal), Anthropic can retry it on another model server-side (fallbacks: "default", beta header server-side-fallback-2026-07-01). Only sent for models that support it (claude-sonnet-5-5, claude-opus-5-5, claude-opus-5, claude-fable-5-1). Declined requests that are not retried are logged with their refusal category.

OpenAI

openaiApiKey

Type: string
Default: (empty)

API key from the OpenAI platform. Required when aiProvider is openai.

openaiModel

Type: string
Default: gpt-4o-mini

Model ID sent to the OpenAI Chat Completions API. For image metadata the model must support image input (gpt-4o-mini does).

Ollama

ollamaModel

Type: string
Default: llama3.2

Name of an Ollama model that has been pulled on the Ollama server, for example llama3.2, phi3 or mistral.

For the file list metadata button the model must support images, for example llava or llama3.2-vision. Text-only models ignore the image.

ollamaUrl

Type: string
Default: http://ollama:11434

Base URL of the Ollama server, without /api. The default is the host name of an ollama container in a Docker Compose or DDEV setup. For a local installation use http://localhost:11434.

Keep API keys out of version control

config/system/settings.php is usually committed to Git. Do not store API keys there. Set them from an environment variable in config/system/additional.php instead:

config/system/additional.php

$GLOBALS['TYPO3_CONF_VARS']['EXTENSIONS']['t3_content_quality']['aiProvider'] = 'openai';
$GLOBALS['TYPO3_CONF_VARS']['EXTENSIONS']['t3_content_quality']['openaiApiKey']
    = (string)getenv('OPENAI_API_KEY');

Values set in additional.php override the values from the backend settings form.

AI button in the rich-text editor

The ✦ AI button needs the system extension rte_ckeditor. The extension does not change existing editor presets. Use one of these three ways to add the button:

Option 1: Use the preset of the extension

The extension registers the RTE preset t3_content_quality in ext_localconf.php. It is the TYPO3 core default preset plus the AI button at the end of the toolbar. Activate it in page TSconfig:

EXT:my_sitepackage/Configuration/page.tsconfig

RTE.default.preset = t3_content_quality

Option 2: Import the plugin into your own preset

Import EXT:t3_content_quality/Configuration/RTE/Plugin.yaml in your own preset:

EXT:my_sitepackage/Configuration/RTE/Default.yaml

imports:
  - { resource: 'EXT:rte_ckeditor/Configuration/RTE/Default.yaml' }
  - { resource: 'EXT:t3_content_quality/Configuration/RTE/Plugin.yaml' }

The TYPO3 YAML loader appends list entries in the order the files are loaded: imported files first, then the preset's own content. If your preset defines its own toolbar.items, the AI button therefore appears at the start of the toolbar. Use option 3 to place it elsewhere.

Option 3: Add the button manually

Add the module and the toolbar item to your preset where you want them:

EXT:my_sitepackage/Configuration/RTE/Default.yaml

editor:
  config:
    importModules:
      - { module: '@woit/t3-content-quality/CKEditor/ai-text-generator.js', exports: ['AiTextGenerator'] }
    toolbar:
      items:
        # … your existing toolbar items …
        - '|'
        - aiTextGenerator

See Configuration for details on RTE presets.

Page field: schema type

The extension adds the field Schema Type (Structured Data) (tx_t3contentquality_schema_type) to the Meta Tags palette on the Metadata tab of the page properties. It overrides the automatic schema.org type detection.

Options: (Auto-detect), Event, NewsArticle, TouristAttraction, LocalBusiness, Organization, FAQPage, JobPosting, Product.

Frontend JSON-LD middleware

The PSR-15 middleware woit/t3-content-quality/json-ld-injector is registered automatically for the frontend. It runs after typo3/cms-frontend/output-compression and before typo3/cms-frontend/send-response.

For every HTML response it looks up an approved schema for the current page and language in tx_t3contentquality_schema. If one exists, it inserts

<script type="application/ld+json">…</script>

directly before </head>. No TypoScript is needed. Pages without an approved schema are not changed.