--- title: "Configuration" manual: "AI Cowriter" version: "3.19" permalink: "https://docs.typo3.org/permalink/netresearch/t3-cowriter:configuration@3.19" source: "Configuration/Index.rst" rendered: "2026-10-29T22:05:46+00:00" --- # Configuration {#configuration} ## LLM provider setup {#llm-provider-setup} The Cowriter extension uses the [nr-llm extension](https://docs.typo3.org/p/netresearch/nr-llm/main/en-us/Index.html#start) for LLM provider configuration. Configure your preferred provider in the [nr-llm backend module](https://docs.typo3.org/p/netresearch/nr-llm/main/en-us/Administration/Index.html#administration-backend-module). ### Supported providers {#supported-providers} - **OpenAI** \- GPT-5.x series, o-series reasoning models - **Anthropic Claude** \- Claude 4.5 Opus, Sonnet, Haiku - **Google Gemini** \- Gemini 3 Pro, Flash - **OpenRouter** \- Access to 100+ models from multiple providers - **Mistral** \- Mistral Large, Medium, and open models - **Groq** \- Ultra-fast inference with Llama, Mixtral ### Configuration steps {#configuration-steps} 1. Navigate to **Admin Tools** \> **LLM Management** 1. Add a provider (e.g., OpenAI with your API key) 1. Create a model configuration 1. Create an LLM configuration and set it as default > [!TIP] > See the > [nr-llm provider configuration](https://docs.typo3.org/p/netresearch/nr-llm/main/en-us/Configuration/ProviderFields.html#configuration-provider) > for detailed provider configuration options. ## RTE configuration {#rte-configuration} There are two ways to configure the CKEditor integration: ### Option 1: Using the shipped preset {#option-1-using-the-shipped-preset} If you don't have a custom RTE configuration, point the RTE at the preset this extension registers: 1. Go to your root page 1. Open **Page Properties** \> **Page TSconfig** 1. Add this line: ```typoscript RTE.default.preset = cowriter ``` The preset is inherited by every page below, and brings all Cowriter toolbar items with it. ![Page TSconfig configuration in TYPO3 v14](../Images/pagetsconfig.png) ### Option 2: Custom RTE configuration {#option-2-custom-rte-configuration} If you have your own RTE configuration file (`your_ext/Configuration/RTE/YourConfig.yaml`), add the Cowriter module: ```yaml editor: config: importModules: - { module: '@netresearch/t3_cowriter/cowriter', exports: ['Cowriter'] } toolbar: items: - cowriter - cowriterVision - cowriterTranslate - cowriterTemplates - cowriterImage - cowriterInlineSuggestions ``` The six toolbar items are: - **`cowriter`** Main dialog — task-based content generation with preview - **`cowriterVision`** Generate image alt text via LLM vision analysis - **`cowriterTranslate`** Inline translation dropdown with the languages of the record's site, see [Translation languages and formality](https://docs.typo3.org/permalink/netresearch/t3-cowriter:configuration-translation-languages@3.19) - **`cowriterTemplates`** Open the Cowriter dialog with a task pre-selected - **`cowriterImage`** Generate an image from a description and insert it, see [Image generation](https://docs.typo3.org/permalink/netresearch/t3-cowriter:configuration-image-generation@3.19). The button appears only while `image.enable` is on, and it works only together with the image plugin of `netresearch/rte-ckeditor-image`. - **`cowriterInlineSuggestions`** Lets each editor switch inline suggestions off and on again, see [Inline suggestions](https://docs.typo3.org/permalink/netresearch/t3-cowriter:configuration-inline-suggestions@3.19). The button appears only while `inlineSuggestions.enable` is on. The suggestions themselves do not need it: without the button in the toolbar they are on for every editor. > [!TIP] > You can include only the toolbar items you need. For example, if you > only want the main dialog and translation, omit `cowriterVision` > and `cowriterTemplates`. > [!WARNING] > An explicit `toolbar.items` list REPLACES the toolbar, it does not add > to it. Importing the module and then listing only `cowriter` gives you > exactly one button — the others are registered but have nowhere to > appear. If buttons are missing, check this list before suspecting the > extension. ## Translation languages and formality {#configuration-translation-languages} The **Translate** dropdown offers the languages of the site the edited record belongs to: every language of the site configuration, labelled with its title. Languages hidden from the frontend (`enabled: false`) are included, so editors can translate into a language before it goes live. When the record belongs to no site (a record on the root page, pid 0, or in a folder outside every site), the dropdown keeps its fixed list: German, English, French, Spanish, Italian, Dutch, Portuguese, Polish, Japanese and Chinese. The same list appears when none of the site's languages has a target code (for example a site whose only language is `fil_PH`, see below). The target language is taken from the language's `locale`: - The two-letter language code is used, the region is dropped: `de_DE.UTF-8` and `de_CH.UTF-8` both translate into `de`. - English and Portuguese keep the region, because DeepL translates into their regional variants: `en_GB` gives `en-GB`, `en_US` gives `en-US`, `pt_BR` gives `pt-BR`, `pt_PT` gives `pt-PT`. Other regions of these languages (`en_AU`) give `en` or `pt`. - Chinese gives `zh`. DeepL translates `zh` into Simplified Chinese; the language model receives only "Chinese" and chooses the script itself. The regions that write Traditional Chinese keep their region: `zh_TW` gives `zh-TW`, `zh_HK` gives `zh-HK`, `zh_MO` gives `zh-MO`; a locale with the script `Hant` and no region gives `zh-TW`, one with the script `Hans` gives `zh`. DeepL has no target for these codes, so nr-llm translates them with the language model. - A language without a two-letter code (`fil_PH`) is not offered, nor is the `C` or `POSIX` locale. ### Formality {#formality} Each site language can set the formality of translations into it with the key `cowriterFormality`: - **`default`** No preference (the default when the key is missing). - **`formal`** Formal address, for German the "Sie" form. - **`informal`** Informal address, for German the "du" form. Any other value counts as `default`. DeepL applies the formality for the languages it supports it for (with nr-llm 0.41: German, French, Italian, Spanish, Dutch, Polish, Portuguese, Russian, Japanese) and ignores it for the others; with a language model as translator it becomes an instruction to keep that tone. Set the key in the Sites module (**Site Management** \> **Sites**, the language's field **Cowriter: formality of translations**) or directly in the site configuration: **config/sites/\/config.yaml** ```yaml languages: - languageId: 0 title: English locale: en_US.UTF-8 base: / - languageId: 1 title: Deutsch locale: de_DE.UTF-8 base: /de/ cowriterFormality: formal ``` With this configuration the dropdown offers "English" (`en-US`) and "Deutsch" (`de`), and German translations use the formal address. > [!NOTE] > Saving a site in the Sites module keeps only the language keys that > have a field in the module. The extension adds a field for > `cowriterFormality`, so the setting survives an edit there. ### Revising DeepL translations {#configuration-translation-polish} A translation without a pinned LLM configuration goes to a specialized translator such as DeepL when nr-llm has one for the language pair. A language model can then revise that translation for fluency, terminology and tone (nr-llm's "polish", ADR-233). Both settings are in the extension configuration (**Admin Tools** \> **Settings** \> **Extension Configuration** \> `t3_cowriter`); a request from the editor cannot switch the revision on or change its instruction. - **translation.polish** - *Type:* boolean - *Default:* `0` Switches the revision on. It applies only when the specialized translator translated the text. A translation that a language model writes anyway (no specialized translator for the pair, or a pinned configuration) is not revised a second time. - **translation.polishInstruction** - *Type:* string - *Default:* empty An additional instruction for the revising model, for example `Address the reader directly.`. It is added to nr-llm's built-in revision instruction, which keeps meaning, formality, markup, placeholders and glossary terms. An empty value uses the built-in instruction alone. Costs and fallback: - A revised translation uses two requests, the translator's and the language model's. Both count against the editor's [nr-llm budget](https://docs.typo3.org/p/netresearch/nr-llm/main/en-us/Administration/UserBudgets.html#administration-user-budgets). - The Cowriter does not use nr-llm's translation cache. Every translation with the revision sends both requests, also when the same text was translated before. - The revision never fails a translation. When the language model fails, the budget refuses the second request, or the answer changes markup, placeholders or glossary terms, the editor gets the translator's text unchanged. The response of the translate route carries `polished` (`true` or `false`) when the setting is on and the specialized translator translated, and `polishReason` with nr-llm's reason when the revision was not applied (for example `budget_refused` or `markup_changed`). Without the setting the response is unchanged. ## Image generation {#configuration-image-generation} With **Cowriter - Generate image** an editor describes an image, nr-llm's OpenAI image service generates it, and the Cowriter stores it as a file and inserts it into the text. File fields such as the images of a content element get a **Generate image** button as well ([In file fields](https://docs.typo3.org/permalink/netresearch/t3-cowriter:configuration-image-generation-file-fields@3.19)). The feature is off until an administrator switches it on in the extension configuration (**Admin Tools** \> **Settings** \> **Extension Configuration** \> `t3_cowriter`). A request from the editor cannot switch it on or choose the model. - **image.enable** - *Type:* boolean - *Default:* `0` Shows the toolbar button and the button in file fields, and opens the route. While it is off, the button is removed from every CKEditor toolbar, file fields do not get it, and the route answers with HTTP 403. - **image.model** - *Type:* string - *Default:* empty The OpenAI image model, for example `gpt-image-2` or `gpt-image-2.5-flare`. Empty uses the active image model that nr-llm marks as default in its model list (**Admin Tools** \> **LLM** \> **Models**), or `gpt-image-2` when there is none. The Cowriter does not read nr-llm's setting `image.dalle.defaultModel`. Requirements and behaviour: - The image plugin of `netresearch/rte-ckeditor-image` (extension key `rte_ckeditor_image`) must be loaded in the editor configuration, for example by importing its plugin configuration into the preset: ```yaml imports: - { resource: 'EXT:rte_ckeditor_image/Configuration/RTE/Plugin.yaml' } ``` The image is inserted as that plugin's image with its file reference (`data-htmlarea-file-uid`), like an image chosen in the file browser. TYPO3's CKEditor has no image plugin of its own, and the core RTE processing removes a plain `` on save (`denyTags: img` in `EXT:rte_ckeditor/Configuration/RTE/Processing.yaml`). Without the plugin the button is shown disabled, says why, and requests nothing. - An OpenAI API key in the nr-llm extension configuration (`providers.openai.apiKeyIdentifier`). - The editor needs a folder to store the image in: the folder TYPO3 uses for the editor's uploads (user TSconfig `options.defaultUploadFolder`, otherwise the first default folder of a storage the editor may write to). The storage must be public. Without such a folder the request is refused before the image is generated, so nothing is paid for. - The file is added with the editor's own file permissions and file mounts. Its name is made from the first words of the description and a random suffix, for example `a-lighthouse-at-dusk-3f9c2a1b.png`. - Only the image data the model returns is stored, never a URL. It must be a PNG, JPEG or WebP image of at most 10 MB; anything else is refused. - Each image counts against the editor's [nr-llm budget](https://docs.typo3.org/p/netresearch/nr-llm/main/en-us/Administration/UserBudgets.html#administration-user-budgets). nr-llm 0.41 has prices for `gpt-image-2`, `gpt-image-1.5`, `gpt-image-1-mini`, `gpt-image-1` and the DALL·E models, but none for the `gpt-image-2.5` models. Unless their record in nr-llm's model list carries a price, their images count with cost 0 against a cost limit; request limits apply in every case. - A generation can take one to two minutes; nr-llm waits up to 300 seconds for OpenAI. What cuts such a request off is PHP-FPM's [request_terminate_timeout](https://www.php.net/manual/en/install.fpm.configuration.php) and the read timeouts of the web server and of proxies in front of TYPO3; allow them more than that. PHP's `max_execution_time` does not count the wait on Linux, because it measures only the script's own execution and not stream operations such as the request to OpenAI ([PHP manual, set_time_limit()](https://www.php.net/manual/en/function.set-time-limit.php); on Windows the real time counts). A request that is cut off inserts nothing. Whether the image was stored depends on when it was cut off: after the image was generated and stored, the file is in the upload folder; before that, there is none. The editor's message says that the image may already be in the upload folder. - The editor may close the dialog while the image is generated. The request goes on, and the image is inserted where the cursor was when the editor clicked the button, also when the text was edited in the meantime. This holds as long as the record stays open: saving and closing the record, or leaving the page, reloads the form, and the answer is no longer awaited. The image is then stored in the upload folder but not inserted; the file field control behaves the same. - The description becomes the image's alternative text and the editor is told to check it. Extensions that react to new files, such as one that writes alternative texts on upload, also run for a generated image. ### In file fields {#configuration-image-generation-file-fields} File fields in the record editor (TCA `type` `file`, for example the images and media of a content element or the media of a page) get a **Generate image** button below their file list. It uses the same route, setting, folder and checks as the rich text editor. The stored file is added to the field as a new file reference, exactly as if the editor had selected it in the file browser; it is kept when the record is saved. The rich text editor's image plugin is not needed for this. The button is shown only when all of these hold: - `image.enable` is on. - The editor has a folder to store the image in (see above). TYPO3's own upload button may use a different folder for a field, set with user TSconfig `options.defaultUploadFolder` per table and field; the generated image always goes to the editor's general upload folder. - The field accepts PNG files by its `allowed` and `disallowed` lists. The route cannot ask the model for another format (nr-llm 0.41 has no option for it), and OpenAI's image models return PNG, so a field that takes only JPEG or WebP gets no button. Should a model still return another format the field does not accept, the image is stored but not added, and the editor is told so. - The field is not read-only, and its appearance does not switch off the add controls (`showFileSelectors: false`) or uploads (`fileUploadAllowed` set to a false value such as `0`, also from TSconfig). When the field holds `maxitems` files, the button is hidden, like TYPO3's own add buttons, and it comes back when a file reference is removed. It is built on the PSR-14 event `\TYPO3\CMS\Backend\Form\Event\CustomFileControlsEvent`, which TYPO3 13.4 and 14.3 both dispatch. The new file reference gets no alternative text of its own. Like a file selected from the file list, it shows the alternative text of the file's metadata, which the Cowriter does not write; an extension that fills it for new files fills it for a generated image as well. ## Inline suggestions {#configuration-inline-suggestions} While an editor writes in the rich text editor, the language model can propose how the text goes on. When the editor stops typing at the end of a paragraph, the proposal appears as grey text after the cursor; `Tab` inserts it, any other key, a click or leaving the editor discards it. The feature is off until an administrator switches it on in the extension configuration (**Admin Tools** \> **Settings** \> **Extension Configuration** \> `t3_cowriter`). A request from the editor cannot switch it on or choose the configuration. - **inlineSuggestions.enable** - *Type:* boolean - *Default:* `0` Switches inline suggestions on. While it is off, the editor sends no request and shows nothing, the toggle button is removed from every CKEditor toolbar, and the route answers with HTTP 403. - **inlineSuggestions.delay** - *Type:* int - *Default:* `1200` How long the editor waits after the last key before it asks, in milliseconds. Values below 800 count as 800, values above 3000 as 3000. - **inlineSuggestions.minChars** - *Type:* int - *Default:* `30` How many characters the paragraph needs before suggestions start, from 10 to 500. The route refuses shorter texts as well. - **inlineSuggestions.configuration** - *Type:* string - *Default:* empty The identifier of the nr-llm configuration that answers (**Admin Tools** \> **LLM** \> **Configurations**). Empty uses the default configuration. A suggestion is only useful when it arrives within a second or two, so a fast and inexpensive model suits it best. The editor must be allowed to use the configuration; otherwise the route answers with HTTP 403 and the editor stops asking until the page is reloaded. When a suggestion is asked for: - Only after typing. Pasting, undo, a dialog's result or an accepted suggestion do not trigger one. - Only when the cursor is at the end of a paragraph or heading without a selection, the editor has the focus, is not read-only and not in source mode, and no input method composition (IME) is in progress. Never in a list item or a table cell: there `Tab` indents the item or moves to the next cell, and a suggestion would compete with it. - Each new key cancels the wait and aborts a running request; an answer that arrives after the text or the cursor changed is dropped. - Each rich text field sends at most one request every five seconds. The route allows 12 per minute per backend user, across all fields and tabs, counted apart from the other Cowriter actions, so typing cannot use up their limit. After a refusal for the rate limit the field waits as long as the route says (`Retry-After`); after another failure it waits 30 seconds. Failures are not shown to the editor. What is sent and returned: - The request carries the text before the cursor as plain text: the current paragraph and the paragraphs before it, at most the last 1000 characters. The route refuses more. With a saved record it also carries the record's table and uid; the server looks up the record's site language with the editor's permissions and asks for a continuation in that language. Without one, the model keeps the language of the text. - The model is asked for at most one sentence of plain text (at most 60 tokens, temperature 0.3, stopping at a line break). The route strips markup, entities, line breaks, control and direction characters and Markdown markers, keeps the first sentence and at most 200 characters, and puts a space in front when the text does not end in one. Only what looks like a tag is removed, so `a