Cowriter overview

The backend module AI > Cowriter (identifier cowriter_status, path /module/cowriter/status) shows what the cowriter does, whether each of its actions can run for you, how much it was used and where it is set up. Up to version 3.24 the same module was the administrator-only "Cowriter Status" page under Admin Tools; bookmarks and links to it keep working.

Who sees the module

Administrators see it without further setup. An editor sees it when one of their backend groups grants the module Cowriter in the field Modules (groupMods) of the group record.

An editor sees their own readiness, including the sites with their own configuration, their own usage and their own budget. An administrator also sees the usage of the whole installation, the extension settings and the links that fix a problem. The page never shows another person's name, prompt or text. The usage per person stays in AI > Operation > Analytics, which only administrators open.

What the cowriter is

A short description and a table of the places editors find the cowriter: the buttons of the cowriter preset in the rich text toolbar, the "Suggest" button next to text fields, the button in the record editor that suggests values for all fields, the image file control and the "Present as tabs or accordion" button. Each place says whether it is on and which setting of the extension configuration switches it.

The toolbar buttons appear in rich text fields whose RTE configuration lists them in its toolbar: the cowriter preset (for example RTE.default.preset = cowriter in the page TSconfig), or a configuration of your own that imports the Cowriter module and names its toolbar items. Which configuration a field uses depends on the page and is not checked on this page.

Readiness

One row per action, checked the way the action runs for the person viewing the page. Nothing is sent to a provider, so whether a provider answers is not checked; the provider list has a connection test for that.

Each row has one of five states:

Ready: The action finds everything it needs.
Partly ready: The action runs on some of its paths and fails on others, and the reason names which. An administrator gets a link to the module that fixes it; an editor is asked to contact an administrator.
Not ready: Something it needs is missing. The reason names it. An administrator gets a link to the module that fixes it; an editor is asked to contact an administrator.
Off: The feature is switched off in the extension configuration.
Not checked: The state cannot be known without making a request: the configuration chooses its model for each request, or the provider uses an adapter the check does not know. Such a row is never shown as ready.

What each action needs:

Write and rewrite in the dialog: The default configuration of the AI section, which the person may use (it is not limited to backend groups they are not in), with an active model whose provider is active and has an API key when its adapter needs one. When the person cannot use the default, the dialog's configuration picker may still offer one that runs: the row is then partly ready and names it, because writing works only after picking it.
Field suggestions: The same default configuration. Field suggestions have no configuration picker, so without a usable default they are not ready.
Tasks: Every active task the dialog offers, checked the way a task runs: on its own configuration, or on the default when it has none. The row counts the tasks that cannot run and names them when there are at most three. It is partly ready when some tasks run and others do not, and off when there is no active task. A configuration picked in the dialog replaces the task's own one, so when the picker offers the person a configuration that runs, failing tasks still run once it is picked: the row is then partly ready and names that configuration.
Describe an image and translate: nr-llm's generic call, which runs on the default configuration only when that has a model and is not limited to backend groups. Image descriptions also need a model marked as able to read images. Translation uses a specialized translator such as DeepL first when one is available; the language pairs it does not support go to the generic call. Without a default the generic call can run on, translation is partly ready.
Generate an image: The setting image.enable, the OpenAI API key in nr-llm's extension configuration and a default upload folder the person may write to.
Inline suggestions: The setting inlineSuggestions.enable and the configuration it names, or the default configuration.
Tabs or accordion: The setting presentation.enable, the tab and accordion elements of bootstrap_package and the default configuration.

A second table lists every site that names its own configuration in its site settings, checked with the access of the person viewing the page. An editor sees only the sites their web mounts reach: a mount on the root page, above it or anywhere below it. A task without its own configuration, run on a record of such a site, uses that configuration instead of the default.

Configuration

Administrators see every setting of the extension configuration with its value and whether it is the default or was changed. The settings are changed in Settings > Extension Configuration > t3_cowriter, which only system maintainers open.

Usage

The period is 7, 30 or 90 days, chosen with the form above the figures. It starts at 00:00 of its first day and ends now.

Estimated cost, requests and tokens: From nr-llm's daily usage record of the cowriter (source extension t3_cowriter).
Actions run and failed, per action: From nr-llm's telemetry. A task is listed with its name; an action name the cowriter does not know is shown as it is stored. The telemetry is deleted after nr-llm's retention period (privacy.retention.telemetry, by default 30 days). When the period reaches further back, the page says from which day the counts per action are complete. When the telemetry is switched off, the page says so.
Your AI budget: Your own usage of all AI features today and this month against your personal limits, from nr-llm's budget.
Shared prompts: For administrators: how many shared prompts wait for approval, with a link to the List module.

Set up and follow up

Links to the modules of the AI section (Configurations, Tasks, Models, Providers, Analytics, AI tasks), to the extension configuration and to this documentation. A link is shown only when the person may open the module.