Configuration

T3AS uses AI Foundation for shared provider setup, model selection, prompts, and core AI services. Complete the parent setup first, then review the T3AS-specific search and training settings below.

Helpful AI Foundation references:

1. Extension Settings

Configure the shared AI Foundation settings first, then return to T3AS for search, training, and source-specific setup.

  1. Go to the TYPO3 backend.

  2. Open AI Foundation > AI Features.

  3. Configure the shared AI Foundation settings.

  4. Return to T3AS to continue with source, training, and search-specific settings.

AI Features

T3AS focuses on AI search, training, and answer delivery on top of the shared AI Foundation setup. Use these features when you want to connect project content to AI search, control answer behavior, and monitor how search performs after rollout.

Key T3AS capabilities include:

  • Search and answer generation based on trained project content

  • Data source syncing and training queue management

  • Scheduler-based background processing

  • Usage analytics, logs, and request statistics

  • Prompt-controlled search answers and instructions

For shared model routing and central AI behavior, see AI Foundation AI Features.

2. Dashboard

Purpose

The Dashboard gives an overview of your AI training pipeline for the current site.

What you see

  • Which AI/embedding model is in use (e.g. OpenAI, Gemini, Mistral, Custom).

  • Status of the Search and Chatbot modules (if installed): active/inactive, AI engine, base model, embedding model.

  • Status of your data sources and training (e.g. how many items are pending, completed, or failed).

  • Training Pipeline section: data sources count, queue size, and a link to CLI reference.

  • Usage Analytics summary (e.g. total interactions, search queries, chat sessions over the last 7 days).

  • A link to the Scheduler to run or check the automatic training task.

Scheduler link

From the Dashboard you can open the TYPO3 Scheduler and locate the automatic training task (typically named T3AS Training for this site). Use Run All or Run Task Now to process the training queue immediately.

3. Data Source

Purpose

Add and manage the sources of content that will be used for AI search, such as website pages, PDFs, and Q&A records.

Adding a data source

  1. Click + Add Source.

  2. Choose the type of source, for example:

    • Sitemap XML – Your sitemap URL(s) (e.g. https://example.com/sitemap.xml).

    • PDF Documents – Folder path where PDFs are stored (and optionally upload PDFs).

    • TYPO3 Pages – Content from specific TYPO3 pages.

    • Web Pages – A website URL; optionally limit to a path (e.g. https://example.com/blog/*).

    • Q&A Pairs – Manual question-and-answer content.

    • Indexed Search / Ke Search / Solr – If the corresponding extensions are installed and indexed content is available.

  3. Fill in the requested details (URLs, folder path, page selection, etc.) and give the source a Name (e.g. Main Website) and optional Description.

  4. Set Sync interval: how often content should be refreshed (e.g. Hourly, Daily, Weekly). Custom means no automatic schedule (manual sync only).

  5. Set Used by (formerly Type) to control where this source is available.

  6. Set Enabled to on if the source should be active.

  7. Click Save.

After saving, T3AS will:

  • Create or update the data source.

  • Sync content into the training queue (new or changed items).

  • Automatically create the T3AS Training Scheduler task for this site (if it does not exist yet) and run it at the frequency you set (e.g. Hourly, Daily, Weekly). You do not need to create the scheduler task manually—it is created when the source is saved and will execute according to the chosen sync interval.

Editing or deleting a source

Use the actions next to each data source to edit (type, URLs, interval, usage) or delete it.

Source Groups | Page-Level Source Configuration

Source Groups allow administrators and editors to organize data sources and control which content is available to AI Search and the AI Chatbot on a page-by-page basis. This gives you precise control over retrieval because only the selected Source Groups are considered for that page.

Source Groups are used only during content retrieval. They do not affect AI training, vector indexing, Scheduler execution, embedding generation, or data synchronization. The filtering is applied by the VectorService during retrieval.

Backend Configuration

Manage Source Groups

Go to Data Sources -> Source Groups to manage Source Groups. Administrators can:

  • create Source Groups

  • edit Source Groups

  • delete Source Groups

Note

The Global Source Group is a system group and cannot be edited or deleted.

Assign Source Groups to Data Sources

When creating or editing a data source, users can:

  • select one or more Source Groups

  • rely on the Global Source Group, which is assigned automatically

  • configure the Used by field (formerly Type) to control where the source is available

Available Used by options:

  • AI Search

  • AI Chatbot

  • Both AI Search and AI Chatbot

This setting controls where the data source can be used after retrieval starts.

Page-Level Configuration (T3AS / T3AC)

  1. Open the desired TYPO3 page.

  2. Open Page Properties.

  3. Go to the AI Chatbot tab.

  4. Find the Source Groups field.

  5. Select the Source Groups that should be available on that page.

Only data sources assigned to the selected Source Groups are used on that page. This filtering applies to both AI Search and AI Chatbot, and different pages can use different Source Groups.

Practical Example

A company has separate documentation for Products, HR, and Internal Policies.

Three Source Groups are created:

  • Products

  • HR

  • Internal

On product pages, only the Products Source Group is selected, so AI Search and AI Chatbot return product-related information. On HR pages, only the HR Source Group is selected, so HR content is used while product information is excluded.

Note

Source Groups only affect content retrieval. Existing training data, embeddings, vector indexing, and scheduled synchronization continue to operate normally.

Note

If no custom Source Groups are selected, the Global Source Group remains available according to the configured behavior.

Header / Footer for Sitemap and Web-Page Source Types

These options help you keep repeated layout content out of the main page body while still making shared site information available to AI retrieval.

Index Site Header

Use Index Site Header when the site header contains useful shared information that should be indexed only once.

Purpose:

  • extract the website <header> content one time for each unique header

  • store that content as a separate training item

  • remove the same header content from the page body before indexing

Benefits:

  • reduces duplicate information across many pages

  • keeps repeated navigation or shared header text from being indexed over and over

  • preserves useful shared site information for retrieval

When to enable:

  • when the header contains meaningful text that supports AI answers

  • when many pages share the same header content

Recommended use cases:

  • websites with shared product navigation or service overviews in the header

  • websites where the header contains reusable company or category information

Index Site Footer

Use Index Site Footer when the site footer contains shared information that should be indexed only once.

Purpose:

  • extract the website <footer> content one time for each unique footer

  • store that content as a separate training item

  • exclude that footer content from the page body during indexing

Benefits:

  • prevents duplicate footer text from being trained again on every page

  • keeps the main page content cleaner for retrieval

  • preserves useful global site information such as company details or support links

When to enable:

  • when the footer contains helpful shared text for search or chatbot answers

  • when the same footer appears on many pages

Recommended use cases:

  • websites with shared contact details, policy references, or company summaries in the footer

  • large sites where repeated footer content would otherwise be indexed many times

Important

Deleting a data source also removes its training queue and embedded data for that source.

Sync

Sync (per source or Sync all) refreshes content from the source into the training queue.

  • Sync does not run AI training by itself.

  • Training is performed by the Scheduler task or manually (see Training Center).

Scheduler

When you create a data source, T3AS will automatically create the T3AS Training Scheduler task for this site (if it does not exist yet) and run it at the frequency you set (e.g. Hourly, Daily, Weekly). You do not need to create the scheduler task manually—it is created when the source is saved and will execute according to the chosen sync interval.

From the Dashboard you can open the TYPO3 Scheduler and locate the automatic training task (typically named T3AS Training for this site). Use Run All or Run Task Now to process the training queue immediately.

Command-line tools

What is command-line scheduler execution?

Command-line scheduler execution means running the TYPO3 Scheduler task directly from your terminal. This is useful when you want to process the training queue immediately, without waiting for the next scheduled interval.

How to run the scheduler task

  1. Open your terminal in the TYPO3 project root (where DDEV is available).

  2. Identify the correct T3AS Training task in the TYPO3 Scheduler first.

  3. Run that task manually from TYPO3 or execute it from the command line with the task ID from your own instance.

  4. Check the terminal output to confirm the task completed successfully.

If you are using DDEV, you can run the T3AS training task manually from the command line after confirming the correct task ID in System -> Scheduler.

Legacy installation:

ddev typo3/sysext/core/bin/typo3 scheduler:execute --task=<your-task-id>

Composer installation:

ddev typo3 scheduler:run --task=<your-task-id>

What happens after running the command?

  • TYPO3 executes the selected T3AS Training scheduler task.

  • Items in the training queue are processed.

  • You can verify progress and completion directly in the command output.

Example output

Example command-line output for TYPO3 scheduler task run

Example of scheduler task output in the terminal.

Example command-line output showing queue processing and training completion

Example showing queue processing and training completion summary.

For more options and details, see the TYPO3 Scheduler documentation: https://docs.typo3.org/c/typo3/cms-scheduler/13.4/en-us/Administration/ConsoleTools/Running.html#providing-options-to-the-shell-script

4.Training Center

Purpose

View the training queue (items collected from all data sources) and control training and cleanup.

What you see

  • Summary counts: Total items, Pending, Embedding (processing), Completed, Failed, and Tokens used.

  • All Sources: List or summary of data sources and item counts.

  • Training Queue table: Items with columns such as Item, Status, Tokens, Created, Actions.

  • Filters: By data source, status (All Statuses), or search text.

  • Actions: Select All, Delete, Re-queue (reset).

Queue item statuses:

  • Pending – Waiting to be processed.

  • Processing / Embedding – Currently being sent to the embeddings service.

  • Completed – Successfully trained.

  • Failed – Error during training.

Actions

Sync

Refreshes content from the data source into the queue (same as in the Data Source tab).

Reset (Re-queue)

Puts a failed or completed item back to Pending so it will be processed again on the next training run.

Delete

Removes selected queue items.

Important

Deleted items will not be trained again unless they are added again by a new sync.

Run training

Use the link to the Scheduler module and run the T3AS Training task for this site.

When the task runs, it:

  • Processes all Pending queue items (generates embeddings).

  • Runs cleanup of old completed/failed items according to the retention setting.

Training behaviour (simple terms)

  • Only items in status Pending are processed when training runs.

  • Processing means: the text is sent to the configured embeddings service, and the result is stored for search usage.

  • After success, the item is marked Completed; on error, Failed.

  • How often training runs depends on the Sync interval of your data sources and on the Scheduler actually being triggered (e.g. via cron).

5. Search tab

The Search tab controls AI search for the whole site. Here you turn search on, set how answers look, style the widget, and manage suggested questions. Settings on a single page plugin can override these defaults.

Step 1: Open the T3AS module.

Step 2: Click the Search tab.

Step 3: Configure Settings, Widget, and Questions.

Settings

Turn AI search on and control answer behaviour.

  • Enable AI Search Globally — Activates AI search across the site

  • Enable Reference Links — Shows source links below the answer (pages, PDFs, etc.)

  • Enable Voiceover — Adds a play button so visitors can hear the answer

  • Enable Search Feedback — Shows thumbs up/down; ratings appear in Usage Analytics

  • Enable Chatbot Mode — Lets visitors ask follow-up questions

  • Result Style — Summarize (short) or Long Answer (detailed)

  • Search Class — CSS class for Ke Search and similar integrations (see Injecting AI Search result in TYPO3 Search Extensions)

  • Instructions — Custom rules for how the AI should write answers

Widget

Style the search box and floating trigger button (for modal or floating layouts).

  • Widget Mode — How the widget opens (e.g. Modal Box (Centered))

  • Widget/Modal Trigger Button Position — Trigger button position (e.g. Left (bottom))

  • Trigger Button Size — Size of the floating trigger button

  • Search Icon — Icon on the search box or trigger

  • Trigger Button Background — Background style of the trigger button

  • Select Style — Default Style (site colours) or Customized Style (plugin)

  • Border Radius — Corner roundness (e.g. Semi Rounded)

  • Select Loader — Skeleton Loader or Typing Loader while the answer loads

  • Primary Color / Secondary Color / Text Color — Only when Customized Style is selected

  • Search Form Type — Input layout (e.g. With Button)

  • Button Type — Search Icon or With Label

Questions

Set up clickable question suggestions in the search box.

  • Predefined Questions — Enable suggested questions

  • Question Position — Where they appear (e.g. Bottom Search)

  • Number of Questions to Show — How many to display (e.g. 5)

  • Questions Storage Folder(s) — Page ID of the folder with question records (e.g. 681)

Note

These settings apply site-wide. To override them on one page, use the T3AS Search frontend plugin. See T3AS Search Plugin.

6. Usage Analytics

The Usage Analytics tab records visitor search activity. You can see what was searched, what answer was given, feedback ratings, and reference links used.

Step 1: Open the T3AS module.

Step 2: Click Usage Analytics.

What you see

Each row in the log list shows:

  • Search term — What the visitor typed

  • AI answer — Short summary of the result

  • Module — e.g. Search

  • Feedback — Thumbs up or down (when Search Feedback is enabled)

  • Reference links — Number of source links shown

  • Page, language, time — Where and when the search happened

Open a log entry

Click a row to see the full detail:

  • Negative feedback and any visitor comment

  • Search badge — Click to filter logs by that search term

  • Reference sources — Pages or files used to build the answer

  • All messages — Full chat history (when Chatbot Mode is on)

  • Delete This Log — Remove a single entry

Filter and export

  • Search queries or responses — Find text in the logs

  • All Modules — Filter by module (e.g. Search only)

  • All Languages — Filter by language

  • Export — Download log data as a file

Note

Enable Search Feedback in Search → Settings or in the plugin Search Results tab to collect thumbs up/down ratings.

When no data exists yet: “No interaction logs yet. Search and search history will appear here when the modules are loaded and users interact.”

7. AILogs

Purpose

View log entries for the current site, including sync, training, and error events.

What you see

  • Search: Use the search box (for example: Search in message...) to find specific log text.

  • Channel: Filter by channel (default: [all]).

  • Level: Filter by log level (for example: Any, Error, Warning, Info).

  • Max rows: Set how many rows are shown per page (default: 50).

  • Entry count: The page shows a summary like Showing up to 50 of 745 entries per page.

  • Log table columns: - Time - Level - User - Details

8. AI Statistics

Purpose

View AI API usage statistics for the current site.

What you see

  • API Usage summary for your search activity.

  • API Requests count.

  • Tokens usage details: - Total tokens - Context tokens - Generated tokens

9. AI Prompts

Use AI Prompts to control how T3AS writes answers, summaries, and search-related responses. This is useful when you want search output to follow a consistent tone, answer style, or instruction set across the whole site.

Best practices:

  • Keep instructions focused on answer quality, tone, and length.

  • Test prompt changes with real user questions.

  • Review AI Foundation AI Prompts when you want shared prompt behavior across multiple AI Universe extensions.

10. Providers & MCP Tools

T3AS uses AI Foundation for shared provider setup and MCP-based integrations. Review this area when you need to confirm that the correct provider, model, and MCP capabilities are available for search and training workflows.

See also:

Solr Settings

If the selected search engine is Solr, please provide the following details in case your Solr server is secured with HTTP authentication.

  • Solr Username Username for Solr authentication.

  • Solr Password Password for Solr authentication.

Hosted-Solr Server Integration for Solr

The Hosted-Solr Server feature enables you to connect your TYPO3 instance directly to the Hosted-Solr service for improved search indexing and data retrieval. The following guide outlines how to configure T3AS and Hosted Solr in your TYPO3 instance using the Site Configuration module.

  • Step 1: Open Site Configuration

  1. In the TYPO3 backend, navigate to Site Management → Sites

  2. Edit site configuration

  • Step 2:Scroll down to the Solr section in the same T3AS tab and Configure Solr Integration

  • Step 4: Define Solr Connection Settings

Extension Banner

  • Switch to the Solr tab within the Site Configuration.

  • Enable the Enable Solr for this site option.

  • Provide the connection details:

    • Scheme: http or https

    • Host: Enter your Hosted-Solr host address (e.g., 562d8a85dc0-icy-tree-111:eecbaaae879b@node-14.hosted-solr.com)

    • Port: Usually 443 for secure connections

    • URL Path: Provide the Solr path without /solr/ (e.g., /562d8a85dc0-icy-tree-111/)

  • Step 5: Save Configuration

Getting Started

  1. Create an Account: Sign up on the Hosted-Solr platform using your email address.

  2. Create a Solr Core: Once your account is active, create a new Solr core.

  3. Configure in TYPO3: Add your Solr core connection details within the TYPO3 Site Settings.

This integration allows you to seamlessly manage your Solr configuration and maintain consistent communication between TYPO3 and the Hosted-Solr environment.

Verifying the Connection

After configuration, ensure that Solr is properly connected:

  • Navigate to the Info module inside the Solr tab within your TYPO3 backend.

  • Verify that the Solr connection status and indexing information appear correctly.

Additional Fields Support

The extension now supports fetching additional fields from Solr beyond the standard predefined set.

This means you can include custom or project-specific fields in your search configuration to enhance indexing and display flexibility.

Configuration Steps

  1. Go to the Extension Settings in TYPO3.

  2. Specify which fields should be retrieved from Solr.

  3. Save your settings to enable greater control over search results and data output.

By leveraging this feature, you can tailor your Solr-based search experience to match the exact needs of your TYPO3 project.

Note

Several default fields are automatically included for content retrieval from Solr: id, site, type, uid, content, pid, url, changed, and access. Ensure that Solr is properly configured and that the content field is available in your Solr-indexed data.

Injecting AI Search result in TYPO3 Search Extensions

This integration allows you to inject AI-generated results into an existing search form

Note

Make sure you have added the Search Class in the T3AS Search > Settings tab as described in 5. Search tab.

For example, If you want to render the AI search result on below the ke_search form, Please add below code at

EXT:ke_search/Resources/Private/Templates/SearchForm.html

<f:cObject typoscriptObjectPath="lib.injectAiSearchResults" />
Extension

Enable AI Search plugin using TypoScript

To enable this, Add the following TypoScript object to any template where you want the AI Search plugin to be rendered. The searchPid value should be set to the page ID where the plugin is placed (for example: 4).

Add the following TypoScript object to any template where you want the AI Search plugin to be rendered:

<f:cObject typoscriptObjectPath="lib.renderAiSearchPlugin" data="{searchPid:4}"/>