Referenz:WikiRAG: Unterschied zwischen den Versionen

Versionsgeschichte interaktiv durchsuchen

Version vom 6. Oktober 2025, 10:29 Uhr

Übersicht
Beschreibung:	Erweiterungsschicht für wiki-gestützte LLMs
Status:	stable	Lizenz:	GPL-3.0-only
Entwickler:	Hallo Welt!	Distribution:	BlueSpice
Kategorie:	Kommunikation	Edition:	BlueSpice pro, BlueSpice farm, BlueSpice cloud 5.2+

Funktion

The extension WikiRAG provides data for RAG (Retrieval-Augmented Generation) applications from the wiki.

Data is not directly indexed to targets, but first scheduled to be processed by a job at regular intervals. Change observers are normally responsible for detecting indexable changes and scheduling jobs to process. Depending on what changed, change observers will only schedule certain DataProviders to be executed.

Types of pages being indexed

Not all pages are suitable for indexing. This determination is done in IndexabilityChecker (WikiRAG._IndexabilityChecker service and accessible over Scheduler::canPageBeScheduled method). By default, following pages are indexed:

Non-talk content pages
File pages where file types are OFFICE or TEXT (docx, doc, odt, pdf, txt, md, ...)
Any page allowed by WikiRAGCanBeIndexed hook

Prevent page index by MAGIC WORD

Use magic word on a page to prevent it from being indexed.

Technische Information[Bearbeiten | Quelltext bearbeiten]

Voraussetzungen[Bearbeiten | Quelltext bearbeiten]

MediaWiki: 1.43.0

Integriert in[Bearbeiten | Quelltext bearbeiten]

BlueSpiceDistributionConnector

BlueSpiceExtendedSearch

BlueSpiceVisualEditorConnector

WikiRAG

Spezialseiten[Bearbeiten | Quelltext bearbeiten]

Konfiguration[Bearbeiten | Quelltext bearbeiten]

Name	Wert
WikiRAGApiAllowedIP	NULL
WikiRAGPipeline	array ( )
WikiRAGTarget	array ( 'type' => 'null-target', 'configuration' => array ( ), )

Hooks[Bearbeiten | Quelltext bearbeiten]

Composer-Anforderungen

all-packages	mwstake/mediawiki-component-runjobstrigger

PDF-Ausschluss - Start

Feedback zur Dokumentation ist im Community-Forum möglich.

PDF-Ausschluss - Ende

@@ Zeile 7: / Zeile 7: @@
 |category=Kommunikation
 |bsvFrom=5.2
-|features=The extension '''ChatBot''' allows users to interact with the wiki from a chat window. The answers of the chatbot are related to the wiki content.
+|features=The extension '''WikiRAG''' provides data for RAG (Retrieval-Augmented Generation) applications from the wiki.
+Data is not directly indexed to targets, but first scheduled to be processed by a job at regular intervals. Change observers are normally responsible for detecting indexable changes and scheduling jobs to process. Depending on what changed, change observers will only schedule certain DataProviders to be executed.
+=== Types of pages being indexed ===
+Not all pages are suitable for indexing. This determination is done in <code>IndexabilityChecker</code> (<code>WikiRAG._IndexabilityChecker</code> service and accessible over <code>Scheduler::canPageBeScheduled</code> method). By default, following pages are indexed:
+* Non-talk content pages
+* File pages where file types are <code>OFFICE</code> or <code>TEXT</code> (<code>docx</code>, <code>doc</code>, <code>odt</code>, <code>pdf</code>, <code>txt</code>, <code>md</code>, ...)
+* Any page allowed by <code>WikiRAGCanBeIndexed</code> hook
+=== Prevent page index by MAGIC WORD ===
+Use <code>__NO_RAG_EXPORT__</code> magic word on a page to prevent it from being indexed.
 }}