Extension:Wanda

**MediaWiki extensions manual**
Wanda; Release status: stable
Implementation	Artificial intelligence, API, Data extraction, Page action, Search, Special page
Description	Wanda is a MediaWiki extension that provides an AI-powered chatbot interface for your wiki. It can answer questions about your wiki content using various LLM providers and includes both a dedicated special page and a floating chat widget.
Author(s)	Sanjay Thiyagarajan (Techwizzietalk)
Maintainer(s)	Techwizzietalk
Latest version	2.0.0 (2025-10-24)
Compatibility policy	For every MediaWiki release that is a Long Term Support release there is a corresponding branch in the extension.
MediaWiki	1.42+
PHP	7.4+
Database changes	No
License	GNU General Public License 2.0 or later
Download	Download extension ; Git [?]: Browse repository (Phabricator · GitHub); Gerrit code review; Git commit log; Download source tarball;
	Hooks used PageSaveComplete; UploadComplete; BeforePageDisplay;
	Translate the Wanda extension if it is available at translatewiki.net
Issues	Open tasks · Report a bug

Wanda is a MediaWiki extension that provides an AI-powered chatbot interface for your wiki. It can answer questions about your wiki content using various LLM providers and includes both a dedicated special page and a floating chat widget.

Wanda also provides an API that allows other extensions and services to interact with the outside LLM as well, providing AI functionality based on the wiki content. Currently two other extensions make use of the Wanda API:

WandaScore - provides feedback on the quality of any wiki page
WandaScribe - offers suggestions on improving the writing of any selected text within a page.

Requirements

MediaWiki 1.42.0 or later
Elasticsearch server (for content indexing and search)
- CirrusSearch is the easiest approach
One of the supported LLM providers:
- Ollama (self-hosted) - recommended for privacy
- OpenAI API access
- Anthropic Claude API access
- Azure OpenAI service

Installation

Download and move the extracted Wanda folder to your extensions/ directory.
Developers and code contributors should install the extension from Git instead, using:
```
cd extensions/
git clone https://gerrit.wikimedia.org/r/mediawiki/extensions/Wanda
```
Add the following code at the bottom of your LocalSettings.php file:
```
wfLoadExtension( 'Wanda' );
```
Ensure the Requirements are met.
If you are adding the extension to an existing wiki, reindex ElasticSearch:
php extensions/Wanda/maintenance/ReindexAllPages.php
Done – Navigate to Special:Version on your wiki to verify that the extension is successfully installed.

Features

Multiple LLM Providers: Support for Ollama (self-hosted), OpenAI, Anthropic Claude, and Azure OpenAI
Text Search: Uses ElasticSearch's text-based search
Floating Chat Widget: Always-accessible chat button on all pages
Special Page: Dedicated chat interface at Special:Wanda
Responsive Design: Works on desktop and mobile devices
Secure Configuration: API key management and timeout controls
Content Indexing: Automatic indexing of wiki pages and file uploads

Configuration

Basic Configuration

Add these configuration variables to your LocalSettings.php after loading the extension:

// Choose your LLM provider
$wgWandaLLMProvider = 'ollama'; // Options: 'ollama', 'openai', 'anthropic', 'azure'

// Provider-specific settings
$wgWandaLLMApiKey = 'your-api-key-here'; // Not needed for Ollama

$wgWandaLLMModel = 'gemma:2b'; // Model name to use for the LLM provider

$wgWandaLLMApiEndpoint = 'http://localhost:11434/api/'; // API endpoint URL for the LLM provider

// Elasticsearch configuration (required for content search)
$wgWandaLLMElasticsearchUrl = 'http://elasticsearch:9200';

// UI and interface settings
$wgWandaShowPopup = true; // Show/hide the floating chat widget on all pages

// Custom prompt settings
$wgWandaCustomPrompt = ''; // Custom prompt template to override default behavior

$wgWandaCustomPromptTitle = ''; // Wiki page title containing custom prompt template

// LLM behavior settings
$wgWandaLLMMaxTokens = 1000; // Maximum tokens in LLM responses

$wgWandaLLMTemperature = '0.7'; // Creativity/randomness setting (0.0-1.0)

$wgWandaLLMTimeout = 30; // Request timeout in seconds for LLM calls

// Indexing settings
$wgWandaAutoReindex = true; // Automatically reindex content after update.php

$wgWandaSkipESQuery = false; // Skip Elastic Search

$wgWandaMaxImageSize = 5242880; // Maximum size allowed for an image

$wgWandaMaxImageCount = 10; // Maximum number of images that can be attached with the single request

Provider-Specific Examples

Ollama (Self-hosted)

$wgWandaLLMProvider = 'ollama';
$wgWandaLLMApiEndpoint = 'http://localhost:11434/api/';
$wgWandaLLMModel = 'gemma:2b';
$wgWandaLLMEmbeddingModel = 'nomic-embed-text';

OpenAI

$wgWandaLLMProvider = 'openai';
$wgWandaLLMApiKey = 'sk-your-openai-api-key';
$wgWandaLLMModel = 'gpt-3.5-turbo';
$wgWandaLLMEmbeddingModel = 'text-embedding-ada-002';

Anthropic Claude

$wgWandaLLMProvider = 'anthropic';
$wgWandaLLMApiKey = 'sk-ant-your-anthropic-key';
$wgWandaLLMModel = 'claude-3-haiku-20240307';

Azure OpenAI

$wgWandaLLMProvider = 'azure';
$wgWandaLLMApiKey = 'your-azure-api-key';
$wgWandaLLMApiEndpoint = 'https://your-resource.openai.azure.com/';
$wgWandaLLMModel = 'gpt-35-turbo';

LLM Configuration Parameters

All configuration parameters with their default values:

Variable	Default Value	Description
`$wgWandaLLMProvider`	`'ollama'`	LLM provider to use
`$wgWandaLLMApiEndpoint`	`'http://ollama:11434/api/'`	API endpoint URL
`$wgWandaLLMApiKey`		API key for external providers
`$wgWandaLLMModel`	`'gemma:2b'`	Model name to use
`$wgWandaLLMEmbeddingModel`	`'nomic-embed-text'`	Embedding model for vector search
`$wgWandaLLMElasticsearchUrl`	`'http://elasticsearch:9200'`	Elasticsearch server URL
`$wgWandaLLMMaxTokens`	`1000`	Maximum tokens in responses
`$wgWandaLLMTemperature`	`0.7`	Response creativity (0.0-1.0)
`$wgWandaLLMTimeout`	`30`	API request timeout in seconds

Usage

Special Page

Visit Special:Wanda on your wiki to access the full-featured chat interface.

Floating Chat Widget

The floating chat widget appears on all pages (except Special:Wanda) as a chat button in the bottom-right corner. Click it to open the chat interface.

Add the following configuration in LocalSettings.php to show/hide the floating chat widget.

$wgWandaShowPopup = true;

Content Indexing

To enable the chatbot to answer questions about your wiki content, you need to index your pages in Elasticsearch with embeddings. Use the maintenance script:

php maintenance/runScript.php extensions/Wanda/maintenance/ReindexAllPages.php

This script will:

Process all pages in your wiki
Generate embeddings for content chunks
Store the indexed content in Elasticsearch
Enable semantic search capabilities

API

The extension provides an API module accessible via api.php?action=wandachat.

Parameters

Parameter	Required	Description
`action`	Yes	Must be set to `wandachat`
`message`	Yes	The user's question or data to be processed
`format`	No	Response format, recommended: `json`
`usepublicknowledge`	No	Allow fallback to public knowledge when wiki context is insufficient (default: `false`)
`provider`	No	Override LLM provider (`ollama`, `openai`, `anthropic`, `azure`, `gemini`)
`model`	No	Override the model name for the request
`apikey`	No	Override API key for the request
`apiendpoint`	No	Override API endpoint URL
`maxtokens`	No	Override maximum tokens in the response
`temperature`	No	Override creativity/randomness setting (0.0–1.0)
`timeout`	No	Override request timeout in seconds
`customprompt`	No	Override the default prompt template
`customprompttitle`	No	Override using content from a wiki page as prompt template
`skipesquery`	No	Skip Elastic search query

Example Request

api.php?action=wandachat&question=How%20do%20I%20create%20a%20new%20page&format=json

Example Response

{
  "wandachat": {
    "response": "To create a new page in MediaWiki, you can...",
    "success": true
  }
}

Hooks

The extension uses several MediaWiki hooks:

PageSaveComplete

Automatically indexes pages when they are saved or updated.

UploadComplete

Indexes uploaded files for content search.

BeforePageDisplay

Adds the floating chat widget to pages.

Architecture

The extension follows this workflow:

Content Indexing: Standard text-based Elasticsearch
Query Processing: Standard text-based Elasticsearch
Similarity Search: Standard text-based Elasticsearch
Response Generation: The LLM generates responses based on the retrieved content

Future workflow for vector embeddings (once MW supports ElasticSearch 8 or another vector database is used):

Content Indexing: Wiki pages are processed and stored in Elasticsearch with embeddings
Query Processing: User queries are converted to embeddings (when available)
Similarity Search: Elasticsearch finds the most relevant content using vector search
Response Generation: The LLM generates responses based on the retrieved content
Fallback Mechanism: Falls back to text-based search when embeddings are unavailable

Security Considerations

API keys are stored securely in MediaWiki configuration files
Content is processed by your chosen LLM provider - consider data privacy implications
For sensitive wikis, consider using self-hosted Ollama instead of external services
Configure appropriate timeouts and rate limits to prevent abuse
The floating chat widget and special page respect MediaWiki user permissions

Troubleshooting

Common Issues

Chat not responding

Verify your LLM provider configuration
Check that the API endpoint is accessible
Ensure API keys are correctly configured for external providers

No relevant results

Run the content indexing script: php maintenance/runScript.php extensions/Wanda/maintenance/ReindexAllPages.php
Verify Elasticsearch is running and accessible
Check that pages contain sufficient content for indexing

Floating chat not appearing

Ensure the extension is properly loaded in LocalSettings.php
Check that JavaScript is enabled in the browser
Verify the page is not Special:Wanda (widget is disabled there)

Version History

1.0.0: Initial release with multi-provider LLM support, Elasticsearch integration, and floating chat widget.

2.0.0: Implement public knowledge addition, OOUI to Codex migration, rename API action to 'wandachat', allow LLM parameters override using API parameters.