madj2k/t3-ai-assistant

AI Assistant is a flexible TYPO3 extension for building AI-powered assistants based on Retrieval-Augmented Generation (RAG), configurable processing pipelines, and pluggable AI and vector store integrations.

Maintainers

Package info

github.com/skroggel/t3-ai-assistant

Type:typo3-cms-extension

pkg:composer/madj2k/t3-ai-assistant

Transparency log

Statistics

Installs: 11

Dependents: 0

Suggesters: 0

Stars: 0

Open Issues: 0

Security

Advisories: 0

Aikido package health analysis

1.0.2 2026-08-19 15:12 UTC

Requires

Requires (Dev)

GPL-2.0-or-later adddd0d7f12ddc4e02ab71c6fdff96a42b71e5a4

  • Steffen Kroggel
  • Maximilian Fäßler

Published on Packagist.org on 2026-08-19 15:13 UTC

This package is auto-updated.

Last update: 2026-08-19 15:14:28 UTC

README

AI Assistant provides configurable Retrieval-Augmented Generation (RAG) assistants for TYPO3. Editors can combine TYPO3 pages, files and custom data sources with configurable chat pipelines and source attribution. Additional source types can be supplied by extensions.

The extension provides the TYPO3 integration: database records, backend configuration, source discovery, persistence, commands, session handling, diagnostics and frontend output. The framework-independent runtime, pipeline contracts, provider connectors and indexing primitives live in madj2k/ai-core, which Composer installs as a dependency.

Features

  • Assistant profiles and configurable processing pipelines
  • AI and vector-store connection records
  • TYPO3 page and file indexing, including structured file adapters
  • Metadata-aware retrieval and source links
  • Chat memory, pipeline traces and backend diagnostics
  • Extension points for processors, connectors, indexers and file adapters

Requirements

  • TYPO3 13.4 or 14.3
  • PHP 8.2 or newer
  • An AI provider for chat and embeddings
  • A reachable vector store such as Qdrant

Installation

composer require madj2k/ai-assistant

After installation:

  1. Apply the TYPO3 database schema updates.
  2. Create and test an AI connection and a vector-store connection in the backend module.
  3. Create an indexer and index its sources.
  4. Create or import an assistant profile using the same vector-store connection and collection.
  5. Add the chat plugin to a page.

See the Documentation for configuration, commands and extension points.

Local Qdrant with DDEV

ddev get netz98/ddev-qdrant
ddev restart

The dashboard is normally available at:

https://<project-name>.ddev.site:6333/dashboard

To test API-key authentication locally, add QDRANT__SERVICE__API_KEY to the Qdrant service environment and enter the same key in the TYPO3 vector-store connection record.

Diagnostics

Set Pipeline log mode to verbose under AI Assistant > Configuration, then inspect a chat request under AI Assistant > Diagnostics. Optional TYPO3 PSR log output is written to var/log/tx_aiassistant.log; general TYPO3 errors are available in var/log/typo3_*.log.

Render the documentation

Run from the extension root:

docker run --rm --pull always -v "$(pwd)":/project -it ghcr.io/typo3-documentation/render-guides:latest --config=Documentation