Reading time: 10 minutes

Apache Solr for TYPO3 – The path to intelligent search

Search has long been the unloved stepchild of many web projects: a small input field somewhere in the corner that was supposed to ‘work somehow’ – as long as no one complained about it. Users type in their questions, get a list of hits and have to laboriously gather the relevant information themselves. That is currently changing fundamentally. With TYPO3 14 in April 2026 and a new generation of EXT:solr, a new chapter is beginning: search is becoming an intelligent knowledge layer.

The TYPO3 14 moment

Every major version is a natural turning point. Adjustments are necessary anyway, budgets are approved, and teams deal with the current situation. TYPO3 14 in April 2026 offers the opportunity not only to establish compatibility, but to take a real leap forward.

The new EXT:solr version for TYPO3 14 is more than just a technical update. It marks the transition from classic full-text search to a system that understands meaning.

Structure, semantics, intelligence: the three levels

The path to AI-enabled search follows a logical sequence. To understand the significance of the different levels in AI-enabled search, we need a clear overview of the process and the underlying technologies. The development of an intelligent search solution is a step-by-step process that begins with structured data, which serves as a solid foundation.

Structured data as a foundation

Before artificial intelligence can play to its strengths, it needs structured source data. JSON-LD makes content machine-readable and defines entities and their relationships. A product page becomes a product with price, availability and manufacturer. An article becomes a creative work with author, date and topic.

The immediate benefit: Search engines such as Google understand the content better, rich snippets become possible, and visibility increases. The strategic benefit: Your own AI integration benefits from the same structure.

Enhancely makes this structuring accessible. Editors do not need to study Schema.org specifications. The extension takes care of the markup based on the existing TYPO3 content types, making the first step practical.

Synonyms on steroids

Classic synonym management in Solr works, but it is tedious. Someone has to decide that ‘home office’ and ‘remote work’ belong together. Someone has to maintain, expand and update this list. In multilingual projects, the effort multiplies.

Large language models are fundamentally changing this game. They understand meaning in context and recognise semantic relationships without explicit configuration. A user searches for ‘working on the go,’ and the system understands the intention and finds relevant content on flexible working models, even if this phrase does not appear anywhere.

This is not a pipe dream. The integration of embedding models in Solr already makes this type of semantic search possible today. The new EXT:solr version brings these capabilities to the TYPO3 world.

From hit lists to answers

The next stage is called Retrieval Augmented Generation, or RAG for short. Instead of a list of links, the system provides answers based on its own content.

An example:A visitor to the company website asks, ‘What certifications does product X have?’ Today, the search might return the product page, the data sheet and three press releases. The user has to piece the information together themselves. With RAG, the system can respond: ‘Product X is certified according to ISO 9001 and ISO 14001. The current certification was renewed in March 2025.’ With source references, of course.

Solr RAG is available as an add-on for enterprise partners. This turns your own website into a knowledge base that can answer questions.

Infrastructure for the new world

AI functions place different demands on infrastructure than traditional search. Embedding models require computing power, vector databases require storage, and RAG pipelines require orchestrated collaboration between multiple components.

Not every project wants or is able to operate this infrastructure itself. The new Hosted Solr Cloud offers an alternative: managed infrastructure that scales with your requirements. New AI tariffs make getting started predictable without the need for your own GPU clusters.

Those who prefer to host themselves can continue to do so. EXT:solr remains open source, and the AI extensions are documented. The decision between self-operation and managed service is a question of your own strategy, not technical possibilities.

A new ecosystem model

A new sponsorship model for EB partners will also be launched in April 2026. The idea behind it is that sustainable open source development requires sustainable financing. Partners who use EXT:solr in customer projects invest in further development and benefit jointly from the results.

This is not a new concept. In anthropology, we are familiar with the kula ring, a barter system in the Pacific region in which value flows in a circle and strengthens all participants. Open source works in a similar way when the ecosystem is right. The new model aims to enable precisely that: a stable cycle of contribution and benefit.

And where are you currently?

Not every project has to climb all the steps at once. A sensible assessment of your current position could look like this:

Basic level: 

• EXT:solr classic.
• Fast, relevant full-text search.
• For many projects, this is already a significant improvement on the TYPO3 on-board search.
   

Structure level:

• JSON-LD via Enhancely.
• Immediate SEO benefits, preparation for AI integration.
• Manageable effort, measurable benefits.
   

Semantics level:

• LLM-supported synonyms and vector search.
• Context understanding without manual maintenance.
• Relevant for projects with large amounts of content or multilingual requirements.
   

Intelligence level:

• Solr RAG.
• Answers instead of hit lists.
• For companies that want to position their website as a knowledge resource.

The transition to TYPO3 14 in April 2026 is a good time to ask these questions. Where does your own project stand? What would be the next sensible step? The technology is ready. So is the infrastructure.

Request advice on intelligent search