Skip to main navigation Skip to main content Skip to page footer

SEAL Ecosystem Moves From Search to Answers: Community Budget Report

Version 3.0 of the Index extension and 2.0 releases of SEAL and SEAL AI bring configuration-driven indexing, external content fetching, WCAG 2.2 templates, chunked embeddings, and a read-only MCP server that makes TYPO3 content available to AI agents.

In May, I reported that the Search Engine Abstraction Layer (seal) ecosystem had reached production readiness: faceted and geographical search, search statistics, and AI-generated summaries on top of vector search. Since then, the extensions have found their way into more and more real projects. And real projects bring new requirements.

With the follow-up budget idea in the third Community Budget round of 2026, I set out to address these requirements. The goal: moving from search to answers. Here is what happened over the past weeks in the SEAL family of extensions.

Index Extension 3.0: Consolidation and Migration Paths

The Index extension (index) is the indexing framework beneath the whole SEAL ecosystem. It decouples indexing from the search technology. Version 3.0 shifts its focus from adding support extension by extension to a scalable, configuration-driven approach.

Configuration Instead of Code: the Record Extender

The new Record Extender is a generic, configuration-driven indexer. Any TCA table can now become indexable via settings.yaml: field mapping, tag derivation, and geo fields are defined in configuration, not in PHP. This replaces the previous "one integration per extension" approach with one that scales to every project.

The first showcase ships right away. Support for the Event Management and registration extension (sf_event_mgt) is delivered as a pure Extender configuration.

New Fetch Technology for External Content

Not every relevant piece of content lives in TYPO3. The new Fetch technology selects external pages recursively, similar to a crawl with curl (recursive), and feeds them into the same index process. Documentation portals, microsites, or legacy systems can now be part of one unified search. It is very flexible in complex search scenarios.

ke_search Bridge: Now Its Own Extension

Many TYPO3 projects run the Faceted Search extension (ke_search), and a full migration is rarely an option on day one. The ke_search bridge offers a migration path: existing ke_search plugins keep working, while indexing moves to the more flexible Index extension. The bridge now lives in its own extension, Bridge between EXT:index and EXT:ke_search (index_ke_search), so projects only install it when they need it. This extension needs first real-world integrations! If you run ke_search, I’m looking for the first projects to try the bridge.

SEAL Extension 2.0: Result Quality and Accessibility

Where the last release was about new search capabilities, this one is about the quality of what users actually see.

Keyword Highlighting

Search terms are now highlighted directly in the result list. The feature builds on the SEAL project and is also available in the TYPO3 adapter, so users immediately see why a result matches their query.

WCAG 2.2 Compatible Templates

The search plugin templates have been reworked for WCAG 2.2 conformance (ARIA, focus management, keyboard-operable filters, fixes in the pagination). With the BFSG (the German Accessibility Strengthening Act) and the European Accessibility Act in force, accessible search is no longer optional. These templates give projects a compliant starting point instead of a retrofit task.

Dashboard Widgets and Tests

The search statistics dashboard widgets received several fixes, and the test suite has been extended further.

SEAL AI Extension 2.0: From Vectors to Answers

The SEAL AI extension received the biggest update of this round.

Symfony AI Alignment and AiM Support

The extension is aligned with the current state of Symfony AI, including the store lifecycle (StoreInterface::clear()), StoreFactory and DSN configuration, the current platform bridges and the FailoverPlatform for production resilience.

As an alternative, the AI manager AiM extension (aim) can now act as model and provider for all AI functions. This gives administrators more transparency about AI usage directly in the TYPO3 backend.

Smarter Chunking

Instead of embedding whole pages, the extension now splits content into chunks. Semantic chunking with adjustable overlap, chunk-level embeddings and parent-document retrieval work together: matches happen at chunk level, results are presented at page level. Chunk size and strategy are configurable.

Content-Hash Embedding Cache

Each chunk gets a hash over its normalized content plus model, dimension, and chunking configuration. Only changed chunks are re-embedded, so incremental re-indexing replaces full rebuilds. Token and cost reporting makes the savings visible.

Read-Only MCP Server

A new read-only MCP (Model Context Protocol) server exposes search and retrieve tools to AI agents, secured with token-based authentication. Your TYPO3 content becomes a source for retrieval-augmented generation (RAG) in agent workflows — without giving agents write access.

Rate Limiting on Critical Paths

AI calls cost money and resources. A rate limiter now protects the critical paths, such as AI summaries and the MCP endpoint, against abuse and unexpected cost spikes.

Thank You

I am very glad that this work is supported by the TYPO3 Community Budget. My sincere thanks go to the TYPO3 Association and to everyone who voted for this idea. Without this support, the ecosystem would not evolve at this pace and with this level of quality.

Thanks as well to everyone who uses the extensions and shares their experience. Many of the changes in this round started as feedback from real-world projects.

Alignment With Strategic Goals

This round supports two goals of the TYPO3 Association strategy. The AiM integration contributes to Improving Governance (5.2): AI usage becomes visible and manageable in the backend instead of hiding in configuration files. The vectorization pipeline and the MCP interface contribute to Strengthening the Innovation Process and Product Ownership (5.4), bringing TYPO3 into AI-agent and RAG scenarios.

Try It Yourself

All extensions are available in the TYPO3 Extension Repository and on Packagist:

I always welcome feedback, whether it is a bug report, a feature idea, or simply a note that the search is running in your project. The best place to reach me is via the GitHub repositories.