Developer Information

This chapter provides technical details for developers and integrators.

Architecture

The extension uses three PSR-15 middlewares:

UrlSuffixMiddleware - Detects .md suffix and rewrites URLs
LlmsTxtMiddleware - Serves the /llms.txt endpoint
ContentFormatMiddleware - Transforms HTML to Markdown

Middleware Chain

Request: /about.md
   │
   ▼
UrlSuffixMiddleware (before site resolver)
   │  Strips .md suffix
   │  Sets request attribute 'llms_txt_format' = 'md'
   │  Rewrites URI to /about
   ▼
TYPO3 Site Resolver
   │
   ▼
LlmsTxtMiddleware (after site, before page resolver)
   │  Handles /llms.txt requests
   ▼
TYPO3 Page Resolver & Frontend
   │
   ▼
ContentFormatMiddleware (after content-length-headers)
   │  Checks for 'llms_txt_format' attribute
   │  Converts HTML response to Markdown
   ▼
Response: Markdown with YAML frontmatter

Services

PageTreeService

RTfirst\LlmsTxt\Service\PageTreeService

Traverses the TYPO3 page tree and collects page data for llms.txt generation.

Supports multi-language sites
Respects page exclusion settings
Handles translated pages with fallback

MarkdownConverterService

RTfirst\LlmsTxt\Service\MarkdownConverterService

Orchestrates content element to Markdown conversion using registered converters.

LlmsTxtGeneratorService

RTfirst\LlmsTxt\Service\LlmsTxtGeneratorService

Generates the llms.txt content for a site.

Content Converters

The extension uses a converter pattern for content element to Markdown conversion. Each converter implements ContentConverterInterface:

interface ContentConverterInterface
{
    public function supports(string $cType): bool;
    public function convert(array $record, string $baseUrl): string;
}

Built-in Converters

Converter	Supported CTypes	Description
HeaderConverter	`header`	Converts header elements
TextConverter	`text`, `textpic`, `textmedia`	Converts text and text+media
ImageConverter	`image`	Converts image galleries
BulletsConverter	`bullets`	Converts bullet lists
TableConverter	`table`	Converts tables
MenuConverter	`menu_*`	Converts menu elements
HtmlConverter	`html`	Converts raw HTML
DefaultConverter	(fallback)	HTML-to-Markdown fallback

Creating Custom Converters

Create a class implementing ContentConverterInterface:

<?php
declare(strict_types=1);

namespace Vendor\MyExtension\Converter;

use RTfirst\LlmsTxt\Converter\ContentConverterInterface;

class MyCustomConverter implements ContentConverterInterface
{
    public function supports(string $cType): bool
    {
        return $cType === 'my_custom_element';
    }

    public function convert(array $record, string $baseUrl): string
    {
        $header = $record['header'] ?? '';
        $bodytext = $record['bodytext'] ?? '';

        return "## {$header}\n\n{$bodytext}";
    }
}

Register the converter in Services.yaml:

services:
  Vendor\MyExtension\Converter\MyCustomConverter:
    tags:
      - name: 'llms_txt.content_converter'
        priority: 100

Higher priority converters are checked first.

Event Listeners

CacheFlushEventListener

Invalidates llms.txt cache when TYPO3 caches are flushed.

HeaderLinkEventListener

Injects the <link rel="alternate"> tag into HTML responses.

BackendNotificationEventListener

Shows a notification in the Backend if robots.txt lacks llms.txt reference.

Caching

The extension uses two cache layers:

llms.txt Index Cache (cache_pages)
- Stores generated llms.txt content per site
- Invalidated on cache flush
Format Output Cache (llms_txt_format)
- Stores Markdown output per page/language
- 24-hour default lifetime
- Part of the pages cache group

Database Schema

The extension adds fields to the pages table:

Field	Type	Description
tx_llmstxt_description	text	LLM-specific page description
tx_llmstxt_summary	text	Extended page summary
tx_llmstxt_keywords	varchar(255)	Comma-separated keywords
tx_llmstxt_exclude	tinyint(1)	Exclude from llms.txt
tx_llmstxt_priority	int(11)	Priority (0-100) for sorting

Code Quality

The extension maintains high code quality standards:

PHPStan Level 8 compliant
PSR-12 code style (php-cs-fixer)
Unit tests for converters and services

Run quality checks:

# Static analysis
vendor/bin/phpstan analyse packages/llms_txt --level=8

# Code style check
vendor/bin/php-cs-fixer fix packages/llms_txt --dry-run

# Fix code style
vendor/bin/php-cs-fixer fix packages/llms_txt

# Run tests
vendor/bin/phpunit -c packages/llms_txt/phpunit.xml

Dependencies

league/html-to-markdown (^5.1) - HTML to Markdown conversion
typo3/cms-core (^13.0 || ^14.0)
typo3/cms-frontend (^13.0 || ^14.0)