From TYPO3 content to verified answers.
Follow content from your TYPO3 pages and FAL file storage through extraction, vector embeddings, and context-augmented retrieval to produce accurate responses.
Connect TYPO3 Content & FAL Assets
T3chatbot taps directly into your TYPO3 database. It scans published frontend page trees, content elements (tt_content), and enterprise media stored in TYPO3’s File Abstraction Layer (FAL).
Index, Chunk & Vectorize
Text is parsed, divided into semantic chunks, and embedded via your configured provider (OpenAI, Gemini, Claude, or local Ollama). Automated TYPO3 Scheduler tasks and CLI commands keep embeddings up to date.
Semantic Vector Retrieval (RAG)
When a visitor asks a question, the query is converted to a vector embedding to locate the most relevant passages in your index, strictly filtered by the visitor’s current TYPO3 Site Identifier.
Grounded Answer with Citations
The AI model synthesizes a factual, conversational answer exclusively from retrieved passages. Responses include clickable links and document references, eliminating hallucinations.
Automated Index Maintenance
Keeping your knowledge base synchronized with daily editorial activity is handled seamlessly via TYPO3 CLI commands and native Scheduler tasks.
./vendor/bin/typo3 aichatbot:crawl
queues pages and FAL attachments.
./vendor/bin/typo3 aichatbot:indexContentClasses --limit=50
generates vectors.
Enable Frontend Indexing
auto-updates missing content as visitors browse.
./vendor/bin/typo3 aichatbot:indexContentClasses --limit=50
