AI memory (RAG)

// embed text, store the vectors, search by meaning
%Qdrant->create('docs')
%Qdrant->upsert('docs', $id, $text, title: $title)

$hits = %Qdrant->search('docs', 'how do refunds work?', top: 5)

// then answer with one call over any model: Claude, OpenAI, Gemini, Grok
$reply = %AI->chat(user: 'Answer using: '.$hits[0]->title)