PDF, DOCX, Markdown, or plain text. We parse, split by section, run each chunk through Gemini to extract typed entries, then embed them. All agents see the new knowledge within seconds.