TGA uses a lightweight implementation that directly concatenates filtered plant data into the context, omitting infrastructure such as vector databases, in an approach similar in principle to RAG.