Hey I m slightly confused about

def create_document(text, metadata):
    document = Document(
        text=" ".join(text),
        metadata=metadata,
        metadata_separator="::",
        metadata_template="{key}=>{value}",
        text_template="Metadata: {metadata_str}\n-----\nContent: {content}",
    )
    return document

metadata_extractor = MetadataExtractor(
    extractors=[
        TitleExtractor(nodes=3, node_template=DEFAULT_TITLE_NODE_TEMPLATE, combine_template=DEFAULT_TITLE_COMBINE_TEMPLATE),
        KeywordExtractor(keywords=2),
        SummaryExtractor(summaries=["prev", "self"], prompt_template=DEFAULT_SUMMARY_EXTRACT_TEMPLATE),
     ],
)

node_parser = SimpleNodeParser(
    metadata_extractor=metadata_extractor
)

tax_nodes= node_parser.get_nodes_from_documents(documents)

from llama_index.schema import MetadataMode
test_node = tax_nodes[2]
print("The LLM sees this: \n", test_node.get_content(metadata_mode=MetadataMode.LLM))
# print(document)

The LLM sees this: 
 Metadata: Hoofdstuk: Hoofdstuk II. Heffing ter zake van leveringen en diensten
Artikel: Artikel 2
Afdeling: Afdeling 1. Belastbaar feit
Paragraaf: 
document_title: Harmonisatie van de Europese Unie Wetgevingen inzake Belasting op Leveringen van Goederen en Diensten, Intracommunautaire Verwervingen en Invoer van Goederen: Wet op de Omzetbelasting 1968
excerpt_keywords:  BTW-richtlijn 2006, accijnsgoederen
prev_section_summary: 
<summary>
section_summary: 
<summary>
-----
Content: Metadata: 
-----
Content: <content>

(Pdb) print(nodes[0].get_content(metadata_mode=MetadataMode.LLM))
Metadata: test=>val
entities=>{'LangChain', 'SQL', 'Flask', 'Docker'}
excerpt_keywords=>LLMs, LlamaIndex
section_summary=>LlamaIndex is a data framework that provides tools to help augment LLMs with private data. It offers data connectors to ingest data sources and formats, provides ways to structure data, an advanced retrieval/query interface, and easy integrations with outer application frameworks. It is suitable for both beginner and advanced users, with a high-level API for the former and lower-level APIs for the latter.
-----
Content: 
LLMs are a phenomenonal piece of technology for knowledge generation and reasoning. 
They are pre-trained on large amounts of publicly available data.
How do we best augment LLMs with our own private data?