[1] node_parser = MarkdownElementNodeParser(llm=llm)
[2] nodes = node_parser.get_nodes_from_documents([document])
[3] base_nodes, objects = node_parser.get_nodes_and_objects(nodes)

node_parser = MarkdownElementNodeParser(num_workers=8, show_progress=False)
nodes = node_parser.get_nodes_from_documents([document])
base_nodes, objects = node_parser.get_nodes_and_objects(nodes)
index =  VectorStoreIndex(nodes=base_nodes+objects)
recursive_query_engine = index.as_query_engine(similarity_top_k=3,node_postprocessor[FlagEmbeddingReranker(top_n=2, model=RERANKER_MODEL)], verbose=False)

recursive_query_engine = recursive_index.as_query_engine(
    similarity_top_k=5,
    node_postprocessors=[reranker],
    verbose=True
)

class DocumentTypeResponse(BaseModel):
    """Data model for the document type"""
    document_type: str

document_type_identifier = QueryEngineTool(
    query_engine=recursive_query_engine,
    metadata=ToolMetadata(
        name='document_type',
        description=(
            "Only use this tool when required. "
            "Answer to question relating to document type. "
            "Identify document types as either purchase order or invoice. "
        ),
        fn_schema=DocumentTypeResponse
    )
)

context_document_agent = """\
You are an expert administraive assistant who specialized in answering questions about document. 
The questions mainly revolves around extracting key information from either an invoice or purchase order. 
Only use the necessary tool to answer the questions. Only use more tools when needed.
"""

document_agent = ReActAgent.from_tools(
    tools=document_type_identifier,
    verbose=True,
    context=context_document_agent,
)

Using `bitsandbytes` 8-bit quantization requires Accelerate: `pip install accelerate` and the latest version of bitsandbytes: `pip install -i https://pypi.org/simple/ bitsandbytes`

TypeError: BFloat16 is not supported on MPS and ImportError: 

# set up llm using HuggingFaceLLM
import torch
from llama_index.llms.huggingface import HuggingFaceLLM
from transformers import BitsAndBytesConfig

quantization_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_compute_dtype=torch.float16,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_use_double_quant=True,
)

llm = HuggingFaceLLM(
    model_name=model_name,
    model_kwargs={
        "token": hf_token,
        "torch_dtype": torch.bfloat16,
        "quantization_config": quantization_config
    },
    generate_kwargs={
        "do_sample": True,
        "temperature": 0,
        "top_p": 0.9,
    },
    tokenizer_name=model_name,
    tokenizer_kwargs={"token": hf_token},
    stopping_ids=stopping_ids,
)

# A
- text for A

# B
- text for b

from llama_index.core.node_parser import MarkdownNodeParser
parser = MarkdownNodeParser()

for page in document:       
    page_nodes = parser.get_nodes_from_documents([page])
    for node in page_nodes:
        if node.metadata['Header_1'] == 'B':
            node.text = 'New text for B'
        print(node.text) # text is changed
    print(page.text) # text is not changed!

---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
Cell In[20], line 1
----> 1 response_header = header_query_engine.query(new_header_query)

File <hidden_path>\.venv\lib\site-packages\llama_index\core\instrumentation\dispatcher.py:260, in Dispatcher.span.<locals>.wrapper(func, instance, args, kwargs)
    252 self.span_enter(
    253     id_=id_,
    254     bound_args=bound_args,
   (...)
    257     tags=tags,
    258 )
    259 try:
--> 260     result = func(*args, **kwargs)
    261 except BaseException as e:
    262     self.event(SpanDropEvent(span_id=id_, err_str=str(e)))

File <hidden_path>\.venv\lib\site-packages\llama_index\core\base\base_query_engine.py:52, in BaseQueryEngine.query(self, str_or_query_bundle)
     50     if isinstance(str_or_query_bundle, str):
     51         str_or_query_bundle = QueryBundle(str_or_query_bundle)
---> 52     query_result = self._query(str_or_query_bundle)
     53 dispatcher.event(
     54     QueryEndEvent(query=str_or_query_bundle, response=query_result)
     55 )
     56 return query_result
...
--> 302 content = content_template.format(**relevant_kwargs)
    304 message: ChatMessage = message_template.copy()
    305 message.content = content

KeyError: "' Item No"

# create a set of tools on top of the query engine
query_engine_tools = [
    QueryEngineTool(
        query_engine=recursive_query_engine,
        metadata=ToolMetadata(
            name = 'document_type_finder',
            description = (
                '''
                This tool finds the type of document and outputs the answer. 
                This tool will look for obvious features in the document such as words like "invoice" or "tax invoice" for invoice and words like "purchase order", "sales order" for purchase order.
                Finally, this tool only outputs answer as "invoice" or "purchase order"
                '''
            )
        )
    ),
    QueryEngineTool(
        query_engine=recursive_query_engine,
        metadata=ToolMetadata(
            name = 'document_origin_finder',
            description = (
                '''
                This tool finds the country of origin of the document
                This finder will first identify the country of origin of the document sender by looking into its company name, address, or any other relevant information to derive the origin of country.
                Finally, this tool format the answer following to ISO 3166-1 alpha-3 and outputs it.
                '''
            )
        )
    )
]

# create a RAG ReAct QueryEngineTool agent
agent = ReActAgent.from_tools(tools=query_engine_tools, llm=llm, verbose=True)

response = agent.chat(
    '''
    What is the document type and origin? Use any of the tools provided to you.
    '''
)

# output
>>>Thought: The user has not provided the document text for analysis. I need to ask for the document text to proceed with the analysis using the appropriate tool.
Answer: Could you please provide the text of the document you would like me to analyze for its type and origin?