Hey everyone, how are you?

I'm new to using LlamaIndex!
I have the following problem and would appreciate it if you could help me think it through:

I have a VectorStoreIndex and when I make a Query it returns a Node, the information I need to synthesize the answer is split between that Node it returns and another Node (which it doesn't return) but that in the Document is following that Chunk.

How can I make the Synthesizer take into consideration the Nodes that surround the Node that the Retriever brought me so that no information is missing when generating the answer?

Hi!
Does anyone get good results using "gpt-35-turbo-16k" with Agents / Query Engine Tools?
I'm not getting good quality when LLM generates the query to call the query engine.
Is there any tips to Prompt the ToolMetadata Description?
Using "gpt-4" works excellent (but slower and expensive)

query_engine_tools = [
QueryEngineTool(
query_engine=doc_summary_index.as_query_engine(
vector_store_query_mode="hybrid",
service_context=service_context,
use_async=True,
verbose=True,
),
metadata=ToolMetadata(
name="doc_summary_index",
description=(
"Answers questions about Health Care program."
"Extract a well-formed question with a lot of detail."
),
),
)
]

agent = OpenAIAgent.from_tools(
llm=llm,
tools=query_engine_tools,
verbose=True
)