Task

• Currently the ingest openai uses two approaches
  • Using the file_search tool with vector store ids when creating a response with llama stack client https://github.com/thepetk/langgraph-agents/blob/ebde9b71f9801642ee98ce753633ebfa78fedeae/examples/langchain-langgraph/ingest_openai.py#L422-L454 
  • Listing and quering the vector store directly and passing the context, prompt without the file search tool to the llama stack client https://github.com/thepetk/langgraph-agents/blob/ebde9b71f9801642ee98ce753633ebfa78fedeae/examples/langchain-langgraph/ingest_openai.py#L457-L523 
• Investigate if we can find an option c, which can perform better than the above two ideas

Background

Dependencies and Blockers

QE impacted work
Documentation impacted work 

Acceptance Criteria