A parks-permit office first needs the document type, then the applicant names, then an answer to a clerk question from that same text. How should the app compose those official NLP tasks?
Select an answer to reveal the explanation.
Short Explanation
Document type, then names, then a clerk question: call classification, NER, and extractive QA as ordinary functions. A fixed sequence does not need a branching agent. One chat completion or a RAPIDS ingest job is not those three heads.
Full Explanation
Classification, named-entity recognition, and extractive QA are official NLP task heads that compose in ordinary application code. A fixed parks-permit sequence is three function calls, not a large branching agent. LangGraph is the right later tool when a path actually forks. A single unconstrained generation and a RAPIDS wrangle do not implement those three heads.