QuestionQ36
VerifiedEvaluating and deploying agentic workflowsYou are building a social media monitoring agent pipeline in Google Cloud that analyzes five million incoming customer feedback posts per day. The pipeline must classify each post into one of three sentiment categories: positive, negative, or neutral. It must also generate a one-sentence summary for each post. You need to configure the language model to minimize inference cost and latency while maintaining high classification accuracy.
What should you do?
Community Discussion