COST OPTIMIZATION TECHNIQUES

ARTIFICIAL INTELLIGENCE ASSISTANT PROMPTS Prompt 1: Model routing scheme. Use this when you are deciding which model should handle each part of a pipeline. [START OF PROMPT] Recommend a model-routing scheme for these subtasks in our pipeline, assigning each to a cheap/fast tier or a capable/expensive tier with justification: subtasks = [LIST SUBTASKS, e.g. "extract fields from resume", "rank candidate against posting", "write final narrative report"]. Output format: a routing table (Subtask, Recommended Tier, Justification, Estimated Cost Impact). [END OF PROMPT] Prompt 2: Caching opportunities. Use this to find which parts of your calls can be cached. [START OF PROMPT] Identify which parts of this session's calls would benefit from prompt caching, and estimate the savings: [PASTE LAB OR PRODUCTION SESSION'S SYSTEM PROMPT, TOOL DEFINITIONS, AND CALL COUNT HERE]. Output format: a caching-opportunity table (Context Block, Cacheable? Y/N, Reason) plus an estimated cost/latency savings calculation for the session. [END OF PROMPT]