Setting the file. One moment.
Subchapter 23.3
references/decision-refs/cost-levers.mdMarkdown2 KBView on GitHub
This document is the only legal source for cost-optimization levers referenced in agent-advisor outputs. When generating estimates or reports, cite only the levers documented in the table below. Do not invent or suggest discounts, optimizations, or cost-reduction strategies that are not explicitly listed here.
| Lever | Applies to | Effect (order of magnitude) | Caveat |
|---|---|---|---|
| Model tier routing (Sonnet→Haiku for triage/simple items) | any LLM unit | ≈5× cheaper on routed traffic | quality-gate the routing |
| Prompt caching | repeated system/preamble tokens | up to ~90% off cached input tokens | cache-friendly prompt structure required |
| Batch inference | latency-tolerant workloads (scheduled/batch) | ~50% off tokens | async only |
| Scale-to-zero runtimes | spiky/idle-heavy units | eliminates idle compute | cold-start tolerance |
| Quota scheduling | shared Bedrock TPM across units | avoids provisioned throughput | schedule heavy runs off-peak |
When referencing a cost lever in estimate.json.drivers[] or in any report or plan section, cite the lever name exactly as it appears in the table above.