Anthropic silently downgraded cache TTL from 1h → 5M on March 6th

Develop a middleware proxy that manages local semantic caching for LLM providers. By intercepting API calls, the tool can cache responses locally and provide a fallback if TTL is reduced, saving costs and latency.

Suggested repo: CacheShield

"Take control of your LLM costs when providers change caching rules."

Estimated effort: 40h