Cache-first AI runtime

Edge Gateway

Workspace-isolated AI gateway with cache evidence, provider-call accounting, and operator-visible benchmark gates.

Sprint 4 Evidence loading
Cache HIT P95
4ms
Measured live runtime
False HIT
0
Compare artifact
Token reduction
0 bps
Threshold not met
Provider calls
0 bps
Threshold not met

No. The live compare decision is FAIL-measured, so no token-saving or provider-call win is claimed.

Cloudflare Pages hosts this UI shell. The Worker remains API/runtime/status.