# HELP ace_tokens_saved_total Upstream tokens avoided via cache/prefix reuse.
# TYPE ace_tokens_saved_total counter
ace_tokens_saved_total 0
# HELP ace_tokens_saved_by_source Upstream tokens avoided, by savings source.
# TYPE ace_tokens_saved_by_source counter
ace_tokens_saved_by_source{source="cache:exact"} 0
ace_tokens_saved_by_source{source="cache:kv"} 0
ace_tokens_saved_by_source{source="cache:semantic"} 0
ace_tokens_saved_by_source{source="compaction"} 0
ace_tokens_saved_by_source{source="prefix-reuse"} 0
# HELP ace_dollars_saved_total Estimated USD reclaimed vs the served-baseline counterfactual.
# TYPE ace_dollars_saved_total counter
ace_dollars_saved_total 0.000000
# HELP ace_cache_hit_ratio Cache hits / total requests.
# TYPE ace_cache_hit_ratio gauge
ace_cache_hit_ratio 0.000000
# HELP ace_requests_total Total requests observed by the proxy.
# TYPE ace_requests_total counter
ace_requests_total 0
# HELP ace_cost_usd_total Actually-spent upstream USD (served requests).
# TYPE ace_cost_usd_total counter
ace_cost_usd_total 0.000000
# HELP ace_gross_usd_total Counterfactual list cost with no ACE (net + reclaimed).
# TYPE ace_gross_usd_total counter
ace_gross_usd_total 0.000000
# HELP ace_goodput_ratio Delivered completion tokens meeting SLA / all delivered.
# TYPE ace_goodput_ratio gauge
ace_goodput_ratio 0.000000
# HELP ace_ttft_ms_p50 Served time-to-first-token, P50 (ms).
# TYPE ace_ttft_ms_p50 gauge
ace_ttft_ms_p50 0.0
# HELP ace_ttft_ms_p90 Served time-to-first-token, P90 (ms).
# TYPE ace_ttft_ms_p90 gauge
ace_ttft_ms_p90 0.0
# HELP ace_ttft_ms_p99 Served time-to-first-token, P99 (ms).
# TYPE ace_ttft_ms_p99 gauge
ace_ttft_ms_p99 0.0
# HELP ace_itl_ms Mean inter-token latency (ms/token).
# TYPE ace_itl_ms gauge
ace_itl_ms 0.00
# HELP ace_fallback_total Requests re-routed after upstream degraded.
# TYPE ace_fallback_total counter
ace_fallback_total 0