=== RUN   TestLiveDualBackend
    live_test.go:45: native_success=1.00 remote_success=0.33 native_p95=16.615955879s remote_p95=6.926484503s p0=0
    live_test.go:48: pi-native doc-brief success=true p0=false targets=1/1 dur=8.216466681s err= tokens=1765
    live_test.go:48: pi-native doc-outline success=true p0=false targets=1/1 dur=16.615955879s err= tokens=2946
    live_test.go:48: pi-native dialogue-hello success=true p0=false targets=1/1 dur=3.001327463s err= tokens=159
    live_test.go:48: pi-native dialogue-status success=true p0=false targets=1/1 dur=2.657008525s err= tokens=180
    live_test.go:48: pi-native reason-risk success=true p0=false targets=1/1 dur=2.403261749s err= tokens=145
    live_test.go:48: pi-native reason-scope success=true p0=false targets=1/1 dur=3.180087344s err= tokens=142
    live_test.go:48: pi-agent doc-brief success=true p0=false targets=1/1 dur=5.974404302s err= tokens=0
    live_test.go:48: pi-agent doc-outline success=true p0=false targets=1/1 dur=6.926484503s err= tokens=0
    live_test.go:48: pi-agent dialogue-hello success=false p0=false targets=0/1 dur=4.726537486s err=MISSING_TARGET tokens=0
    live_test.go:48: pi-agent dialogue-status success=false p0=false targets=0/1 dur=3.02361258s err=MISSING_TARGET tokens=0
    live_test.go:48: pi-agent reason-risk success=false p0=false targets=0/1 dur=2.696804095s err=MISSING_TARGET tokens=0
    live_test.go:48: pi-agent reason-scope success=false p0=false targets=0/1 dur=2.828393237s err=MISSING_TARGET tokens=0
    live_test.go:52: replacement gate failed: [pi-native P95 latency exceeds 1.3x pi-agent]
--- FAIL: TestLiveDualBackend (62.25s)
FAIL
FAIL	chat-agent-system/internal/agentx/abtest	62.256s
FAIL
