=== RUN   TestLiveDualBackend
    live_test.go:45: native_success=1.00 remote_success=0.33 native_p95=9.809172339s remote_p95=5.47877184s p0=0
    live_test.go:48: pi-native doc-brief success=true p0=false targets=1/1 dur=9.605385216s err= tokens=0
    live_test.go:48: pi-native doc-outline success=true p0=false targets=1/1 dur=17.922113116s err= tokens=0
    live_test.go:48: pi-native dialogue-hello success=true p0=false targets=1/1 dur=3.145449606s err= tokens=0
    live_test.go:48: pi-native dialogue-status success=true p0=false targets=1/1 dur=4.785127435s err= tokens=0
    live_test.go:48: pi-native reason-risk success=true p0=false targets=1/1 dur=3.695562766s err= tokens=0
    live_test.go:48: pi-native reason-scope success=true p0=false targets=1/1 dur=9.809172339s err= tokens=0
    live_test.go:48: pi-agent doc-brief success=true p0=false targets=1/1 dur=15.794770878s err= tokens=0
    live_test.go:48: pi-agent doc-outline success=true p0=false targets=1/1 dur=5.47877184s err= tokens=0
    live_test.go:48: pi-agent dialogue-hello success=false p0=false targets=0/1 dur=3.165209406s err=MISSING_TARGET tokens=0
    live_test.go:48: pi-agent dialogue-status success=false p0=false targets=0/1 dur=3.478807754s err=MISSING_TARGET tokens=0
    live_test.go:48: pi-agent reason-risk success=false p0=false targets=0/1 dur=3.909467737s err=MISSING_TARGET tokens=0
    live_test.go:48: pi-agent reason-scope success=false p0=false targets=0/1 dur=3.077720644s err=MISSING_TARGET tokens=0
    live_test.go:52: replacement gate failed: [pi-native P95 latency exceeds 1.3x pi-agent]
--- FAIL: TestLiveDualBackend (83.87s)
FAIL
FAIL	chat-agent-system/internal/agentx/abtest	83.872s
FAIL
