feat: add unified jailbreak classifier with LazyLock optimization and benchmarks

- Implement unified jailbreak model factory with auto-detection from config.json
  - Supports ModernBERT, DeBERTa V3, and Qwen3Guard models
  - Automatic architecture detection and model loading

- Performance optimizations:
  - Use LazyLock for static default labels (zero-cost after init)
  - Use parking_lot::Mutex instead of std::sync::Mutex for Qwen3Guard
  - Lock-free classification for ModernBERT and DeBERTa (Arc-wrapped)
  - Early lock release in Qwen3Guard to minimize hold time

- Add comprehensive Go benchmark suite:
  - Test all jailbreak models (ModernBERT, DeBERTa, Unified, Qwen3Guard)
  - Measure accuracy, confidence, and latency percentiles (p50/p95/p99)
  - Test both CPU and GPU performance
  - Benchmark results show DeBERTa V3 achieves 95% accuracy

- Update FFI bindings:
  - Add init_unified_jailbreak_classifier and classify_unified_jailbreak_text
  - Update ClassificationResult to include label field
  - Rename 'class' to 'predicted_class' for consistency

- Add HuggingFace model ID support in unified factory
  - Auto-fetch config.json from HuggingFace Hub
  - Support both local paths and HF model IDs

- Add unit tests for unified jailbreak classifier
  - Tests in semantic-router_test.go, config_test.go, extproc_test.go
  - Fix test compilation errors with proper struct field usage

- Update Go interfaces to use unified classifier by default
  - Deprecate useModernBERT flag in favor of auto-detection

Signed-off-by: Yue Zhu <[email protected]>

Author: yuezhu1

bench/.gitignore

@@ -0,0 +1,17 @@
+# Benchmark results
+results/*.txt
+results/*.prof
+results/*.out
+
+# Test binaries
+*.test
+jailbreak_bench.test
+
+# Go coverage files
+*.coverprofile
+coverage.html
+
+# Temporary files
+*.tmp
+*.log
+

bench/Makefile

@@ -0,0 +1,97 @@
+.PHONY: all bench bench-quick bench-full bench-concurrent bench-memory bench-compare clean help
+
+# Default target
+all: help
+
+# Run quick benchmarks (shorter time)
+bench-quick:
+	@echo "Running quick jailbreak benchmarks..."
+	go test -bench=. -benchmem -benchtime=3s
+
+# Run full benchmarks (longer time for more accuracy)
+bench-full:
+	@echo "Running full jailbreak benchmarks..."
+	@mkdir -p results
+	go test -bench=. -benchmem -benchtime=10s | tee results/bench_$$(date +%Y%m%d_%H%M%S).txt
+
+# Run only concurrent benchmarks
+bench-concurrent:
+	@echo "Running concurrency benchmarks..."
+	go test -bench=Concurrent -benchmem -benchtime=30s
+
+# Run with memory profiling
+bench-memory:
+	@echo "Running benchmarks with memory profiling..."
+	@mkdir -p results
+	go test -bench=. -benchmem -memprofile=results/mem.prof -cpuprofile=results/cpu.prof
+	@echo "View memory profile: go tool pprof results/mem.prof"
+	@echo "View CPU profile: go tool pprof results/cpu.prof"
+
+# Compare with previous results (requires benchstat)
+bench-compare:
+	@if ! command -v benchstat >/dev/null 2>&1; then \
+		echo "Installing benchstat..."; \
+		go install golang.org/x/perf/cmd/benchstat@latest; \
+	fi
+	@if [ -z "$$(ls -t results/bench_*.txt 2>/dev/null | head -2 | tail -1)" ]; then \
+		echo "No previous results found. Run 'make bench-full' first."; \
+		exit 1; \
+	fi
+	@echo "Comparing with previous results..."
+	@OLD=$$(ls -t results/bench_*.txt | head -2 | tail -1); \
+	 NEW=$$(ls -t results/bench_*.txt | head -1); \
+	 echo "Old: $$OLD"; \
+	 echo "New: $$NEW"; \
+	 benchstat $$OLD $$NEW
+
+# Run benchmarks for specific model
+bench-modernbert:
+	@echo "Running ModernBERT benchmarks..."
+	go test -bench=BenchmarkModernBert -benchmem
+
+bench-deberta:
+	@echo "Running DeBERTa benchmarks..."
+	go test -bench=BenchmarkDeberta -benchmem
+
+bench-unified:
+	@echo "Running Unified classifier benchmarks..."
+	go test -bench=BenchmarkUnified -benchmem
+
+# Run only initialization benchmarks
+bench-init:
+	@echo "Running initialization benchmarks..."
+	go test -bench=BenchmarkInit -benchmem
+
+# Run benchmarks with race detector (slower but checks for race conditions)
+bench-race:
+	@echo "Running benchmarks with race detector..."
+	go test -race -bench=Concurrent -benchtime=5s
+
+# Clean results
+clean:
+	@echo "Cleaning benchmark results..."
+	rm -rf results/*.txt results/*.prof
+	rm -f /tmp/jailbreak_bench.test
+
+# Help target
+help:
+	@echo "Jailbreak Classifier Benchmarks"
+	@echo ""
+	@echo "Available targets:"
+	@echo "  make bench-quick       - Run quick benchmarks (3s each)"
+	@echo "  make bench-full        - Run full benchmarks (10s each, save results)"
+	@echo "  make bench-concurrent  - Run concurrency benchmarks only"
+	@echo "  make bench-memory      - Run with CPU and memory profiling"
+	@echo "  make bench-compare     - Compare with previous results"
+	@echo ""
+	@echo "Model-specific benchmarks:"
+	@echo "  make bench-modernbert  - Benchmark ModernBERT only"
+	@echo "  make bench-deberta     - Benchmark DeBERTa only"
+	@echo "  make bench-unified     - Benchmark Unified classifier only"
+	@echo ""
+	@echo "Other targets:"
+	@echo "  make bench-init        - Benchmark initialization only"
+	@echo "  make bench-race        - Run with race detector"
+	@echo "  make clean             - Clean benchmark results"
+	@echo "  make help              - Show this help message"
+