Upload backend/scripts/benchmark_search.py with huggingface_hub

backend/scripts/benchmark_search.py

@@ -0,0 +1,104 @@
+import os
+import sys
+import time
+import json
+from pathlib import Path
+import statistics
+
+# Ensure project root on path
+ROOT_DIR = Path(__file__).resolve().parents[2]
+BACKEND_DIR = ROOT_DIR / "backend"
+HUE_PORTAL_DIR = BACKEND_DIR / "hue_portal"
+
+for path in (HUE_PORTAL_DIR, BACKEND_DIR, ROOT_DIR):
+    if str(path) not in sys.path:
+        sys.path.insert(0, str(path))
+
+os.environ.setdefault("DJANGO_SETTINGS_MODULE", "hue_portal.hue_portal.settings")
+
+import django
+
+django.setup()
+
+from django.db import connection
+from hue_portal.core.models import Procedure, Fine, Office, Advisory
+from hue_portal.core.search_ml import search_with_ml
+
+
+QUERIES = {
+    "procedure": [
+        "đăng ký cư trú",
+        "thủ tục pccc",
+        "giấy tờ antt",
+    ],
+    "fine": [
+        "mức phạt nồng độ cồn",
+        "vượt đèn đỏ",
+        "không đội mũ bảo hiểm",
+    ],
+    "office": [
+        "công an phường",
+        "điểm tiếp dân",
+    ],
+    "advisory": [
+        "cảnh báo lừa đảo",
+        "giả mạo công an",
+    ],
+}
+
+
+def run_benchmark(iterations: int = 3):
+    results = {
+        "database_vendor": connection.vendor,
+        "timestamp": time.time(),
+        "iterations": iterations,
+        "entries": [],
+    }
+
+    datasets = {
+        "procedure": (Procedure.objects.all(), ["title", "domain", "conditions", "dossier"]),
+        "fine": (Fine.objects.all(), ["name", "code", "article", "decree", "remedial"]),
+        "office": (Office.objects.all(), ["unit_name", "address", "district", "service_scope"]),
+        "advisory": (Advisory.objects.all(), ["title", "summary"]),
+    }
+
+    for dataset, queries in QUERIES.items():
+        qs, fields = datasets[dataset]
+        for query in queries:
+            durations = []
+            counts = []
+            for _ in range(iterations):
+                start = time.perf_counter()
+                items = list(search_with_ml(qs, query, fields, top_k=20))
+                durations.append(time.perf_counter() - start)
+                counts.append(len(items))
+
+            results["entries"].append(
+                {
+                    "dataset": dataset,
+                    "query": query,
+                    "avg_duration_ms": statistics.mean(durations) * 1000,
+                    "p95_duration_ms": statistics.quantiles(durations, n=20)[18] * 1000 if len(durations) >= 20 else max(durations) * 1000,
+                    "min_duration_ms": min(durations) * 1000,
+                    "max_duration_ms": max(durations) * 1000,
+                    "avg_results": statistics.mean(counts),
+                }
+            )
+
+    return results
+
+
+def main():
+    iterations = int(os.environ.get("BENCH_ITERATIONS", "3"))
+    benchmark = run_benchmark(iterations=iterations)
+
+    output_dir = ROOT_DIR / "logs" / "benchmarks"
+    output_dir.mkdir(parents=True, exist_ok=True)
+    output_file = output_dir / f"search_benchmark_{int(benchmark['timestamp'])}.json"
+    output_file.write_text(json.dumps(benchmark, ensure_ascii=False, indent=2))
+
+    print(f"Benchmark completed. Results saved to {output_file}")
+
+
+if __name__ == "__main__":
+    main()