Upload test_semantic_search.py with huggingface_hub
Browse files- test_semantic_search.py +152 -0
test_semantic_search.py
ADDED
|
@@ -0,0 +1,152 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
End-to-end smoke test for the Sentinel-AI API.
|
| 3 |
+
Make sure the backend is running before executing this script.
|
| 4 |
+
|
| 5 |
+
Usage:
|
| 6 |
+
python test_semantic_search.py
|
| 7 |
+
"""
|
| 8 |
+
|
| 9 |
+
import json
|
| 10 |
+
import os
|
| 11 |
+
import sys
|
| 12 |
+
from typing import Any
|
| 13 |
+
|
| 14 |
+
import requests
|
| 15 |
+
|
| 16 |
+
API_BASE = os.getenv("SENTINEL_API_BASE_URL", "http://localhost:8000").rstrip("/")
|
| 17 |
+
|
| 18 |
+
# Optional shared-deployment token (start the backend with SENTINEL_API_KEY).
|
| 19 |
+
_API_KEY = os.getenv("SENTINEL_API_KEY", "").strip()
|
| 20 |
+
_HEADERS = {"Authorization": f"Bearer {_API_KEY}"} if _API_KEY else {}
|
| 21 |
+
|
| 22 |
+
|
| 23 |
+
def _print_header(title: str) -> None:
|
| 24 |
+
print("\n" + "=" * 70)
|
| 25 |
+
print(title)
|
| 26 |
+
print("=" * 70)
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
def test_health() -> bool:
|
| 30 |
+
"""Test if the backend is running."""
|
| 31 |
+
_print_header("TEST: Health Check")
|
| 32 |
+
try:
|
| 33 |
+
response = requests.get(f"{API_BASE}/health", headers=_HEADERS, timeout=5)
|
| 34 |
+
response.raise_for_status()
|
| 35 |
+
data = response.json()
|
| 36 |
+
|
| 37 |
+
print("Backend is running")
|
| 38 |
+
print(f" Status: {data['status']}")
|
| 39 |
+
print(f" Index ready: {data['index_ready']}")
|
| 40 |
+
print(f" Model: {data['model']}")
|
| 41 |
+
return bool(data.get("status") == "ok")
|
| 42 |
+
except requests.exceptions.ConnectionError:
|
| 43 |
+
print(f"Cannot connect to {API_BASE}")
|
| 44 |
+
print(" Start the backend with:")
|
| 45 |
+
print(" cd backend && python -m uvicorn main:app --host 127.0.0.1 --port 8000 --reload")
|
| 46 |
+
return False
|
| 47 |
+
except Exception as exc:
|
| 48 |
+
print(f"Health check failed: {exc}")
|
| 49 |
+
return False
|
| 50 |
+
|
| 51 |
+
|
| 52 |
+
def test_semantic_search(query: str, top_k: int = 5) -> dict[str, Any] | None:
|
| 53 |
+
"""Run one semantic-search query and return the response (or None)."""
|
| 54 |
+
_print_header("TEST: Semantic Search")
|
| 55 |
+
print(f"Query: '{query}'")
|
| 56 |
+
print(f"Top K: {top_k}")
|
| 57 |
+
|
| 58 |
+
try:
|
| 59 |
+
response = requests.post(
|
| 60 |
+
f"{API_BASE}/api/v1/semantic-search",
|
| 61 |
+
json={"query": query, "top_k": top_k},
|
| 62 |
+
headers=_HEADERS,
|
| 63 |
+
timeout=30,
|
| 64 |
+
)
|
| 65 |
+
response.raise_for_status()
|
| 66 |
+
data = response.json()
|
| 67 |
+
|
| 68 |
+
attrs = data.get("extracted_attrs") or []
|
| 69 |
+
excluded = data.get("excluded_attrs") or []
|
| 70 |
+
matches = data.get("matches") or []
|
| 71 |
+
|
| 72 |
+
print("Request successful")
|
| 73 |
+
print(f" Status: {data.get('status')}")
|
| 74 |
+
print(f" Extracted attributes: {', '.join(attrs) if attrs else 'None'}")
|
| 75 |
+
print(f" Excluded attributes: {', '.join(excluded) if excluded else 'None'}")
|
| 76 |
+
print(f" Matches found: {len(matches)}")
|
| 77 |
+
|
| 78 |
+
metrics = data.get("cascadeflow_metrics")
|
| 79 |
+
if metrics:
|
| 80 |
+
print(f" Latency: {metrics.get('latency_ms', 0):.1f} ms")
|
| 81 |
+
print(f" Cost: ${metrics.get('cost_usd', 0):.6f}")
|
| 82 |
+
|
| 83 |
+
if matches:
|
| 84 |
+
print(f"\nTop {min(3, len(matches))} matches:")
|
| 85 |
+
for i, match in enumerate(matches[:3], 1):
|
| 86 |
+
shown = ", ".join(match["attributes"][:5])
|
| 87 |
+
print(f" {i}. Image ID: {match['image_id']} (faiss_id={match['faiss_id']})")
|
| 88 |
+
print(f" Attributes: {shown}...")
|
| 89 |
+
|
| 90 |
+
return data
|
| 91 |
+
except requests.exceptions.Timeout:
|
| 92 |
+
print("Request timed out (>30 seconds)")
|
| 93 |
+
return None
|
| 94 |
+
except requests.exceptions.HTTPError as exc:
|
| 95 |
+
print(f"HTTP Error: {exc}")
|
| 96 |
+
try:
|
| 97 |
+
print(f" Detail: {json.dumps(response.json(), indent=2)}")
|
| 98 |
+
except Exception:
|
| 99 |
+
print(f" Response: {response.text}")
|
| 100 |
+
return None
|
| 101 |
+
except Exception as exc:
|
| 102 |
+
print(f"Request failed: {exc}")
|
| 103 |
+
return None
|
| 104 |
+
|
| 105 |
+
|
| 106 |
+
def main() -> int:
|
| 107 |
+
"""Run all tests and return a process exit code."""
|
| 108 |
+
_print_header("SENTINEL-AI SEMANTIC SEARCH TEST SUITE")
|
| 109 |
+
|
| 110 |
+
if not test_health():
|
| 111 |
+
print("\nBackend is not running. Please start it first.")
|
| 112 |
+
return 1
|
| 113 |
+
|
| 114 |
+
results: list[tuple[str, bool]] = []
|
| 115 |
+
|
| 116 |
+
# A response is only a pass if the request succeeded *and* the extracted
|
| 117 |
+
# attributes are sensible for the query.
|
| 118 |
+
data = test_semantic_search("smiling woman with blond hair", top_k=5)
|
| 119 |
+
ok = bool(data) and "Male" in data.get("excluded_attrs", [])
|
| 120 |
+
if ok:
|
| 121 |
+
ok = not any("Male" in m.get("attributes", []) for m in data.get("matches", []))
|
| 122 |
+
results.append(("Smiling woman query (excludes Male from results)", ok))
|
| 123 |
+
|
| 124 |
+
data = test_semantic_search("young male with wavy hair and eyeglasses", top_k=5)
|
| 125 |
+
ok = bool(data) and {"Male", "Young"} <= set(data.get("extracted_attrs", []))
|
| 126 |
+
results.append(("Young male with glasses query", ok))
|
| 127 |
+
|
| 128 |
+
data = test_semantic_search("person with beard and mustache", top_k=5)
|
| 129 |
+
ok = bool(data) and "No_Beard" not in data.get("extracted_attrs", [])
|
| 130 |
+
results.append(("Beard/mustache query (no No_Beard inversion)", ok))
|
| 131 |
+
|
| 132 |
+
data = test_semantic_search("alien with three eyes", top_k=5)
|
| 133 |
+
ok = bool(data) and not data.get("matches")
|
| 134 |
+
results.append(("Edge-case query returns no matches", ok))
|
| 135 |
+
|
| 136 |
+
_print_header("SUMMARY")
|
| 137 |
+
for label, passed in results:
|
| 138 |
+
print(f" [{'PASS' if passed else 'FAIL'}] {label}")
|
| 139 |
+
|
| 140 |
+
passed_count = sum(1 for _, passed in results if passed)
|
| 141 |
+
print(f"\nTests passed: {passed_count}/{len(results)}")
|
| 142 |
+
|
| 143 |
+
if passed_count == len(results):
|
| 144 |
+
print("\nALL TESTS PASSED")
|
| 145 |
+
return 0
|
| 146 |
+
|
| 147 |
+
print(f"\n{len(results) - passed_count} test(s) failed.")
|
| 148 |
+
return 1
|
| 149 |
+
|
| 150 |
+
|
| 151 |
+
if __name__ == "__main__":
|
| 152 |
+
sys.exit(main())
|