fix: Isolate test database to prevent production cache.db access (#116)

This commit resolves issue #115 by implementing test database isolation:

1. Added auto-use fixture `isolated_test_cache` in tests/conftest.py that:
   - Creates a temporary test database for each test function
   - Sets it as the global cache manager via set_cache_manager()
   - Automatically cleans up after test completion via reset_cache_manager()

2. Updated cache instantiation to use global cache manager:
   - Changed normalizer.py to use get_cache_manager() instead of CacheManager()
   - Changed dispatcher.py to use get_cache_manager() instead of CacheManager()

These changes ensure:
- Tests never access the production cache.db in .aletheia-probe/
- Each test gets a clean, isolated database
- No test pollution between test runs
- Production database schema changes don't break tests

All 323 tests now pass successfully.

Fixes #115

Co-authored-by: florath-ai-assistant[bot] <Andreas.Florath@telekom.de>

Author: coding-ai-assistant[bot]

src/aletheia_probe/dispatcher.py

@@ -107,11 +107,11 @@ async def assess_journal(self, query_input: QueryInput) -> AssessmentResult:
         # Acronym fallback: If initial query yields no confident results and input looks
         # like an acronym with a cached expansion, retry with the expanded name
         if self._should_try_acronym_fallback(assessment_result, query_input):
-            from .cache import CacheManager
+            from .cache import get_cache_manager
             from .normalizer import InputNormalizer
 
             normalizer = InputNormalizer()
-            cache = CacheManager()
+            cache = get_cache_manager()
 
             # Check if input is acronym-like and has expansion
             if normalizer._is_standalone_acronym(query_input.raw_input):

src/aletheia_probe/normalizer.py

@@ -487,9 +487,9 @@ def _lookup_acronym_expansion(self, acronym: str) -> str | None:
         Returns:
             Full name if found, None otherwise
         """
-        from .cache import CacheManager
+        from .cache import get_cache_manager
 
-        cache = CacheManager()
+        cache = get_cache_manager()
         return cache.get_full_name_for_acronym(acronym)
 
     def _store_acronym_mappings_from_text(
@@ -507,9 +507,9 @@ def _store_acronym_mappings_from_text(
         if not extracted_acronyms:
             return
 
-        from .cache import CacheManager
+        from .cache import get_cache_manager
 
-        cache = CacheManager()
+        cache = get_cache_manager()
 
         # For each acronym, try to extract the full name before the parentheses
         for acronym in extracted_acronyms:

tests/conftest.py

@@ -1,14 +1,45 @@
 # SPDX-License-Identifier: MIT
 """Pytest configuration and shared fixtures."""
 
+import tempfile
 from pathlib import Path
 
 import pytest
 
 import aletheia_probe.backends  # Import backends to register them
+from aletheia_probe.cache import CacheManager, reset_cache_manager, set_cache_manager
 from aletheia_probe.models import BackendResult, BackendStatus, QueryInput
 
 
+@pytest.fixture(scope="function", autouse=True)
+def isolated_test_cache(tmp_path):
+    """
+    Automatically provide an isolated test cache for every test.
+
+    This fixture:
+    1. Creates a temporary database file for each test
+    2. Sets it as the global cache manager
+    3. Cleans up after the test completes
+
+    This prevents tests from accessing the production cache.db file.
+    """
+    # Create a temporary database file
+    cache_path = tmp_path / "test_cache.db"
+
+    # Create a cache manager instance with the test database
+    test_cache = CacheManager(db_path=cache_path)
+
+    # Set it as the global cache manager
+    set_cache_manager(test_cache)
+
+    # Yield control to the test
+    yield test_cache
+
+    # Clean up after the test
+    reset_cache_manager()
+    # The tmp_path fixture automatically cleans up the temp directory
+
+
 @pytest.fixture
 def sample_query_input():
     """Sample QueryInput for testing."""