Fix characterization tests: portable imports + tmp_path data_dir

Characterization tests were written during an old refactor and hardcoded:
- sys.path manipulation referencing C:/ROAMPAL-REFACTOR (long-defunct dir)
- data_dir="C:/ROAMPAL/ui-implementation/src-tauri/binaries/data" absolute path

On CI runners (D:\a\...) both of these fail. Locally they worked only because those paths happened to exist on the dev machine.

Changes:
- Replaced get_original_memory_system() / get_original_action_outcome() / get_original_wilson_score() helpers with plain top-of-file imports. The sys.path juggling was scaffolding to swap between a pre-refactor tree and a post-refactor tree — that refactor shipped in v0.3.1 so the juggling is useless.
- Replaced hardcoded data_dir with pytest's tmp_path fixture. Fixture scope bumped from "class" to default so each test gets a fresh temp directory.

Kept every test assertion — threshold values, method existence, API shape checks, ActionOutcome dataclass behavior — because those are real regression guards worth keeping.

23 characterization tests pass (was 0 collectable on fresh env). Full backend: 529 passed / 1 skipped / 0 failed.

Author: roampal-ai

ui-implementation/src-tauri/backend/tests/characterization/test_outcome_behavior.py

@@ -2,88 +2,24 @@
 Characterization Tests for UnifiedMemorySystem outcome recording.
 
 These tests capture the CURRENT behavior of record_outcome() and related methods.
-They serve as a regression safety net during refactoring.
-
-Uses lazy imports inside fixtures to ensure path is set correctly.
+They serve as a regression safety net against accidental threshold / API drift.
 """
 
 import asyncio
 from datetime import datetime
 
 import pytest
 
-
-def clear_memory_modules():
-    """Clear all cached modules.memory.* modules to force fresh import."""
-    import sys
-    to_delete = [key for key in sys.modules if key.startswith("modules.memory") or key == "modules"]
-    for key in to_delete:
-        del sys.modules[key]
-
-
-def get_original_memory_system():
-    """Import and return the original UnifiedMemorySystem.
-
-    Uses importlib to ensure fresh import with correct path.
-    """
-    import sys
-
-    # Clear cached modules first
-    clear_memory_modules()
-
-    # Remove refactor path if present and ensure original is first
-    refactor_path = "C:/ROAMPAL-REFACTOR"
-    backend_path = "C:/ROAMPAL/ui-implementation/src-tauri/backend"
-
-    # Remove refactor path if it's there
-    while refactor_path in sys.path:
-        sys.path.remove(refactor_path)
-
-    # Ensure original backend path is first
-    if backend_path not in sys.path:
-        sys.path.insert(0, backend_path)
-    elif sys.path.index(backend_path) != 0:
-        sys.path.remove(backend_path)
-        sys.path.insert(0, backend_path)
-
-    from modules.memory.unified_memory_system import UnifiedMemorySystem
-    return UnifiedMemorySystem
-
-
-def get_original_action_outcome():
-    """Import and return the original ActionOutcome dataclass."""
-    import sys
-
-    # Clear cached modules first
-    clear_memory_modules()
-
-    # Remove refactor path if present and ensure original is first
-    refactor_path = "C:/ROAMPAL-REFACTOR"
-    backend_path = "C:/ROAMPAL/ui-implementation/src-tauri/backend"
-
-    # Remove refactor path if it's there
-    while refactor_path in sys.path:
-        sys.path.remove(refactor_path)
-
-    # Ensure original backend path is first
-    if backend_path not in sys.path:
-        sys.path.insert(0, backend_path)
-    elif sys.path.index(backend_path) != 0:
-        sys.path.remove(backend_path)
-        sys.path.insert(0, backend_path)
-
-    from modules.memory.unified_memory_system import ActionOutcome
-    return ActionOutcome
+from modules.memory.unified_memory_system import UnifiedMemorySystem, ActionOutcome
 
 
 class TestOutcomeBehavior:
     """Capture current outcome recording behavior."""
 
-    @pytest.fixture(scope="class")
-    def memory_system(self):
-        UnifiedMemorySystem = get_original_memory_system()
+    @pytest.fixture
+    def memory_system(self, tmp_path):
         ms = UnifiedMemorySystem(
-            data_dir="C:/ROAMPAL/ui-implementation/src-tauri/binaries/data",
+            data_dir=str(tmp_path),
             use_server=False
         )
         return ms
@@ -139,11 +75,10 @@ def test_threshold_values(self, memory_system):
 class TestPromotionBehavior:
     """Capture current promotion/demotion behavior."""
 
-    @pytest.fixture(scope="class")
-    def memory_system(self):
-        UnifiedMemorySystem = get_original_memory_system()
+    @pytest.fixture
+    def memory_system(self, tmp_path):
         ms = UnifiedMemorySystem(
-            data_dir="C:/ROAMPAL/ui-implementation/src-tauri/binaries/data",
+            data_dir=str(tmp_path),
             use_server=False
         )
         return ms
@@ -168,12 +103,10 @@ class TestActionOutcome:
 
     def test_action_outcome_import(self):
         """ActionOutcome should be importable."""
-        ActionOutcome = get_original_action_outcome()
         assert ActionOutcome is not None
 
     def test_action_outcome_fields(self):
         """ActionOutcome should have expected fields."""
-        ActionOutcome = get_original_action_outcome()
 
         # Create instance with actual API (v0.2.1 Causal Learning)
         ao = ActionOutcome(
@@ -188,7 +121,6 @@ def test_action_outcome_fields(self):
 
     def test_action_outcome_to_dict(self):
         """ActionOutcome.to_dict() should work."""
-        ActionOutcome = get_original_action_outcome()
 
         ao = ActionOutcome(
             action_type="search_memory",
@@ -205,7 +137,6 @@ def test_action_outcome_to_dict(self):
 
     def test_action_outcome_from_dict(self):
         """ActionOutcome.from_dict() should work."""
-        ActionOutcome = get_original_action_outcome()
 
         data = {
             "action_type": "search_memory",

ui-implementation/src-tauri/backend/tests/characterization/test_search_behavior.py

@@ -1,14 +1,8 @@
-r"""
+"""
 Characterization Tests for UnifiedMemorySystem.search()
 
 These tests capture the CURRENT behavior of the search() method.
-They serve as a regression safety net during refactoring.
-
-Uses lazy imports inside fixtures to ensure path is set correctly.
-
-Usage:
-    cd C:/ROAMPAL-REFACTOR
-    python -m pytest tests/characterization/ -v
+They serve as a regression safety net against API / return-shape drift.
 """
 
 import asyncio
@@ -18,68 +12,8 @@
 
 import pytest
 
-
-def clear_memory_modules():
-    """Clear all cached modules.memory.* modules to force fresh import."""
-    import sys
-    to_delete = [key for key in sys.modules if key.startswith("modules.memory") or key == "modules"]
-    for key in to_delete:
-        del sys.modules[key]
-
-
-def get_original_memory_system():
-    """Import and return the original UnifiedMemorySystem.
-
-    Uses importlib to ensure fresh import with correct path.
-    """
-    import sys
-
-    # Clear cached modules first
-    clear_memory_modules()
-
-    # Remove refactor path if present and ensure original is first
-    refactor_path = "C:/ROAMPAL-REFACTOR"
-    backend_path = "C:/ROAMPAL/ui-implementation/src-tauri/backend"
-
-    # Remove refactor path if it's there
-    while refactor_path in sys.path:
-        sys.path.remove(refactor_path)
-
-    # Ensure original backend path is first
-    if backend_path not in sys.path:
-        sys.path.insert(0, backend_path)
-    elif sys.path.index(backend_path) != 0:
-        sys.path.remove(backend_path)
-        sys.path.insert(0, backend_path)
-
-    from modules.memory.unified_memory_system import UnifiedMemorySystem
-    return UnifiedMemorySystem
-
-
-def get_original_wilson_score():
-    """Import and return the original wilson_score_lower function."""
-    import sys
-
-    # Clear cached modules first
-    clear_memory_modules()
-
-    # Remove refactor path if present and ensure original is first
-    refactor_path = "C:/ROAMPAL-REFACTOR"
-    backend_path = "C:/ROAMPAL/ui-implementation/src-tauri/backend"
-
-    # Remove refactor path if it's there
-    while refactor_path in sys.path:
-        sys.path.remove(refactor_path)
-
-    # Ensure original backend path is first
-    if backend_path not in sys.path:
-        sys.path.insert(0, backend_path)
-    elif sys.path.index(backend_path) != 0:
-        sys.path.remove(backend_path)
-        sys.path.insert(0, backend_path)
-
-    from modules.memory.scoring_service import wilson_score_lower
-    return wilson_score_lower
+from modules.memory.unified_memory_system import UnifiedMemorySystem
+from modules.memory.scoring_service import wilson_score_lower
 
 
 # Characterization queries from the plan
@@ -98,13 +32,11 @@ def get_original_wilson_score():
 class TestSearchBehavior:
     """Capture current search() behavior for regression testing."""
 
-    @pytest.fixture(scope="class")
-    def memory_system(self):
-        """Create a single instance for all tests in this class."""
-        UnifiedMemorySystem = get_original_memory_system()
-        # Use production data directory (read-only access)
+    @pytest.fixture
+    def memory_system(self, tmp_path):
+        """Create a single instance for each test."""
         ms = UnifiedMemorySystem(
-            data_dir="C:/ROAMPAL/ui-implementation/src-tauri/binaries/data",
+            data_dir=str(tmp_path),
             use_server=False  # Don't start embedding server
         )
         return ms
@@ -191,18 +123,15 @@ class TestWilsonScore:
 
     def test_wilson_score_import(self):
         """Wilson score function should be importable."""
-        wilson_score_lower = get_original_wilson_score()
         assert callable(wilson_score_lower)
 
     def test_wilson_score_zero_uses(self):
         """Zero uses should return 0.5 (neutral)."""
-        wilson_score_lower = get_original_wilson_score()
         score = wilson_score_lower(0, 0)
         assert score == 0.5, f"Expected 0.5 for zero uses, got {score}"
 
     def test_wilson_score_perfect_record(self):
         """Perfect record with few uses should be lower than proven record."""
-        wilson_score_lower = get_original_wilson_score()
 
         # 1/1 = 100% but low confidence
         new_score = wilson_score_lower(1, 1)
@@ -216,7 +145,6 @@ def test_wilson_score_perfect_record(self):
 
     def test_wilson_score_range(self):
         """Wilson score should always be between 0 and 1."""
-        wilson_score_lower = get_original_wilson_score()
 
         test_cases = [
             (0, 1),
@@ -235,11 +163,10 @@ def test_wilson_score_range(self):
 class TestMCPToolShapes:
     """Verify MCP tool response shapes match expected format."""
 
-    @pytest.fixture(scope="class")
-    def memory_system(self):
-        UnifiedMemorySystem = get_original_memory_system()
+    @pytest.fixture
+    def memory_system(self, tmp_path):
         ms = UnifiedMemorySystem(
-            data_dir="C:/ROAMPAL/ui-implementation/src-tauri/binaries/data",
+            data_dir=str(tmp_path),
             use_server=False
         )
         return ms