fix: stochastic failure in test_autonomous_goals.py

tests/e2e/test_system_goap_navigation.py

@@ -344,30 +344,30 @@ class TestActionMasking:
 
     def test_action_masked_after_max_retries(self):
         """After 2 failures, the action must be excluded from available_actions."""
-        action_failures = {"tap reels tab": 2}
+        action_failures = {(ScreenType.HOME_FEED, "tap reels tab"): 2}
         original_actions = ["tap home tab", "tap reels tab", "tap profile tab"]
         MAX_RETRIES = 2
 
-        masked = [a for a in original_actions if action_failures.get(a, 0) < MAX_RETRIES]
+        masked = [a for a in original_actions if action_failures.get((ScreenType.HOME_FEED, a), 0) < MAX_RETRIES]
         assert "tap reels tab" not in masked
         assert "tap home tab" in masked
         assert "tap profile tab" in masked
 
     def test_action_not_masked_under_threshold(self):
         """1 failure is not enough to mask."""
-        action_failures = {"tap reels tab": 1}
+        action_failures = {(ScreenType.HOME_FEED, "tap reels tab"): 1}
         original_actions = ["tap reels tab", "tap profile tab"]
         MAX_RETRIES = 2
 
-        masked = [a for a in original_actions if action_failures.get(a, 0) < MAX_RETRIES]
+        masked = [a for a in original_actions if action_failures.get((ScreenType.HOME_FEED, a), 0) < MAX_RETRIES]
         assert "tap reels tab" in masked
 
     def test_success_resets_failure_count(self):
         """A successful execution must reset the failure counter."""
-        action_failures = {"tap reels tab": 1}
+        action_failures = {(ScreenType.HOME_FEED, "tap reels tab"): 1}
         # Simulate success
-        action_failures["tap reels tab"] = 0
-        assert action_failures["tap reels tab"] == 0
+        action_failures[(ScreenType.HOME_FEED, "tap reels tab")] = 0
+        assert action_failures[(ScreenType.HOME_FEED, "tap reels tab")] == 0
 
     def test_hd_map_unreachable_with_masked_actions(self):
         """

tests/unit/test_autonomous_goals.py

@@ -30,15 +30,15 @@ def test_autonomous_goal_weighting():
     available_goals = ["goal_A", "goal_B", "goal_C"]
 
     # Simulate that goal_B has been incredibly successful, goal_A moderately, goal_C not at all.
-    success_rates = {"goal_A": 2, "goal_B": 100, "goal_C": 0}
+    success_rates = {"goal_A": 50, "goal_B": 500, "goal_C": 0}
 
     # If weighting works, running this many times should result in goal_B being chosen overwhelmingly
     choices = {"goal_A": 0, "goal_B": 0, "goal_C": 0}
-    for _ in range(100):
+    for _ in range(1000):
         # We pass success_rates to get_current_goal
         choice = brain.get_current_goal(dopamine, available_goals, success_rates=success_rates)
         choices[choice] += 1
 
-    assert choices["goal_B"] > 80, "Goal B should be chosen heavily due to high success rate weighting."
-    assert choices["goal_A"] < 20, "Goal A should be chosen rarely."
-    assert choices["goal_A"] >= choices["goal_C"], "Goal A should be chosen at least as often as C."
+    assert choices["goal_B"] > 800, f"Goal B should be chosen heavily: {choices['goal_B']}"
+    assert choices["goal_A"] > 50, f"Goal A should be chosen sometimes: {choices['goal_A']}"
+    assert choices["goal_C"] < 50, f"Goal C should be chosen rarely: {choices['goal_C']}"