chore: restructure test env, cassettes, and conftest; fix flaky tests

Consolidates pytest config, standardizes env handling, reorganizes cassette layout, removes outdated VCR configs, improves sync with threading.Condition, updates event-waiting logic, ensures cleanup, regenerates Gemini cassettes, and reverts unintended test changes.

lib/crewai/tests/memory/test_external_memory.py

@@ -194,7 +194,7 @@ def test_crew_external_memory_reset(mem_type, crew_with_external_memory):
 
 
 @pytest.mark.parametrize("mem_method", ["search", "save"])
-@pytest.mark.vcr(filter_headers=["authorization"])
+@pytest.mark.vcr()
 def test_crew_external_memory_save_with_memory_flag(
     mem_method, crew_with_external_memory
 ):
@@ -206,7 +206,7 @@ def test_crew_external_memory_save_with_memory_flag(
 
 
 @pytest.mark.parametrize("mem_method", ["search", "save"])
-@pytest.mark.vcr(filter_headers=["authorization"])
+@pytest.mark.vcr()
 def test_crew_external_memory_save_using_crew_without_memory_flag(
     mem_method, crew_with_external_memory_without_memory_flag
 ):
@@ -259,19 +259,22 @@ def test_external_memory_custom_storage(custom_storage, crew_with_external_memor
 def test_external_memory_search_events(
     custom_storage, external_memory_with_mocked_config
 ):
-    events = defaultdict(list)
-    event_received = threading.Event()
+    events: dict[str, list] = defaultdict(list)
+    condition = threading.Condition()
 
     external_memory_with_mocked_config.storage = custom_storage
 
     @crewai_event_bus.on(MemoryQueryStartedEvent)
     def on_search_started(source, event):
-        events["MemoryQueryStartedEvent"].append(event)
+        with condition:
+            events["MemoryQueryStartedEvent"].append(event)
+            condition.notify()
 
     @crewai_event_bus.on(MemoryQueryCompletedEvent)
     def on_search_completed(source, event):
-        events["MemoryQueryCompletedEvent"].append(event)
-        event_received.set()
+        with condition:
+            events["MemoryQueryCompletedEvent"].append(event)
+            condition.notify()
 
     external_memory_with_mocked_config.search(
         query="test value",
@@ -279,7 +282,13 @@ def test_external_memory_search_events(
         score_threshold=0.35,
     )
 
-    assert event_received.wait(timeout=5), "Timeout waiting for search events"
+    with condition:
+        success = condition.wait_for(
+            lambda: len(events["MemoryQueryStartedEvent"]) >= 1
+            and len(events["MemoryQueryCompletedEvent"]) >= 1,
+            timeout=10,
+        )
+    assert success, "Timeout waiting for search events"
     assert len(events["MemoryQueryStartedEvent"]) == 1
     assert len(events["MemoryQueryCompletedEvent"]) == 1
 
@@ -323,26 +332,35 @@ def test_external_memory_search_events(
 def test_external_memory_save_events(
     custom_storage, external_memory_with_mocked_config
 ):
-    events = defaultdict(list)
-    event_received = threading.Event()
+    events: dict[str, list] = defaultdict(list)
+    condition = threading.Condition()
 
     external_memory_with_mocked_config.storage = custom_storage
 
     @crewai_event_bus.on(MemorySaveStartedEvent)
     def on_save_started(source, event):
-        events["MemorySaveStartedEvent"].append(event)
+        with condition:
+            events["MemorySaveStartedEvent"].append(event)
+            condition.notify()
 
     @crewai_event_bus.on(MemorySaveCompletedEvent)
     def on_save_completed(source, event):
-        events["MemorySaveCompletedEvent"].append(event)
-        event_received.set()
+        with condition:
+            events["MemorySaveCompletedEvent"].append(event)
+            condition.notify()
 
     external_memory_with_mocked_config.save(
         value="saving value",
         metadata={"task": "test_task"},
     )
 
-    assert event_received.wait(timeout=5), "Timeout waiting for save events"
+    with condition:
+        success = condition.wait_for(
+            lambda: len(events["MemorySaveStartedEvent"]) >= 1
+            and len(events["MemorySaveCompletedEvent"]) >= 1,
+            timeout=10,
+        )
+    assert success, "Timeout waiting for save events"
     assert len(events["MemorySaveStartedEvent"]) == 1
     assert len(events["MemorySaveCompletedEvent"]) == 1
 

lib/crewai/tests/memory/test_long_term_memory.py

@@ -23,25 +23,19 @@ def long_term_memory():
 
 def test_long_term_memory_save_events(long_term_memory):
     events = defaultdict(list)
-    all_events_received = threading.Event()
+    condition = threading.Condition()
 
     @crewai_event_bus.on(MemorySaveStartedEvent)
     def on_save_started(source, event):
-        events["MemorySaveStartedEvent"].append(event)
-        if (
-            len(events["MemorySaveStartedEvent"]) == 1
-            and len(events["MemorySaveCompletedEvent"]) == 1
-        ):
-            all_events_received.set()
+        with condition:
+            events["MemorySaveStartedEvent"].append(event)
+            condition.notify()
 
     @crewai_event_bus.on(MemorySaveCompletedEvent)
     def on_save_completed(source, event):
-        events["MemorySaveCompletedEvent"].append(event)
-        if (
-            len(events["MemorySaveStartedEvent"]) == 1
-            and len(events["MemorySaveCompletedEvent"]) == 1
-        ):
-            all_events_received.set()
+        with condition:
+            events["MemorySaveCompletedEvent"].append(event)
+            condition.notify()
 
     memory = LongTermMemoryItem(
         agent="test_agent",
@@ -53,7 +47,13 @@ def test_long_term_memory_save_events(long_term_memory):
     )
     long_term_memory.save(memory)
 
-    assert all_events_received.wait(timeout=5), "Timeout waiting for save events"
+    with condition:
+        success = condition.wait_for(
+            lambda: len(events["MemorySaveStartedEvent"]) >= 1
+            and len(events["MemorySaveCompletedEvent"]) >= 1,
+            timeout=5,
+        )
+    assert success, "Timeout waiting for save events"
     assert len(events["MemorySaveStartedEvent"]) == 1
     assert len(events["MemorySaveCompletedEvent"]) == 1
     assert len(events["MemorySaveFailedEvent"]) == 0
@@ -98,31 +98,31 @@ def test_long_term_memory_save_events(long_term_memory):
 
 def test_long_term_memory_search_events(long_term_memory):
     events = defaultdict(list)
-    all_events_received = threading.Event()
+    condition = threading.Condition()
 
     @crewai_event_bus.on(MemoryQueryStartedEvent)
     def on_search_started(source, event):
-        events["MemoryQueryStartedEvent"].append(event)
-        if (
-            len(events["MemoryQueryStartedEvent"]) == 1
-            and len(events["MemoryQueryCompletedEvent"]) == 1
-        ):
-            all_events_received.set()
+        with condition:
+            events["MemoryQueryStartedEvent"].append(event)
+            condition.notify()
 
     @crewai_event_bus.on(MemoryQueryCompletedEvent)
     def on_search_completed(source, event):
-        events["MemoryQueryCompletedEvent"].append(event)
-        if (
-            len(events["MemoryQueryStartedEvent"]) == 1
-            and len(events["MemoryQueryCompletedEvent"]) == 1
-        ):
-            all_events_received.set()
+        with condition:
+            events["MemoryQueryCompletedEvent"].append(event)
+            condition.notify()
 
     test_query = "test query"
 
     long_term_memory.search(test_query, latest_n=5)
 
-    assert all_events_received.wait(timeout=5), "Timeout waiting for search events"
+    with condition:
+        success = condition.wait_for(
+            lambda: len(events["MemoryQueryStartedEvent"]) >= 1
+            and len(events["MemoryQueryCompletedEvent"]) >= 1,
+            timeout=5,
+        )
+    assert success, "Timeout waiting for search events"
     assert len(events["MemoryQueryStartedEvent"]) == 1
     assert len(events["MemoryQueryCompletedEvent"]) == 1
     assert len(events["MemoryQueryFailedEvent"]) == 0

lib/crewai/tests/memory/test_short_term_memory.py

@@ -107,27 +107,33 @@ def test_short_term_memory_search_events(short_term_memory):
 
 
 def test_short_term_memory_save_events(short_term_memory):
-    events = defaultdict(list)
-    save_started = threading.Event()
-    save_completed = threading.Event()
+    events: dict[str, list] = defaultdict(list)
+    condition = threading.Condition()
 
     @crewai_event_bus.on(MemorySaveStartedEvent)
     def on_save_started(source, event):
-        events["MemorySaveStartedEvent"].append(event)
-        save_started.set()
+        with condition:
+            events["MemorySaveStartedEvent"].append(event)
+            condition.notify()
 
     @crewai_event_bus.on(MemorySaveCompletedEvent)
     def on_save_completed(source, event):
-        events["MemorySaveCompletedEvent"].append(event)
-        save_completed.set()
+        with condition:
+            events["MemorySaveCompletedEvent"].append(event)
+            condition.notify()
 
     short_term_memory.save(
         value="test value",
         metadata={"task": "test_task"},
     )
 
-    assert save_started.wait(timeout=2), "Timeout waiting for save started event"
-    assert save_completed.wait(timeout=2), "Timeout waiting for save completed event"
+    with condition:
+        success = condition.wait_for(
+            lambda: len(events["MemorySaveStartedEvent"]) >= 1
+            and len(events["MemorySaveCompletedEvent"]) >= 1,
+            timeout=5,
+        )
+    assert success, "Timeout waiting for save events"
 
     assert len(events["MemorySaveStartedEvent"]) == 1
     assert len(events["MemorySaveCompletedEvent"]) == 1