Fix STLTMemory consolidation discarding the oldest memory entry before LLM summarization

mesa_llm/memory/st_lt_memory.py

@@ -136,7 +136,6 @@ def _process_step_core(self, pre_step: bool):
             > self.capacity + (self.consolidation_capacity or 0)
             and self.consolidation_capacity
         ):
-            self.short_term_memory.popleft()
             should_consolidate = True
 
         elif (
@@ -155,6 +154,7 @@ def process_step(self, pre_step: bool = False):
 
         if should_consolidate:
             self._update_long_term_memory()
+            self.short_term_memory.popleft()
 
         if new_entry and self.display:
             new_entry.display()
@@ -167,6 +167,7 @@ async def aprocess_step(self, pre_step: bool = False):
 
         if should_consolidate:
             await self._aupdate_long_term_memory()
+            self.short_term_memory.popleft()
 
         if new_entry and self.display:
             new_entry.display()

tests/test_memory/test_STLT_memory.py

@@ -200,3 +200,50 @@ def test_get_prompt_ready_returns_str_when_empty(self, mock_agent):
         )
         assert "Short term memory:" in result
         assert "Long term memory:" in result
+
+    def test_consolidation_includes_evicted_entry(
+        self, mock_agent, mock_llm, llm_response_factory
+    ):
+        """Test that the entry being evicted from short-term memory is included
+        in the consolidation prompt so its content is preserved in long-term
+        memory.
+
+        Before the fix, popleft() removed the oldest entry *before*
+        consolidation ran, so the evicted entry was permanently lost.
+        """
+        mock_llm.generate.return_value = llm_response_factory("Consolidated summary")
+
+        memory = STLTMemory(
+            agent=mock_agent,
+            short_term_capacity=2,
+            consolidation_capacity=1,
+            llm_model="provider/test_model",
+        )
+        memory.llm = mock_llm
+
+        with patch("rich.console.Console"):
+            # Fill memory to just below consolidation threshold
+            for i in range(3):
+                memory.add_to_memory("observation", {"content": f"event_{i}"})
+                memory.process_step(pre_step=True)
+                memory.process_step(pre_step=False)
+
+            # At this point, capacity(2) + consolidation_capacity(1) = 3
+            # We have 3 entries, so no consolidation yet.
+            assert not mock_llm.generate.called
+
+            # Add one more entry to trigger consolidation
+            memory.add_to_memory("observation", {"content": "event_3"})
+            memory.process_step(pre_step=True)
+            memory.process_step(pre_step=False)
+
+            # Consolidation should have been triggered
+            assert mock_llm.generate.called
+
+            # The consolidation prompt must contain the oldest entry
+            # (event_0) that is about to be evicted
+            prompt_sent = mock_llm.generate.call_args[0][0]
+            assert "event_0" in prompt_sent, (
+                "The evicted entry's content must appear in the consolidation "
+                "prompt so the LLM can preserve it in long-term memory"
+            )