feat: release 2.4.0 with original cases and automatic human review

tests/unit/test_token_meter.py

@@ -126,6 +126,14 @@ def test_record_llm_usage_accumulates_across_calls_same_session():
     assert meter.last_prompt_tokens == 250
 
 
+def test_record_llm_usage_preserves_the_call_model():
+    engine = Mock()
+    engine.current_session_id = uuid4()
+    engine.current_step_id = None
+    payload = record_llm_usage(engine, _msg_with_usage(), model="gpt-6.1-terra")
+    assert payload["model"] == "gpt-6.1-terra"
+
+
 def test_record_llm_usage_never_raises_and_skips_gracefully():
     # No engine
     assert record_llm_usage(None, _msg_with_usage()) is None