fix: ignore middleware prompts in run journal input (#3780)

backend/packages/harness/deerflow/runtime/journal.py

@@ -197,12 +197,12 @@ class RunJournal(BaseCallbackHandler):
             [len(batch) for batch in messages],
         )
 
-        # Capture the first human message sent to any LLM in this run.
+        # Capture the first user message sent to the lead agent in this run.
-        if not self._first_human_msg and messages:
+        caller = self._identify_caller(tags)
+        if caller == "lead_agent" and not self._first_human_msg and messages:
             for batch in reversed(messages):
                 for m in reversed(batch):
                     if isinstance(m, HumanMessage) and m.name != "summary" and m.additional_kwargs.get("hide_from_ui") is not True:
-                        caller = self._identify_caller(tags)
                         self.set_first_human_message(m.text)
                         self._put(
                             event_type="llm.human.input",

backend/tests/test_run_journal.py

@@ -924,6 +924,57 @@ class TestChatModelStartHumanMessage:
         assert len(human_events) == 1
         assert human_events[0]["content"]["content"] == "Real question"
 
+    @pytest.mark.anyio
+    async def test_summarization_prompt_does_not_capture_first_human_message(self, journal_setup):
+        """Internal summarization prompts must not replace the run's real user input."""
+        from langchain_core.messages import HumanMessage
+
+        j, store = journal_setup
+        summarization_prompt = HumanMessage(
+            content="<role>\nContext Extraction Assistant\n</role>\n\n<primary_objective>\nExtract context...",
+        )
+        j.on_chat_model_start(
+            {},
+            [[summarization_prompt]],
+            run_id=uuid4(),
+            tags=["middleware:summarize"],
+        )
+        j.on_chat_model_start(
+            {},
+            [[HumanMessage(content="Real user follow-up")]],
+            run_id=uuid4(),
+            tags=["lead_agent"],
+        )
+        await j.flush()
+
+        assert j._first_human_msg == "Real user follow-up"
+        assert j.get_completion_data()["message_count"] == 1
+        events = await store.list_events("t1", "r1")
+        human_events = [e for e in events if e["event_type"] == "llm.human.input"]
+        assert len(human_events) == 1
+        assert human_events[0]["content"]["content"] == "Real user follow-up"
+        assert human_events[0]["metadata"]["caller"] == "lead_agent"
+
+    @pytest.mark.anyio
+    @pytest.mark.parametrize("tags", [["middleware:summarize"], ["subagent:research"]])
+    async def test_non_lead_human_prompts_are_not_captured_as_user_input(self, journal_setup, tags):
+        """Only lead-agent LLM starts create UI-facing human input events."""
+        from langchain_core.messages import HumanMessage
+
+        j, store = journal_setup
+        j.on_chat_model_start(
+            {},
+            [[HumanMessage(content="Internal prompt")]],
+            run_id=uuid4(),
+            tags=tags,
+        )
+        await j.flush()
+
+        assert j._first_human_msg is None
+        assert j.get_completion_data()["message_count"] == 0
+        events = await store.list_events("t1", "r1")
+        assert not any(e["event_type"] == "llm.human.input" for e in events)
+
     @pytest.mark.anyio
     async def test_only_first_human_message_captured(self, journal_setup):
         """Subsequent on_chat_model_start calls do not overwrite the first message."""