fix(researcher): Handle non-string search results in summarizer node

backend/src/ea_chatbot/graph/workers/researcher/nodes/summarizer.py

@@ -17,7 +17,15 @@ def summarizer_node(state: WorkerState) -> dict:
         callbacks=[LangChainLoggingHandler(logger=logger)]
     )
     
-    results_str = "\n---\n".join(raw_results)
+    # Ensure all results are strings (Gemini/OpenAI might return complex content)
+    processed_results = []
+    for res in raw_results:
+        if isinstance(res, list):
+            processed_results.append(str(res))
+        else:
+            processed_results.append(str(res))
+
+    results_str = "\n---\n".join(processed_results)
     
     prompt = f"""You are a Research Specialist sub-agent. You have completed a research sub-task.
 Task: {task}

backend/tests/test_orchestrator_loop.py

@@ -84,3 +84,53 @@ def test_orchestrator_full_flow():
     assert mock_reflector.called
     assert mock_synthesizer.called
     assert "Final synthesized answer" in [m.content for m in final_state["messages"]]
+
+def test_orchestrator_researcher_flow():
+    """Verify that the Orchestrator can route to the researcher worker."""
+    
+    mock_analyzer = MagicMock()
+    mock_planner = MagicMock()
+    mock_delegate = MagicMock()
+    mock_researcher = MagicMock()
+    mock_reflector = MagicMock()
+    mock_synthesizer = MagicMock()
+    
+    mock_analyzer.return_value = {"next_action": "plan"}
+    mock_planner.return_value = {
+        "checklist": [{"task": "Search news", "worker": "researcher"}],
+        "current_step": 0
+    }
+    mock_delegate.side_effect = [
+        {"next_action": "researcher"},
+        {"next_action": "summarize"}
+    ]
+    mock_researcher.return_value = {"messages": [AIMessage(content="News found")]}
+    mock_reflector.return_value = {"current_step": 1, "next_action": "delegate"}
+    mock_synthesizer.return_value = {"messages": [AIMessage(content="Final News Summary")], "next_action": "end"}
+    
+    app = create_workflow(
+        query_analyzer=mock_analyzer,
+        planner=mock_planner,
+        delegate=mock_delegate,
+        researcher_worker=mock_researcher,
+        reflector=mock_reflector,
+        synthesizer=mock_synthesizer
+    )
+    
+    initial_state = AgentState(
+        messages=[HumanMessage(content="What's the news?")],
+        question="What's the news?",
+        analysis={},
+        next_action="",
+        iterations=0,
+        checklist=[],
+        current_step=0,
+        vfs={},
+        plots=[],
+        dfs={}
+    )
+    
+    final_state = app.invoke(initial_state)
+    
+    assert mock_researcher.called
+    assert "Final News Summary" in [m.content for m in final_state["messages"]]