Converted the evolve_question property of the state object into a List and created a function to return the last evolved question.

app.py

@@ -88,8 +88,8 @@ if st.button("Generate Synthetic Data"):
         # Display evolved questions
         st.markdown("### Evolved Questions")
         evolved_questions = [
-            {"id": f"q{i}", "question": q, "evolution_type": "simple"} 
-            for i, q in enumerate([result.evolved_question])  # Currently only one question
+            {"id": f"q{i}", "question": q, "evolution_type": "simple"}
+            for i, q in enumerate(result.evolved_questions)
         ]
         st.json(evolved_questions)
         

graph/nodes/answer.py

@@ -17,9 +17,10 @@ def generate_answer(state: SDGState) -> SDGState:
     new_state = SDGState(
         input=state.input,
         documents=state.documents,
-        evolved_question=state.evolved_question,
+        evolved_questions=state.evolved_questions,
         context=state.context,
-        answer=f"Based on the retrieved context:\n{context_snippet}"
+        answer=f"Based on the retrieved context:\n{context_snippet}",
+        num_evolve_passes=state.num_evolve_passes
     )
     
     logger.debug(f"Answer node returning state: {new_state}")

graph/nodes/evolve.py

@@ -10,17 +10,20 @@ def evolve_question(state: SDGState, llm) -> SDGState:
         "Rewrite or evolve the following question to be more challenging or insightful:\n\n{}",
         "Rewrite or evolve the following question to be more creative or original:\n\n{}"
     ]
-    evolved = state.input
+    evolved_questions = list(state.evolved_questions) if state.evolved_questions else [state.input]
+    evolved = evolved_questions[-1]
     for i in range(num_passes):
         prompt = prompts[i % len(prompts)].format(evolved)
         response = llm.invoke(prompt)
         evolved = response.content if hasattr(response, 'content') else str(response)
+        evolved_questions.append(evolved)
     new_state = SDGState(
         input=state.input,
         documents=state.documents,
-        evolved_question=evolved,
+        evolved_questions=evolved_questions,
         context=state.context,
-        answer=state.answer
+        answer=state.answer,
+        num_evolve_passes=state.num_evolve_passes
     )
     logger.debug(f"Evolve node returning state: {new_state}")
     return new_state

graph/nodes/retrieve.py

@@ -14,9 +14,10 @@ def retrieve_relevant_context(state: SDGState, vectorstore) -> SDGState:
     new_state = SDGState(
         input=state.input,
         documents=state.documents,
-        evolved_question=state.evolved_question,
+        evolved_questions=state.evolved_questions,
         context=[doc.page_content for doc in retrieved_docs],
-        answer=state.answer
+        answer=state.answer,
+        num_evolve_passes=state.num_evolve_passes
     )
     
     logger.debug(f"Retrieve node returning state: {new_state}")

graph/types.py

@@ -5,7 +5,11 @@ from pydantic import BaseModel, Field
 class SDGState(BaseModel):
     input: str = Field(default="")
     documents: List[Document] = Field(default_factory=list)
-    evolved_question: str = Field(default="")
+    evolved_questions: List[str] = Field(default_factory=list)
     context: List[str] = Field(default_factory=list)
     answer: str = Field(default="")
-    num_evolve_passes: int = Field(default=2)
+    num_evolve_passes: int = Field(default=2)
+
+    @property
+    def evolved_question(self):
+        return self.evolved_questions[-1] if self.evolved_questions else ""

main.py

@@ -20,6 +20,7 @@ class DocumentEncoder(json.JSONEncoder):
         if isinstance(obj, SDGState):
             return {
                 "input": obj.input,
+                "evolved_questions": obj.evolved_questions,
                 "evolved_question": obj.evolved_question,
                 "context": obj.context,
                 "answer": obj.answer

tests/graph/nodes/test_evolve.py

@@ -54,4 +54,40 @@ def test_evolve_question_three_passes():
         call("Rewrite or evolve the following question to be more creative or original:\n\nChallenging: What were the top LLMs in 2023?"),
         call("Rewrite or evolve the following question to be more challenging or insightful:\n\nCreative: What were the top LLMs in 2023?")
     ]
-    mock_llm.invoke.assert_has_calls(expected_calls)
+    mock_llm.invoke.assert_has_calls(expected_calls)
+
+def test_evolved_questions_list_populated_correctly():
+    state = SDGState(input="Base question", num_evolve_passes=3)
+    mock_llm = MagicMock()
+    mock_llm.invoke.side_effect = [
+        MagicMock(content="Challenging: Base question"),
+        MagicMock(content="Creative: Challenging: Base question"),
+        MagicMock(content="Challenging Again: Creative: Challenging: Base question")
+    ]
+    updated_state = evolve_question(state, mock_llm)
+    # The evolved_questions list should contain the initial input plus one entry per pass
+    assert updated_state.evolved_questions == [
+        "Base question",
+        "Challenging: Base question",
+        "Creative: Challenging: Base question",
+        "Challenging Again: Creative: Challenging: Base question"
+    ]
+    # The property should return the last one
+    assert updated_state.evolved_question == "Challenging Again: Creative: Challenging: Base question"
+
+def test_evolved_questions_list_with_existing_evolutions():
+    # If the state already has evolved_questions, it should continue from the last
+    state = SDGState(input="Base question", evolved_questions=["Base question", "First evolution"], num_evolve_passes=2)
+    mock_llm = MagicMock()
+    mock_llm.invoke.side_effect = [
+        MagicMock(content="Second evolution"),
+        MagicMock(content="Third evolution")
+    ]
+    updated_state = evolve_question(state, mock_llm)
+    assert updated_state.evolved_questions == [
+        "Base question",
+        "First evolution",
+        "Second evolution",
+        "Third evolution"
+    ]
+    assert updated_state.evolved_question == "Third evolution"

tests/graph/test_build_graph.py

@@ -17,6 +17,8 @@ def test_build_sdg_graph_runs():
     result = graph.invoke(state)
 
     assert isinstance(result, dict)
-    assert "evolved_question" in result
+    assert "evolved_questions" in result
+    if result["evolved_questions"]:
+        assert result["evolved_questions"][-1] == "Evolved test question"
     assert result["context"]
     assert "Relevant content" in result["context"][0]