Feature/multi turn

src/agents/sql_agent_graph.py

@@ -91,6 +91,7 @@ def db_classifier_node(state: SqlAgentState):
     chain = DB_CLASSIFIER_PROMPT | llm_instance | StrOutputParser()
     selected_db_name = chain.invoke({
         "db_options": db_options,
+        "chat_history": state['chat_history'],
         "question": state['question']
     })
 
@@ -199,6 +200,7 @@ def response_synthesizer_node(state: SqlAgentState):
 
     prompt = RESPONSE_SYNTHESIZER_PROMPT.format(
         question=state['question'],
+        chat_history=state['chat_history'],
         context_message=context_message
     )
     response = llm_instance.invoke(prompt)

src/api/v1/endpoints/chat.py

@@ -23,5 +23,6 @@ def handle_chat_request(
     Returns:
         ChatRespone: 챗봇 응답
     """
-    final_answer = service.handle_request(request.question)
+    final_answer = service.handle_request(request.question, request.chat_history)
+
     return ChatResponse(answer=final_answer)

src/api/v1/schemas/chatbot_schemas.py

@@ -1,9 +1,16 @@
 # src/api/v1/schemas/chatbot_schemas.py
 
 from pydantic import BaseModel
+from typing import List, Optional
+
+class ChatMessage(BaseModel):
+    """대화 기록의 단일 메시지를 나타내는 모델"""
+    role: str  # "user" 또는 "assistant"
+    content: str
 
 class ChatRequest(BaseModel):
     question: str
+    chat_history: Optional[List[ChatMessage]] = None
 
 class ChatResponse(BaseModel):
-    answer: str
+    answer: str

src/prompts/v1/sql_agent/db_classifier.yaml

@@ -1,13 +1,17 @@
 _type: "prompt"
 input_variables:
   - db_options
+  - chat_history
   - question
 template: |
   Based on the user's question, which of the following databases is most likely to contain the answer?
   Please respond with only the database name.
 
   Available databases:
   {db_options}
+
+  conversation History:
+  {chat_history}
 
   User Question:
   {question}

src/prompts/v1/sql_agent/response_synthesizer.yaml

@@ -1,6 +1,7 @@
 _type: prompt
 input_variables:
   - question
+  - chat_history
   - context_message
 template: |
   You are a friendly and helpful database assistant chatbot.
@@ -11,6 +12,9 @@ template: |
 
   Context:
   {context_message}
+
+  conversation History:
+  {chat_history}
 
   Instructions:
   - If the process was successful:

src/services/chatbot_service.py

@@ -1,24 +1,34 @@
 # src/services/chatbot_service.py
 
 from agents.sql_agent_graph import sql_agent_app
-from core.db_manager import schema_instance
+from api.v1.schemas.chatbot_schemas import ChatMessage # --- 추가된 부분 ---
+from langchain_core.messages import HumanMessage, AIMessage, BaseMessage # --- 추가된 부분 ---
+from typing import List, Optional # --- 추가된 부분 ---
+#from core.db_manager import schema_instance
 
 class ChatbotService():
-    def __init__(self):
-        self.db_schema = schema_instance
+    # TODO: schema API 요청
+    # def __init__(self):
+    #     self.db_schema = schema_instance
 
-    def handle_request(self, user_question: str) -> str:
-        # 1. 에이전트 그래프에 전달할 초기 상태 구성
+    def handle_request(self, user_question: str, chat_history: Optional[List[ChatMessage]] = None) -> dict:
+
+        langchain_messages: List[BaseMessage] = []
+        if chat_history:
+            for message in chat_history:
+                if message.role == 'user':
+                    langchain_messages.append(HumanMessage(content=message.content))
+                elif message.role == 'assistant':
+                    langchain_messages.append(AIMessage(content=message.content))
+
         initial_state = {
             "question": user_question,
-            "chat_history": [],
-            "db_schema": self.db_schema,
+            "chat_history": langchain_messages,
+            # "db_schema": self.db_schema,
             "validation_error_count": 0,
             "execution_error_count": 0
         }
 
-        # 2. 그래프 실행
         final_state = sql_agent_app.invoke(initial_state)
-        final_response = final_state['final_response']
 
-        return final_response
+        return final_state['final_response']