chore: Update agent tool descriptions and fix missing variable in chart tool

src/agent/rag_agent.py

@@ -17,12 +17,12 @@ from src.config.consts import (
 class IssuanceArgs(BaseModel):
     question: str = Field(description="对话问题")
     history: str = Field(description="历史对话记录")
-    location: list = Field(description="行政区名称")
+    location: list = Field(description="行政区划名称")
 
 
 class RAGQuery(BaseTool):
     name = "rag_query"
-    description = """这是一个区（县）级的水文气象地质知识库，当咨询区（县）的水文气象地质等相关信息的时候，你可以提供数据和资料。注意，这个查询只能获取单个区（县）的水文气象地质等相关信息，当需要查询省市的详细信息时，需要获取改省市下的具体行政规划，并获一一取具体的区（县）的水文气象地质等相关信息"""
+    description = """这是一个区（县）级的水文气象地质知识库，当咨询区（县）的水文气象地质等相关信息的时候，你可以提供数据和资料。注意，这个查询只能获取单个区（县）的水文气象地质等相关信息，当需要查询省市的详细信息时，需要获取改省市下的具体行政规划，并一一获取具体的区（县）的水文气象地质等相关信息。这个知识库中信息并不全面，有可能缺失。这个工具生成的结果要求保存到 action_cache 中。键值为 'rag_query'，值为工具输出。形式为 {"rag_query":$RESULT}"""
     args_schema: Type[BaseModel] = IssuanceArgs
     rerank: Any  # 替换 Any 为适当的类型
     rerank_model: Any  # 替换 Any 为适当的类型
@@ -52,9 +52,9 @@ class RAGQuery(BaseTool):
         # for l in split_str:
         #     split_list.append(l)
         answer = self.db.find_like_doc(location)
-        print(len(answer))
+        print(len(answer) if answer else 0)
         split_docs = []
-        for a in answer:
+        for a in answer if answer else []:
             d = Document(page_content=a[0], metadata=json.loads(a[1]))
             split_docs.append(d)
         print(len(split_docs))
@@ -67,12 +67,14 @@ class RAGQuery(BaseTool):
         print(matches)
         similarity = self.get_similarity_with_ext_origin(matches)
         # cur_similarity = similarity.get_rerank(self.rerank_model)
-        cur_similarity = similarity.get_rerank_with_doc(self.rerank_model,split_docs)
-        print(cur_similarity)
-        # geo_result = "以下是详细的水文气象地质资料："+cur_similarity+"\n 以下是原问题"+question
-        # cur_question = self.prompt.format(history=history, context=cur_similarity, question=question)
-        cur_answer = self.llm_chain.run(context=cur_similarity, question=question)
-        print(cur_answer)
-        return cur_answer
+        # cur_similarity = similarity.get_rerank_with_doc(self.rerank_model,split_docs)
+        docs = similarity.get_rerank_docs()
+        # print(cur_similarity)
+        # # geo_result = "以下是详细的水文气象地质资料："+cur_similarity+"\n 以下是原问题"+question
+        # # cur_question = self.prompt.format(history=history, context=cur_similarity, question=question)
+        # cur_answer = self.llm_chain.run(context=cur_similarity, question=question)
+        # print(cur_answer)
+        # return cur_answer
+        return {"参考资料": docs, "原问题": question}
 
 

src/agent/tool_chart.py

@@ -3,7 +3,7 @@ from pydantic import BaseModel, Field
 from langchain_core.tools import BaseTool
 
 class ChartArgs(BaseModel):
-    name: str = Field(..., description="图表名称，用于显示在图表上方")
+    title: str = Field(..., description="图表名称，用于显示在图表上方的字符串")
     chart_type: str = Field(..., description="图表类型，如 line, bar, scatter, pie 等")
     x: list = Field(..., description="x 轴数据，列表形式")
     y: list = Field(..., description="y 轴数据，列表形式")
@@ -12,7 +12,7 @@ class ChartArgs(BaseModel):
 
 class Chart(BaseTool):
     name = "chart"
-    description = "组装生成图表的中间数据"
+    description = "组装生成图表的中间数据。此工具生成的数据需要保存到 action_cache 中。键值为 'chart_data'。"
     args_schema: Type[BaseModel] = ChartArgs
 
     def _run(
@@ -37,7 +37,7 @@ def chart_image(chart_data):
     Args:
         chart_data: dict 图表数据 
             {
-                "name": str, 图表名称
+                "title": str, 图表名称
                 "chart_type": str, 图表类型，如 line, bar, scatter, pie 等
                 "x": list, x 轴数据，列表形式
                 "y": list, y 轴数据，列表形式

src/config/prompts.py

@@ -44,7 +44,7 @@ PROMPT_AGENT_CHART_SYS = """请尽量帮助人类并准确回答问题。您可
 
 {tools}
 
-使用 JSON 对象指定工具，提供一个 action 键（工具名称）和一个 action_input 键（工具输入）, 以及 action_cache 键（必要时存储工具中间结果） 。
+使用 JSON 对象指定工具，提供一个 action 键（工具名称）和一个 action_input 键（工具输入）, 以及 action_cache 键（有些工具要求缓存其返回的中间结果） 。
 
 有效的 "action" 值： "Final Answer" 或 {tool_names}
 
@@ -53,7 +53,6 @@ PROMPT_AGENT_CHART_SYS = """请尽量帮助人类并准确回答问题。您可
     {{
     "action": $TOOL_NAME,
     "action_input": $INPUT,
-    "action_cache": $CACHE
     }}
     ```
 
@@ -73,17 +72,29 @@ Action:
     {{
     "action": "Final Answer",
     "action_input": "最终回复给人类",
-    "action_cache": "图表工具结果缓存"
+    "action_cache": "所有要求保存中间结果的工具操作结果汇总"
     }}
     ```
-开始！始终以有效的单个操作的 JSON 对象回复。如有必要，请使用工具，如果你知道答案，请直接回复。如果用户有生成图表的需求，请使用图表生成工具 {chart_tool}，并将结果记录为 $CACHE 中。
-你的回复格式为 Action:```$JSON_BLOB```然后 Observation
-
+开始！始终以有效的单个操作的 JSON 对象回复。如有必要，请使用工具，并在最后一步按照工具要求将工具操作结果汇总到最后一个 Action 中的 action_cache。如果你知道答案，请直接回复。
+你的回复格式为 Action:```$JSON_BLOB```然后 Observation，并在必要时将 Observation 结果更新到下一个 action_cache 中。
 
 """
 
+PROMPT_AGENT_CHART_SYS_VARS = [ "tool_names", "tools"]
+
 PROMPT_AGENT_HUMAN = """{input}\n\n{agent_scratchpad}\n (请注意，无论如何都要以 JSON 对象回复)"""
 
+PROMPT_AGENT_CHAT_HUMAN = """历史纪录:
+'''
+{histories}
+'''
+问题：{input}
+
+{agent_scratchpad}
+(请注意，无论如何都要以 JSON 对象回复)"""
+
+PROMPT_AGENT_CHAT_HUMAN_VARS = ["agent_scratchpad", "input", "histories"]
+
 ##################################################################################################################
 
 # 结合历史对话信息，对用户提问进行扩展，生成不同角度的多个提问。用于 RAG 场景中

src/server/agent.py

@@ -81,7 +81,7 @@ def create_chart_agent(
 ) -> Runnable:
     """Create an agent aimed at supporting chart tools with multiple inputs.
     """
-    missing_vars = {"tools", "tool_names", "agent_scratchpad","chart_tool"}.difference(
+    missing_vars = {"tools", "tool_names", "agent_scratchpad"}.difference(
         prompt.input_variables + list(prompt.partial_variables)
     )
     if missing_vars:

test/rag_agent_test.py

+import sys
+sys.path.append('../')
+
 from typing import List,Union
+
 from langchain_openai import ChatOpenAI
 from langchain.agents import AgentExecutor,create_structured_chat_agent
 from langchain_core.prompts.chat import ChatPromptTemplate,HumanMessagePromptTemplate,SystemMessagePromptTemplate,MessagesPlaceholder
 from langchain_core.prompts import PromptTemplate
+from langchain.chains import LLMChain
 import langchain_core
+
 from src.pgdb.knowledge.similarity import VectorStore_FAISS
 from src.server.get_similarity import QAExt
+from src.server.agent import create_chart_agent
 from src.pgdb.knowledge.k_db import PostgresDB
 from src.pgdb.knowledge.txt_doc_table import TxtDoc
-from langchain.chains import LLMChain
 from src.agent.tool_divisions import AdministrativeDivision
 from src.agent.rag_agent import RAGQuery
 from src.config.consts import (
@@ -24,6 +30,9 @@ from src.config.consts import (
     prompt_enhancement_history_template,
     prompt1
 )
+
+from src.config.prompts import PROMPT_AGENT_CHART_SYS_VARS,PROMPT_AGENT_CHART_SYS,PROMPT_AGENT_CHAT_HUMAN,PROMPT_AGENT_CHAT_HUMAN_VARS
+
 base_llm = ChatOpenAI(
     openai_api_key='xxxxxxxxxxxxx',
     openai_api_base='http://192.168.10.14:8000/v1',
@@ -48,25 +57,31 @@ k_db.connect()
 
 llm_chain = LLMChain(llm=base_llm, prompt=PromptTemplate(input_variables=["history","context", "question"], template=prompt1), llm_kwargs= {"temperature": 0})
 
+tool_rag = RAGQuery(vecstore_faiss,ext,PromptTemplate(input_variables=["history","context", "question"], template=prompt_enhancement_history_template),_db=TxtDoc(k_db),_llm_chain=llm_chain)
 tools = [AdministrativeDivision(),RAGQuery(vecstore_faiss,ext,PromptTemplate(input_variables=["history","context", "question"], template=prompt_enhancement_history_template),_db=TxtDoc(k_db),_llm_chain=llm_chain)]
-input_variables=['agent_scratchpad', 'input', 'tool_names', 'tools']
+# input_variables=['agent_scratchpad', 'input', 'tool_names', 'tools','chart_tool']
+input_variables=[]
+input_variables.extend(PROMPT_AGENT_CHAT_HUMAN_VARS)
+input_variables.extend(PROMPT_AGENT_CHART_SYS_VARS)
 input_types={'chat_history': List[Union[langchain_core.messages.ai.AIMessage, langchain_core.messages.human.HumanMessage, langchain_core.messages.chat.ChatMessage, langchain_core.messages.system.SystemMessage, langchain_core.messages.function.FunctionMessage, langchain_core.messages.tool.ToolMessage]]}
-metadata={'lc_hub_owner': 'hwchase17', 'lc_hub_repo': 'structured-chat-agent', 'lc_hub_commit_hash': 'ea510f70a5872eb0f41a4e3b7bb004d5711dc127adee08329c664c6c8be5f13c'}
 messages=[
-  SystemMessagePromptTemplate(prompt=PromptTemplate(input_variables=['tool_names', 'tools'], template='Respond to the human as helpfully and accurately as possible. You have access to the following tools:\n\n{tools}\n\nUse a json blob to specify a tool by providing an action key (tool name) and an action_input key (tool input).\n\nValid "action" values: "Final Answer" or {tool_names}\n\nProvide only ONE action per $JSON_BLOB, as shown:\n\n```\n{{\n  "action": $TOOL_NAME,\n  "action_input": $INPUT\n}}\n```\n\nFollow this format:\n\nQuestion: input question to answer\nThought: consider previous and subsequent steps\nAction:\n```\n$JSON_BLOB\n```\nObservation: action result\n... (repeat Thought/Action/Observation N times)\nThought: I know what to respond\nAction:\n```\n{{\n  "action": "Final Answer",\n  "action_input": "Final response to human"\n}}\n\nBegin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation')),
+#   SystemMessagePromptTemplate(prompt=PromptTemplate(input_variables=['tool_names', 'tools'], template='Respond to the human as helpfully and accurately as possible. You have access to the following tools:\n\n{tools}\n\nUse a json blob to specify a tool by providing an action key (tool name) and an action_input key (tool input).\n\nValid "action" values: "Final Answer" or {tool_names}\n\nProvide only ONE action per $JSON_BLOB, as shown:\n\n```\n{{\n  "action": $TOOL_NAME,\n  "action_input": $INPUT\n}}\n```\n\nFollow this format:\n\nQuestion: input question to answer\nThought: consider previous and subsequent steps\nAction:\n```\n$JSON_BLOB\n```\nObservation: action result\n... (repeat Thought/Action/Observation N times)\nThought: I know what to respond\nAction:\n```\n{{\n  "action": "Final Answer",\n  "action_input": "Final response to human"\n}}\n\nBegin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation')),
+  SystemMessagePromptTemplate(prompt=PromptTemplate(input_variables=PROMPT_AGENT_CHART_SYS_VARS, template=PROMPT_AGENT_CHART_SYS)),
   MessagesPlaceholder(variable_name='chat_history', optional=True),
-  HumanMessagePromptTemplate(prompt=PromptTemplate(input_variables=['agent_scratchpad', 'input'], template='{input}\n\n{agent_scratchpad}\n (reminder to respond in a JSON blob no matter what)'))
+#   HumanMessagePromptTemplate(prompt=PromptTemplate(input_variables=['agent_scratchpad', 'input'], template='{input}\n\n{agent_scratchpad}\n (reminder to respond in a JSON blob no matter what)'))
+  HumanMessagePromptTemplate(prompt=PromptTemplate(input_variables=PROMPT_AGENT_CHAT_HUMAN_VARS, template=PROMPT_AGENT_CHAT_HUMAN))
 ]
 
 prompt = ChatPromptTemplate(
     input_variables=input_variables,
     input_types=input_types,
-    metadata=metadata,
+    # metadata=metadata,
     messages=messages
 )
 
 
-agent = create_structured_chat_agent(llm=base_llm, tools=tools, prompt=prompt)
+# agent = create_structured_chat_agent(llm=base_llm, tools=tools, prompt=prompt)
+agent = create_chart_agent(base_llm, tools, prompt, chart_tool="chart")
 agent_executor = AgentExecutor(agent=agent, tools=tools,verbose=True,handle_parsing_errors=True)
 history = []
 h1 = []
@@ -81,7 +96,8 @@ prompt = ""
 for h in history:
     prompt += "问：{}\n答：{}\n".format(h[0], h[1])
 print(prompt)
-res = agent_executor.invoke({"input":"以下历史对话记录： "+prompt+"以下是问题："+"攸县、长沙县、化隆县和大通县谁的年平均降雨量大"})
+# res = agent_executor.invoke({"input":"以下历史对话记录： "+prompt+"以下是问题："+"攸县、长沙县、化隆县和大通县谁的年平均降雨量大"})
+res = agent_executor.invoke({"input":"攸县、长沙县、化隆县和大通县谁的年平均降雨量大","histories":history})
 
 print("====== result: ======")
 print(res)