github/DB-GPT
1
0
Fork 0
mirror of https://github.com/csunny/DB-GPT.git synced 2025-08-10 04:37:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat:summarize

Browse Source
This commit is contained in:
aries_ckt 2025-07-11 22:00:56 +08:00
parent a54eefefa2
commit 74850cbe09
2 changed files with 200 additions and 132 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

111
packages/dbgpt-core/src/dbgpt/agent/expand/actions/deep_search_action.py
View File

@ -13,12 +13,12 @@ from dbgpt._private.pydantic import BaseModel, Field, model_to_dict
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
class DeepSearchModel(BaseModel): class SplitQueryModel(BaseModel):
"""Chart item model.""" """Model for splitting queries in deep search actions."""
status: str = Field( # status: str = Field(
..., # ...,
description="The status of the current action, can be split_query, summary, or reflection.", # description="The status of the current action, can be split_query, summary, or reflection.",
) # )
tools: List[dict] = Field( tools: List[dict] = Field(
default_factory=list, default_factory=list,
description="List of tools to be used in the action.", description="List of tools to be used in the action.",
@ -37,11 +37,40 @@ class DeepSearchModel(BaseModel):
return model_to_dict(self) return model_to_dict(self)
class ReflectionModel(BaseModel):
"""Model for Reflection."""
status: str = Field(
default_factory=list,
description="List of tools to be used in the action.",
)
knowledge_gap: str = Field(
...,
description="The intention of the current action, describing what you want to achieve.",
)
sub_queries: List[str] = Field(
default_factory=list,
description="List of sub-queries generated from the current action.",
)
tools: List[dict] = Field(
default_factory=list,
description="List of tools to be used in the action.",
)
thought: str = Field(
...,
description="The thought of the current action, describing what you want to achieve.",
)
def to_dict(self):
"""Convert to dict."""
return model_to_dict(self)
class DeepSearchAction(ToolAction): class DeepSearchAction(ToolAction):
"""React action class.""" """React action class."""
def __init__(self, **kwargs): def __init__(self, **kwargs):
"""Tool action init.""" """Tool action init."""
self.state = "split_query"
super().__init__(**kwargs) super().__init__(**kwargs)
@property @property
@ -73,8 +102,9 @@ class DeepSearchAction(ToolAction):
) -> ActionOutput: ) -> ActionOutput:
"""Perform the action.""" """Perform the action."""
try: try:
action_param: DeepSearchModel = self._input_convert( # state = "split_query"
ai_message, DeepSearchModel action_param: ReflectionModel = self._input_convert(
ai_message, ReflectionModel
) )
except Exception as e: except Exception as e:
logger.exception(str(e)) logger.exception(str(e))
@ -83,29 +113,34 @@ class DeepSearchAction(ToolAction):
content="The requested correctly structured answer could not be found.", content="The requested correctly structured answer could not be found.",
) )
if action_param.status == "split_query": sub_queries = action_param.sub_queries
sub_queries = action_param.sub_queries if action_param.status == "summarize":
# execute knowledge search return ActionOutput(
if not action_param.tools: is_exe_success=True,
return ActionOutput( content=action_param.thought,
is_exe_success=False, terminate=True,
content="No tools available for knowledge search.", )
) if not action_param.tools:
if action_param.tools: return ActionOutput(
for tool in action_param.tools: is_exe_success=False,
if tool.get("tool_type") == "KnowledgeRetrieve": content="No tools available for knowledge search.",
knowledge_args = action_param.get("args", {}) )
if not knowledge_args: if action_param.tools:
return ActionOutput( for tool in action_param.tools:
is_exe_success=False, # state = "knowledge_search"
content="No arguments provided for knowledge search.", if tool.get("tool_type") == "KnowledgeRetrieve":
) knowledge_args = tool.get("args", {})
act_out = await self.knowledge_retrieve( if not knowledge_args:
sub_queries, return ActionOutput(
knowledge_args, is_exe_success=False,
self.resource, content="No arguments provided for knowledge search.",
) )
act_out = await self.knowledge_retrieve(
sub_queries,
knowledge_args,
self.resource,
)
act_out.terminate = False
# if "parser" in kwargs and isinstance(kwargs["parser"], ReActOutputParser): # if "parser" in kwargs and isinstance(kwargs["parser"], ReActOutputParser):
# parser = kwargs["parser"] # parser = kwargs["parser"]
@ -138,21 +173,17 @@ class DeepSearchAction(ToolAction):
lang=self.language, question=query lang=self.language, question=query
) )
query_context_map[query] = resource_prompt query_context_map[query] = resource_prompt
content = "\n".join([
f"{query}:{context}" for query, context in query_context_map.items()]
)
action_output = ActionOutput( action_output = ActionOutput(
is_exe_success=True, is_exe_success=True,
content="\n".join([ content=content,
f"{query}:{context}" for query, context in query_context_map.items()] view=content,
), observations=content,
view="\n".join([
f"{query}:{context}" for query, context in query_context_map.items()]
),
observations=query_context_map,
) )
return action_output return action_output
async def _do_run( async def _do_run(
self, self,
ai_message: str, ai_message: str,

221
packages/dbgpt-core/src/dbgpt/agent/expand/deep_search.py
View File

@ -74,29 +74,36 @@ selecting the right search tools.
# The current time is: {{ now_time }}. # The current time is: {{ now_time }}.
# """ # """
_DEEPSEARCH_SYSTEM_TEMPLATE = """ _DEEPSEARCH_SYSTEM_TEMPLATE = """
你是一个深度搜索助手。你的任务是你将用户原始问题一个或者多个子问题，并且给出可用知识库工具和搜索工具来回答问题或解决问题。 你是一个深度搜索助手。
<目标>
你的任务是根据用户的问题或任务，选择合适的知识检索工具和搜索工具来回答问题或解决问题。
你需要根据已经搜到的知识和搜索到的信息:
{{most_recent_memories}}判断是否需要更多的知识或信息来回答问题。
如果需要更多的知识或信息，你需要提出后续的子问题来扩展你的理解。
</目标>
<可用工具> <可用工具>
1. KnowledgeRetrieve: 查询内部知识库以获取信息。\n可用知识库: {{knowledge_tools}} 1. KnowledgeRetrieve: 查询内部知识库以获取信息。\n可用知识库: {{knowledge_tools}}
2. WebSearch: 进行互联网搜索以获取最新或额外信息。\n 可用搜索工具: {{search_tools}} 2. WebSearch: 进行互联网搜索以获取最新或额外信息。\n 可用搜索工具: {{search_tools}}
3. 总结: 对多个来源的信息进行总结和综合。
</可用工具> </可用工具>
<流程> <流程>
1. 分析任务并创建搜索计划。 1. 分析任务并创建搜索计划。
2. 选择使用一个或多个工具收集信息。 2. 选择使用一个或多个工具收集信息。
3. 对收集到的信息进行反思，判断是否足够回答问题。
</流程> </流程>
<回复格式> <回复格式>
严格按以下JSON格式输出，确保可直接解析： 严格按以下JSON格式输出，确保可直接解析：
{ {
"status": "split_query (拆解搜索计划) | summary (仅当可用知识可以回答用户问题) | reflection (反思) "
"tools": [{ "tools": [{
"tool_type": "工具类型" "tool_type": "工具类型"
"args": "args1", "args": "args1",
}], }],
"intention": "当前你的意图, "intention": "当前你的意图,
"sub_queries": [], "sub_queries": ["子问题1", "子问题2"],
"knowledge_gap": "总结缺乏关于性能指标和基准的信息",
"status": "reflection(反思) | summarize(最后总结)",
} }
</回复格式> </回复格式>
@ -116,13 +123,14 @@ _DEEPSEARCH_SYSTEM_TEMPLATE = """
: 2022年诺贝尔文学奖得主 : 2022年诺贝尔文学奖得主
返回 返回
{ {
"status": "split_query" "status": "reflection"
"tools"?: [{ "tools"?: [{
"tool_type": "KnowledgeRetrieve" "tool_type": "KnowledgeRetrieve"
"args": "knowledge_name", "args": "knowledge_name",
}], }],
"intention": "你的拆解意图, "intention": "你的拆解意图,
"sub_queries": [], "knowledge_gap": "总结缺乏关于2022年诺贝尔文学奖得主的信息",
"sub_queries": ["子问题1","子问题2"],
} }
</示例> </示例>
@ -135,6 +143,36 @@ _DEEPSEARCH_SYSTEM_TEMPLATE = """
""" """
_DEEPSEARCH_USER_TEMPLATE = """""" _DEEPSEARCH_USER_TEMPLATE = """"""
_DEEPSEARCH_FINIAL_SUMMARY_TEMPLATE = """
<GOAL>
Generate a high-quality summary of the provided context.
</GOAL>
<REQUIREMENTS>
When creating a NEW summary:
1. Highlight the most relevant information related to the user topic from the search results
2. Ensure a coherent flow of information
When EXTENDING an existing summary:
{{most_recent_memories}}
1. Read the existing summary and new search results carefully.
2. Compare the new information with the existing summary.
3. For each piece of new information:
a. If it's related to existing points, integrate it into the relevant paragraph.
b. If it's entirely new but relevant, add a new paragraph with a smooth transition.
c. If it's not relevant to the user topic, skip it.
4. Ensure all additions are relevant to the user's topic.
5. Verify that your final output differs from the input summary.
< /REQUIREMENTS >
< FORMATTING >
- Start directly with the updated summary, without preamble or titles. Do not use XML tags in the output.
< /FORMATTING >
<Task>
Think carefully about the provided Context first. Then generate a summary of the context to address the User Input.
</Task>
"""
_REACT_WRITE_MEMORY_TEMPLATE = """\ _REACT_WRITE_MEMORY_TEMPLATE = """\
@ -176,42 +214,7 @@ class DeepSearchAgent(ConversableAgent):
"""Init indicator AssistantAgent.""" """Init indicator AssistantAgent."""
super().__init__(**kwargs) super().__init__(**kwargs)
self._init_actions([DeepSearchAction, Terminate]) self._init_actions([DeepSearchAction])
# async def _a_init_reply_message(
# self,
# received_message: AgentMessage,
# rely_messages: Optional[List[AgentMessage]] = None,
# ) -> AgentMessage:
# reply_message = super()._init_reply_message(received_message, rely_messages)
#
# tool_packs = ToolPack.from_resource(self.resource)
# action_space = []
# action_space_names = []
# action_space_simple_desc = []
# if tool_packs:
# tool_pack = tool_packs[0]
# for tool in tool_pack.sub_resources:
# tool_desc, _ = await tool.get_prompt(lang=self.language)
# action_space_names.append(tool.name)
# action_space.append(tool_desc)
# if isinstance(tool, BaseTool):
# tool_simple_desc = tool.description
# else:
# tool_simple_desc = tool.get_prompt()
# action_space_simple_desc.append(f"{tool.name}: {tool_simple_desc}")
# else:
# for action in self.actions:
# action_space_names.append(action.name)
# action_space.append(action.get_action_description())
# # self.actions
# reply_message.context = {
# "max_steps": self.max_retry_count,
# "action_space": "\n".join(action_space),
# "action_space_names": ", ".join(action_space_names),
# "action_space_simple_desc": "\n".join(action_space_simple_desc),
# }
# return reply_message
async def preload_resource(self) -> None: async def preload_resource(self) -> None:
await super().preload_resource() await super().preload_resource()
@ -289,14 +292,44 @@ class DeepSearchAgent(ConversableAgent):
"knowledge_desc": self.resource.retriever_desc, "knowledge_desc": self.resource.retriever_desc,
}) })
# new_resource = self.resource.apply(apply_func=_remove_tool)
# if new_resource:
# resource_prompt, resource_reference = await new_resource.get_prompt(
# lang=self.language, question=question
# )
# return resource_prompt, resource_reference
return json.dumps(abilities, ensure_ascii=False), [] return json.dumps(abilities, ensure_ascii=False), []
async def build_system_prompt(
self,
question: Optional[str] = None,
most_recent_memories: Optional[str] = None,
resource_vars: Optional[Dict] = None,
context: Optional[Dict[str, Any]] = None,
is_retry_chat: bool = False,
):
"""Build system prompt."""
system_prompt = None
if self.bind_prompt:
prompt_param = {}
if resource_vars:
prompt_param.update(resource_vars)
if context:
prompt_param.update(context)
if self.bind_prompt.template_format == "f-string":
system_prompt = self.bind_prompt.template.format(
**prompt_param,
)
elif self.bind_prompt.template_format == "jinja2":
system_prompt = Template(self.bind_prompt.template).render(prompt_param)
else:
logger.warning("Bind prompt template not exsit or format not support!")
if not system_prompt:
param: Dict = context if context else {}
system_prompt = await self.build_prompt(
question=question,
is_system=True,
most_recent_memories=most_recent_memories,
resource_vars=resource_vars,
is_retry_chat=is_retry_chat,
**param,
)
return system_prompt
def prepare_act_param( def prepare_act_param(
self, self,
received_message: Optional[AgentMessage], received_message: Optional[AgentMessage],
@ -350,6 +383,8 @@ class DeepSearchAgent(ConversableAgent):
rely_action_out=last_out, rely_action_out=last_out,
**kwargs, **kwargs,
) )
if not last_out.terminate:
self.profile.system_prompt_template = _DEEPSEARCH_FINIAL_SUMMARY_TEMPLATE
span.metadata["action_out"] = last_out.to_dict() if last_out else None span.metadata["action_out"] = last_out.to_dict() if last_out else None
if not last_out: if not last_out:
raise ValueError("Action should return value！") raise ValueError("Action should return value！")
@ -366,7 +401,7 @@ class DeepSearchAgent(ConversableAgent):
) -> Union[str, List["AgentMessage"]]: ) -> Union[str, List["AgentMessage"]]:
memories = await self.memory.read(observation) memories = await self.memory.read(observation)
not_json_memories = [] not_json_memories = []
messages = [] # messages = []
structured_memories = [] structured_memories = []
for m in memories: for m in memories:
if m.raw_observation: if m.raw_observation:
@ -381,46 +416,48 @@ class DeepSearchAgent(ConversableAgent):
except Exception: except Exception:
not_json_memories.append(m.raw_observation) not_json_memories.append(m.raw_observation)
for mem_dict in structured_memories: # for mem_dict in structured_memories:
question = mem_dict.get("question") # question = mem_dict.get("question")
thought = mem_dict.get("thought") # thought = mem_dict.get("thought")
action = mem_dict.get("action") # action = mem_dict.get("action")
action_input = mem_dict.get("action_input") # action_input = mem_dict.get("action_input")
observation = mem_dict.get("observation") # observation = mem_dict.get("observation")
if question: # if question:
messages.append( # messages.append(
AgentMessage( # AgentMessage(
content=f"Question: {question}", # content=f"Question: {question}",
role=ModelMessageRoleType.HUMAN, # role=ModelMessageRoleType.HUMAN,
) # )
) # )
ai_content = [] # ai_content = []
if thought: # if thought:
ai_content.append(f"Thought: {thought}") # ai_content.append(f"Thought: {thought}")
if action: # if action:
ai_content.append(f"Action: {action}") # ai_content.append(f"Action: {action}")
if action_input: # if action_input:
ai_content.append(f"Action Input: {action_input}") # ai_content.append(f"Action Input: {action_input}")
messages.append( # messages.append(
AgentMessage( # AgentMessage(
content="\n".join(ai_content), # content="\n".join(ai_content),
role=ModelMessageRoleType.AI, # role=ModelMessageRoleType.AI,
) # )
) # )
#
if observation: # if observation:
messages.append( # messages.append(
AgentMessage( # AgentMessage(
content=f"Observation: {observation}", # content=f"Observation: {observation}",
role=ModelMessageRoleType.HUMAN, # role=ModelMessageRoleType.HUMAN,
) # )
) # )
#
if not messages and not_json_memories: # if not messages and not_json_memories:
messages.append( # messages.append(
AgentMessage( # AgentMessage(
content="\n".join(not_json_memories), # content="\n".join(not_json_memories),
role=ModelMessageRoleType.HUMAN, # role=ModelMessageRoleType.HUMAN,
) # )
) # )
return messages return "\n".join([
mem_dict.get("observation") for mem_dict in structured_memories
])