4 Commits

Author SHA1 Message Date
ly 72ddf46fc7 Merge pull request '增加新的前端子模块' (#4) from dev into main
Reviewed-on: #4
2024-08-29 10:51:50 +08:00
ly f57c0c84ef Merge pull request 'dev' (#3) from dev into main
Reviewed-on: #3
2024-08-29 10:13:10 +08:00
ly 9ee24627c2 Merge pull request 'dev' (#2) from dev into main
Reviewed-on: #2
2024-08-23 09:37:06 +08:00
ly 88761a5d10 Merge pull request 'dev' (#1) from dev into main
Reviewed-on: #1
2024-08-22 09:41:13 +08:00
20 changed files with 295 additions and 2149 deletions
-1
View File
@@ -80,4 +80,3 @@ SYSTEM_PROMPT="You are a weather forecast agent. You help users to get the weath
- You can install any pip package (if it exists) by running a cell with pip install.
"
PROJECT_TITLE = "您好,我是博微工程理解小助手,您可以问我有关[线路工程]工程数据的相关问题!"
-1
View File
@@ -111,4 +111,3 @@ SYSTEM_PROMPT="You are a weather forecast agent. You help users to get the weath
- You can install any pip package (if it exists) by running a cell with pip install.
"
PROJECT_TITLE = "您好,我是博微工程理解小助手,您可以问我有关[线路工程]工程数据的相关问题!"
+225 -228
View File
@@ -3,7 +3,6 @@ import json
import logging
import time
from typing import Dict, List, Any, Optional, AsyncGenerator
from collections import deque
from aiostream import stream
from fastapi import APIRouter, Request
@@ -14,8 +13,7 @@ from llama_index.core.callbacks import CBEventType
from llama_index.core.chat_engine.types import StreamingAgentChatResponse
from llama_index.core.tools import ToolOutput
from pydantic import BaseModel
from app.api.routers.request.base import userMng, conversations,message,parameter,feedback
from app.api.routers.request.baseConfig import *
from app.api.routers.request.base import userMng, conversations,message,parameter
from app.api.routers.request.models import ChatRequestData,ChatFileUploadRequest
from app.engine import get_chat_engine
import uuid
@@ -26,138 +24,78 @@ api_router = r = APIRouter()
v1_router = v = APIRouter()
class ChatCallbackEvent(BaseModel):
event_type: ChatEventType
event_type: CBEventType
payload: Optional[Dict[str, Any]] = None
event_id: str = ""
def get_common_param(self)-> dict:
def get_retrieval_message(self) -> dict | None:
if self.payload:
nodes = self.payload.get("nodes")
if nodes:
msg = f"根据查询检索到 {len(nodes)} 源文件"
else:
msg = f"查询检索中: '{self.payload.get('query_str')}'"
return {
'event': self.event_type.name,
'conversation_id':self.payload.get("conversation_id"),
'message_id': self.payload.get("message_id"),
'created_at': int(time.time()),
'task_id': self.payload.get("task_id")
"type": "events",
"data": {"title": msg},
}
else:
return None
def get_WorkflowStart_param(self) -> dict:
params = self.get_common_param()
params.update({
'workflow_run_id':self.payload.get('workflow_run_id'),
'data':{
"id": self.payload.get('workflow_run_id'),
"workflow_id": self.payload.get('workflow_id'),
"sequence_number": 1709,
"inputs": {
"sys.query": self.payload.get('query'),
"sys.files": [],
"sys.conversation_id": self.payload.get('conversation_id'),
"sys.user_id": self.payload.get('use_id')
def get_tool_message(self) -> dict | None:
func_call_args = self.payload.get("function_call")
if func_call_args is not None and "tool" in self.payload:
tool = self.payload.get("tool")
return {
"type": "events",
"data": {
"title": f"调用工具 {tool.name} ,参数: {func_call_args}",
},
"created_at": int(time.time())
}
})
return params
def get_WorkflowFinished_param(self) -> dict:
params = self.get_common_param()
params.update({
'workflow_run_id':self.payload.get('workflow_run_id'),
'data':{
"id": self.payload.get('workflow_run_id'),
"workflow_id": self.payload.get('workflow_id'),
"sequence_number": 1709,
"status": "succeeded",
"outputs": {
"answer": self.payload.get('response')
def _is_output_serializable(self, output: Any) -> bool:
try:
json.dumps(output)
return True
except TypeError:
return False
def get_agent_tool_response(self) -> dict | None:
response = self.payload.get("response")
if response is not None:
sources = response.sources
for source in sources:
# Return the tool response here to include the toolCall information
if isinstance(source, ToolOutput):
if self._is_output_serializable(source.raw_output):
output = source.raw_output
else:
output = source.content
return {
"type": "tools",
"data": {
"toolOutput": {
"output": output,
"isError": source.is_error,
},
"toolCall": {
"id": None, # There is no tool id in the ToolOutput
"name": source.tool_name,
"input": source.raw_input,
},
"error": '',
"elapsed_time": 36.03764106379822,
"total_tokens": 11707,
"total_steps": 10,
"created_by": {
"id": str(uuid.uuid4()),
"user": self.payload.get('use_id')
},
"created_at": int(time.time()),
"finished_at": int(time.time()),
"files": []
}
})
return params
def get_NodeStart_param(self) -> dict:
params = self.get_common_param()
params.update({
'workflow_run_id':self.payload.get('workflow_run_id'),
'data':{
"id": self.payload.get('nodeid'),
"node_id": self.payload.get('nodeid'),
"node_type": "http-request",
"title": self.payload.get('title'),
"index": self.payload.get('index'),
"predecessor_node_id": self.payload.get('predecessor_node_id'),
"inputs": '',
"created_at": 1724398751,
"extras": {}
}
})
return params
def get_NodeFinished_param(self) -> dict:
params = self.get_common_param()
params.update({
'workflow_run_id':self.payload.get('workflow_run_id'),
'data':{
"id": self.payload.get('nodeid'),
"node_id": self.payload.get('nodeid'),
"node_type": "http-request",
"title": self.payload.get('title'),
"index": self.payload.get('index'),
"predecessor_node_id": self.payload.get('predecessor_node_id'),
"inputs": '',
"process_data": '',
"outputs": '',
"status": "succeeded",
"error": '',
"elapsed_time": 0.10402441816404462,
"execution_metadata": '',
"created_at": 1724398751,
"finished_at": 1724398751,
"files": []
}
})
return params
def get_Message_param(self) -> dict:
params = self.get_common_param()
params.update({
'id':self.payload.get('message_id'),
'answer':self.payload.get('answer')
})
return params
def get_MessageEnd_param(self) -> dict:
params = self.get_common_param()
params.update({
'id':self.payload.get('message_id'),
'metadata':self.payload.get('metadata')
})
return params
def to_response(self)-> dict|None:
def to_response(self):
try:
match self.event_type:
case "workflow_started":
return self.get_WorkflowStart_param()
case "workflow_finished":
return self.get_WorkflowFinished_param()
case "node_started":
return self.get_NodeStart_param()
case 'node_finished':
return self.get_NodeFinished_param()
case 'message':
return self.get_Message_param()
case 'message_end':
return self.get_MessageEnd_param()
case "retrieve":
return self.get_retrieval_message()
case "function_call":
return self.get_tool_message()
case "agent_step":
return self.get_agent_tool_response()
case _:
return None
except Exception as e:
@@ -168,7 +106,9 @@ class ChatEventCallbackHandler(BaseCallbackHandler):
_aqueue: asyncio.Queue
is_done: bool = False
def __init__(self,**params):
def __init__(
self,
):
"""Initialize the base callback handler."""
ignored_events = [
# CBEventType.CHUNKING,
@@ -179,23 +119,6 @@ class ChatEventCallbackHandler(BaseCallbackHandler):
]
super().__init__(ignored_events, ignored_events)
self._aqueue = asyncio.Queue()
self._response:str = ''
self._params:Dict[str,Any] = params
self._nodeStack:deque = deque()
#添加工作流开始事件
data:ChatRequestData = self._params['data']
args:Dict[str,Any] = self._params['ids']
args.update(
{
'use_id': data.user,
'query': data.query,
'conversation_id': data.conversation_id
}
)
wf_event = ChatCallbackEvent(event_type = ChatEventType.WORKFLOW_START,payload = args)
if wf_event.to_response() is not None:
self._aqueue.put_nowait(wf_event)
def on_event_start(
self,
@@ -206,21 +129,9 @@ class ChatEventCallbackHandler(BaseCallbackHandler):
) -> str:
logger.info("event_start:{} type:{} payload:{}\n".format(event_id, event_type, payload))
self._nodeStack.append(event_id)
nindex = self._nodeStack.count() - 1
args:Dict[str,Any] = self._params['ids']
args.update(
{
'nodeid':event_id,
'title':event_type.name,
'index':nindex + 1,
'predecessor_node_id': self._nodeStack[nindex - 1] if nindex > 0 else ''
}
)
nd_event = ChatCallbackEvent(event_type = ChatEventType.NODE_START,payload = args)
if nd_event.to_response() is not None:
self._aqueue.put_nowait(nd_event)
event = ChatCallbackEvent(event_id=event_id, event_type=event_type, payload=payload)
if event.to_response() is not None:
self._aqueue.put_nowait(event)
def on_event_end(
self,
@@ -230,25 +141,9 @@ class ChatEventCallbackHandler(BaseCallbackHandler):
**kwargs: Any,
) -> None:
logger.info("event_end:{} type:{} payload:{}\n".format(event_id, event_type, payload))
#self.response = payload.get("response","")
args:Dict[str,Any] = self._params['ids']
nodeID = self._nodeStack[-1]
if nodeID == event_id:
nindex = self._nodeStack.count() - 1
args.update(
{
'nodeid':event_id,
'title':event_type.name,
'index':nindex + 1,
'predecessor_node_id':self._nodeStack[nindex - 1] if nindex > 0 else ''
}
)
nd_event = ChatCallbackEvent(event_type = ChatEventType.NODE_FINISHED,payload = args)
if nd_event.to_response() is not None:
self._aqueue.put_nowait(nd_event)
self._nodeStack.pop()
event = ChatCallbackEvent(event_id=event_id, event_type=event_type, payload=payload)
if event.to_response() is not None:
self._aqueue.put_nowait(event)
def start_trace(self, trace_id: Optional[str] = None) -> None:
"""No-op."""
@@ -261,23 +156,6 @@ class ChatEventCallbackHandler(BaseCallbackHandler):
) -> None:
"""No-op."""
logger.info("trace_end:{} trace_map:{}\n".format(trace_id, trace_map))
data:ChatRequestData = self._params['data']
args:Dict[str,Any] = self._params['ids']
args.update(
{
'response':self._response,
'conversation_id': data.conversation_id
}
)
wf_event = ChatCallbackEvent(event_type = ChatEventType.WORKFLOW_FINISHED,payload = args)
if wf_event.to_response() is not None:
self._aqueue.put_nowait(wf_event)
args:Dict[str,Any] = self._params['ids']
msgEnt_event = ChatCallbackEvent(event_type = ChatEventType.MESSAGE_END,payload = args)
if msgEnt_event.to_response() is not None:
self._aqueue.put_nowait(msgEnt_event)
async def async_event_gen(self) -> AsyncGenerator[ChatCallbackEvent, None]:
while not self._aqueue.empty() or not self.is_done:
@@ -295,26 +173,95 @@ class IDManager:
"workflow_id": str(uuid.uuid4())
}
class DifyChatResponseEvent(BaseModel):
event: str
conversation_id: str
message_id: str
created_at: int = int(time.time())
task_id: str
class Workflow_started_DifyChatResponseEvent(DifyChatResponseEvent):
workflow_run_id:str
data:Dict[str,Any]
def __init__(self,**args):
args['data'] = {
"id": args['workflow_run_id'],
"workflow_id": args['workflow_id'],
"sequence_number": 1709,
"inputs": {
"sys.query": args['query'],
"sys.files": [],
"sys.conversation_id": args['conversation_id'],
"sys.user_id": args['use_id']
},
"created_at": int(time.time())
}
args['event'] = 'workflow_started'
super().__init__(**args)
class Workflow_finished_DifyChatResponseEvent(DifyChatResponseEvent):
workflow_run_id:str
data:Dict[str,Any]
def __init__(self,**args):
args['event'] = 'workflow_finished'
args['data'] = {
"id": args['workflow_run_id'],
"workflow_id": args['workflow_id'],
"sequence_number": 1709,
"status": "succeeded",
"outputs": {
"answer": args['response']
},
"error": '',
"elapsed_time": 36.03764106379822,
"total_tokens": 11707,
"total_steps": 10,
"created_by": {
"id": str(uuid.uuid4()),
"user": args['use_id']
},
"created_at": int(time.time()),
"finished_at": int(time.time()),
"files": []
}
super().__init__(**args)
class Message_DifyChatResponseEvent(DifyChatResponseEvent):
id:str
answer:str
def __init__(self,**args):
args['id'] = args['message_id']
args['event'] = 'message'
super().__init__(**args)
class MessageEnd_DifyChatResponseEvent(DifyChatResponseEvent):
id:str
metadata:Dict[str,Any] = {}
def __init__(self,**args):
args['id'] = args['message_id']
args['event'] = 'message_end'
super().__init__(**args)
class ChatStreamResponse(StreamingResponse):
TEXT_PREFIX = "data: "
DATA_PREFIX = "data: "
ids:Dict[str,Any] = {}
data:ChatRequestData = None
@classmethod
def convert_Message(cls, token: str):
params = cls.ids
params.update({
'answer':token,
'conversation_id':cls.data.conversation_id
})
event = ChatCallbackEvent(event_type = ChatEventType.MESSAGE,payload = params)
data_str = json.dumps(event.to_response())
def convert_text(cls, token: str):
# Escape newlines and double quotes to avoid breaking the stream
#token = json.dumps(token)
#return f"data: {{"event": "message", "conversation_id": "80d85523-de92-4b9d-aca0-c48a5eacb068", "message_id": "16a06b1b-a89b-49c0-bc15-123bd999f6d6", "created_at": 1724406492, "task_id": "802f3064-030d-42ac-a882-0e1293712d04", "id": "16a06b1b-a89b-49c0-bc15-123bd999f6d6", "answer": "{token}"}}"
return "\n"
@classmethod
def convert_data(cls, data: dict):
data_str = json.dumps(data)
return f"{cls.DATA_PREFIX}{data_str}\n\n"
@classmethod
def convert_Event(cls, data: dict):
data_str = json.dumps(data)
def convert_event(cls, event: DifyChatResponseEvent):
data_str = json.dumps(event.dict())
return f"{cls.DATA_PREFIX}{data_str}\n\n"
def __init__(
@@ -322,11 +269,8 @@ class ChatStreamResponse(StreamingResponse):
request: Request,
event_handler: ChatEventCallbackHandler,
response: StreamingAgentChatResponse,
data: ChatRequestData,
ids:Dict[str,Any]
data: ChatRequestData
):
ChatStreamResponse.ids = ids
ChatStreamResponse.data = data
content = ChatStreamResponse.content_generator(
request, event_handler, response, data
)
@@ -340,26 +284,41 @@ class ChatStreamResponse(StreamingResponse):
response: StreamingAgentChatResponse,
data: ChatRequestData
):
ids = IDManager().createID()
# Yield the text response
async def _chat_response_generator():
final_response = ""
async for token in response.async_response_gen():
final_response += token
yield ChatStreamResponse.convert_Message(token)
args = ids
args['answer'] = token
args['conversation_id'] = data.conversation_id
event = Message_DifyChatResponseEvent(**args)
yield ChatStreamResponse.convert_event(event)
#yield ChatStreamResponse.convert_text(token)
# 存储消息历史
message().add(user_id=data.user,conversation_id=data.conversation_id,query=data.query,answer=final_response)
# the text_generator is the leading stream, once it's finished, also finish the event stream
event_handler.is_done = True
# 发送工作流结束事件
args = ids
args['response'] = final_response
args['conversation_id'] = data.conversation_id
wf_event = Workflow_finished_DifyChatResponseEvent(**args)
yield ChatStreamResponse.convert_event(wf_event)
msgEnt_event = MessageEnd_DifyChatResponseEvent(**ids)
yield ChatStreamResponse.convert_event(msgEnt_event)
# Yield the events from the event handler
async def _event_generator():
async for event in event_handler.async_event_gen():
event_response = event.to_response()
if event_response is not None:
yield ChatStreamResponse.convert_Event(event_response)
yield ChatStreamResponse.convert_text("")
combine = stream.merge(_chat_response_generator(), _event_generator())
is_stream_started = False
@@ -368,11 +327,25 @@ class ChatStreamResponse(StreamingResponse):
if not is_stream_started:
is_stream_started = True
# 发送工作流开始事件
args = ids
args['use_id'] = data.user
args['query'] = data.query
args['conversation_id'] = data.conversation_id
wf_event = Workflow_started_DifyChatResponseEvent(**args)
yield ChatStreamResponse.convert_event(wf_event)
# Stream a blank message to start the stream
# 发送一个空消息事件
#yield ChatStreamResponse.convert_text("")
yield output
if await request.is_disconnected():
break
@v.post("/chat-messages")
async def post_conversations(request: Request, data: ChatRequestData):
userMng.findNoExistCreate(data.user)
@@ -392,15 +365,14 @@ async def post_conversations(request: Request, data: ChatRequestData):
chat_engine = get_chat_engine(filters=filters, params=params)
# 启动聊天事件监听
ids = IDManager().createID()
event_handler = ChatEventCallbackHandler(ids = ids,data = data)
event_handler = ChatEventCallbackHandler()
chat_engine.callback_manager.handlers.append(event_handler) # type: ignore
# 执行异步聊天
response = await chat_engine.astream_chat(data.query)
# 返回异步消息回应
return ChatStreamResponse(request, event_handler, response, data,ids)
return ChatStreamResponse(request, event_handler, response, data)
@v.get("/messages")
async def query_messages(user:str, conversation_id:str):
@@ -416,9 +388,8 @@ async def query_messages(user:str, conversation_id:str):
for record in records:
res = record.dict()
feeds = feedback().query(res['id'])
res["message_files"] = []
res["feedback"] = {'rating':feeds['rating'] } if feeds != None else ''
res["feedback"] = ''
res["retriever_resources"] = []
res["created_at"] = 1723444905
res["agent_thoughts"] = []
@@ -469,22 +440,48 @@ async def query_conversations(user:str, first_id:str = None, limit:str = None, p
async def query_parameters(user:str):
params = parameter().get(user)
if len(params) == 0:
params = BaseConfig().ParamterCfg()
params = {
"opening_statement": "您好,我是配网D3造价软件小助手,您可以问我有关配网造价软件的相关问题!",
"suggested_questions": [],
"suggested_questions_after_answer": {
"enabled": False
},
"speech_to_text": {
"enabled": False
},
"text_to_speech": {
"enabled": False,
"language": "",
"voice": ""
},
"retriever_resource": {
"enabled": True
},
"annotation_reply": {
"enabled": False
},
"more_like_this": {
"enabled": False
},
"user_input_form": [],
"sensitive_word_avoidance": {
"enabled": False
},
"file_upload": {
"image": {
"enabled": False,
"number_limits": 3,
"transfer_methods": [
"remote_url"
]
}
},
"system_parameters": {
"image_file_size_limit": "10"
}
}
return params
@v.post("/messages/{message_id}/feedbacks")
async def post_feedbacks(request: Request,message_id:str,params:Dict[str,Any]):
if params['rating'] =='null':
feedback().delete(message_id)
else:
condition = {'id':message_id}
results = message().query(**condition)
if len(results) > 0:
result = results[0]
feedback().add(message_id=message_id,query=result['query'],
answer=result['answer'],rating=params['rating'])
@r.post("")
def upload_file(request: ChatFileUploadRequest) -> List[str]:
pass
+2 -32
View File
@@ -25,7 +25,7 @@ class conversations:
return None
def add(self,id:str, user_id:str, name:str):
template = BaseConfig().ConversationCfg()
template = BaseConfig.ConversationCfg
template['id'] = id
template['user_id'] = user_id
template['name'] = name
@@ -111,7 +111,7 @@ class message:
return datas
def add(self,user_id:str,conversation_id:str,query:str,answer:str):
template = BaseConfig.MessageCfg()
template = BaseConfig.MessageCfg
template['id'] = str(uuid.uuid4())
template['user_id'] = user_id
template['conversation_id'] = conversation_id
@@ -122,34 +122,4 @@ class message:
def delete(self,user_id:str):
dbManage.delete(self._tableName,user_id = user_id)
def query(self,**condition):
results = []
records = dbManage.query(self._tableName,**condition)
for record in records:
results.append(record.dict())
return results
class feedback:
def __init__(self) -> None:
self._tableName = 'feedbacks'
dbManage.createTable(self._tableName)
def add(self,message_id:str,query:str,answer:str,rating:str):
record = {
'message_id': message_id,
'query': query,
'answer': answer,
'rating': rating,
}
dbManage.addRecord(self._tableName,record)
def delete(self,message_id:str):
cond = {'message_id':message_id}
dbManage.delete(self._tableName,**cond)
def query(self,message_id:str):
cond = {'message_id':message_id}
records = dbManage.query(self._tableName,**cond)
if len(records) > 0:
return records[0].dict()
return None
+8 -26
View File
@@ -1,15 +1,8 @@
from pydantic import BaseModel
import os
from enum import Enum
class BaseConfig(BaseModel):
projectInfo:str = os.getenv("PROJECT_TITLE","您好,我是博微工程理解小助手,您可以问我有关[线路工程]工程数据的相关问题!")
def ParamterCfg(self):
questions = os.getenv("CONVERSATION_STARTERS", "dev")
return{
"opening_statement": self.projectInfo,
"suggested_questions": questions.split('\n'),
class BaseConfig:
ParamterCfg = {
"opening_statement": "您好,我是配网D3造价软件小助手,您可以问我有关配网造价软件的相关问题!",
"suggested_questions": [],
"suggested_questions_after_answer": {
"enabled": False
},
@@ -48,20 +41,18 @@ class BaseConfig(BaseModel):
}
}
def ConversationCfg(self):
return{
ConversationCfg = {
"id": "",
'user_id':'',
"name": "",
"inputs": {},
"status": "normal",
"introduction": self.projectInfo,
"introduction": ParamterCfg['opening_statement'],
"created_at":''
}
@classmethod
def MessageCfg(cls):
return {
MessageCfg = {
"id": "",
'user_id':'',
"conversation_id": "",
@@ -69,12 +60,3 @@ class BaseConfig(BaseModel):
"query": "",
"answer": ""
}
class ChatEventType(str, Enum):
WORKFLOW_START = "workflow_started"
WORKFLOW_FINISHED = "workflow_finished"
NODE_START = "node_started"
NODE_FINISHED = "node_finished"
MESSAGE = "message"
MESSAGE_END = "message_end"
+9 -22
View File
@@ -2,7 +2,7 @@ import os
from typing import Dict, List, Any
from pydantic import BaseModel
from sqlalchemy import create_engine, Column, String, Integer, JSON,Float
from sqlalchemy import create_engine, Column, String, Integer, JSON
from sqlalchemy.engine.reflection import Inspector
from sqlalchemy.orm import sessionmaker, declarative_base
@@ -24,6 +24,10 @@ class ConversationOrm(Base):
if 'name' in data:
self.name = data['name']
class UserOrm(Base):
__tablename__ = "user"
@@ -47,14 +51,6 @@ class MessagesOrm(Base):
query = Column(String)
answer = Column(String)
class FeedBackOrm(Base):
__tablename__ = "feedbacks"
message_id = Column(String,primary_key=True)
query = Column(String)
answer = Column(String)
rating = Column(String)
#数据结构
class ConversationModel(BaseModel):
id: str
@@ -65,6 +61,7 @@ class ConversationModel(BaseModel):
created_at: int
class Config:
#orm_mode = True
from_attributes=True
@classmethod
@@ -76,6 +73,7 @@ class UserModel(BaseModel):
createtime: str
class Config:
#orm_mode = True
from_attributes=True
@classmethod
@@ -88,6 +86,7 @@ class ParametersModel(BaseModel):
value : Dict[str, Any]
class Config:
#orm_mode = True
from_attributes=True
@classmethod
@@ -102,25 +101,13 @@ class MessagesModel(BaseModel):
answer : str
class Config:
#orm_mode = True
from_attributes=True
@classmethod
def orm(cls):
return MessagesOrm
class FeedBackModel(BaseModel):
message_id :str
query :str
answer :str
rating :str
class Config:
from_attributes=True
@classmethod
def orm(cls):
return FeedBackOrm
class DBManager:
def __init__(self) -> None:
DATABASE_URL = os.getenv("SQLITE_DATABASE_URL")
+1 -3
View File
@@ -1,7 +1,7 @@
from typing import Dict, Any
from pydantic import BaseModel
from typing import Optional
class ChatRequestData(BaseModel):
inputs: Dict[str,Any]
@@ -13,5 +13,3 @@ class ChatRequestData(BaseModel):
class ChatFileUploadRequest(BaseModel):
base64: str
+1 -8
View File
@@ -17,7 +17,7 @@ aiostream = "^0.6.2"
llama-index = "0.10.63"
cachetools = "^5.3.3"
protobuf = "4.25.4"
nltk = "^3.9.1"
nltk = "^3.8.2"
jieba = "^0.42.1"
#arize-phoenix = "^4.12.0"
@@ -35,7 +35,6 @@ chroma="^0.2.0"
llama-index-vector-stores-chroma = "^0.1.10"
llama-index-readers-json = "^0.1.5"
llama-index-retrievers-bm25 = "^0.2.2"
llama-index-experimental = "^0.2.0"
duckduckgo_search = "^6.2.6"
@@ -63,12 +62,6 @@ version = "^0.8"
version = "0.0.7"
[[tool.poetry.source]]
name = "mirrors"
url = "https://pypi.tuna.tsinghua.edu.cn/simple/"
priority = "default"
[build-system]
requires = [ "poetry-core" ]
build-backend = "poetry.core.masonry.api"
-138
View File
@@ -1,138 +0,0 @@
import nest_asyncio
nest_asyncio.apply()
from llama_index.core import SimpleDirectoryReader
from llama_index.core.node_parser import SentenceSplitter
from llama_index.core import VectorStoreIndex
from llama_index.core.evaluation import (
FaithfulnessEvaluator,
DatasetGenerator,
CorrectnessEvaluator,
SemanticSimilarityEvaluator,
)
from llama_index.experimental.param_tuner import ParamTuner
from llama_index.experimental.param_tuner.base import RunResult
from llama_index.llms.openai import OpenAI
import asyncio
# 初始化环境
from app.observability import init_observability
from app.settings import init_settings
from dotenv import load_dotenv
load_dotenv()
init_settings()
init_observability()
# 读取文档
documents = SimpleDirectoryReader("D:/LLM_model/text2sql/zjdataai-app-test/backend/data-test").load_data()
# 参数字典
param_dict = {
"chunk_size": [512, 1024],
"top_k": [1, 5],
"temperature": [0.1, 1.0]
}
# 辅助函数
def _build_index(chunk_size, documents):
# 构建索引
splitter = SentenceSplitter(chunk_size=chunk_size)
vector_index = VectorStoreIndex.from_documents(
documents, transformations=[splitter],
)
return vector_index
# 评估函数
def evaluate_query_engine(query_engine, questions):
loop = asyncio.get_event_loop()
correct, total = loop.run_until_complete(_evaluate_query_engine_async(query_engine, questions))
return correct, total
async def _evaluate_query_engine_async(query_engine, questions):
c = [query_engine.aquery(q) for q in questions]
gathering_future = asyncio.gather(*c)
results = await gathering_future
total_correct = 0
for r in results:
eval_result = (
1 if FaithfulnessEvaluator().evaluate_response(response=r).passing else 0
)
total_correct += eval_result
return total_correct, len(results)
# 生成问题
question_generator = DatasetGenerator.from_documents(documents)
eval_questions = question_generator.generate_questions_from_nodes(1) # 假设生成10个问题
# 打印生成的问题
for i, q in enumerate(eval_questions, start=1):
print(f"问题 {i}: {q}")
# 目标函数
def objective_function(params_dict, documents, questions):
chunk_size = params_dict["chunk_size"]
top_k = params_dict["top_k"]
temperature = params_dict["temperature"]
# 构建索引
vector_index = _build_index(chunk_size, documents)
# 查询引擎
query_engine = vector_index.as_query_engine(
similarity_top_k=top_k, temperature=temperature
)
# 评估查询引擎
correct, total = 0, len(questions)
question_answers = [] # 添加列表来收集问题和答案
for question in questions:
response = query_engine.query(question)
if response is not None:
question_answers.append((question, response.response))
eval_result = FaithfulnessEvaluator().evaluate_response(response=response, query_str=question)
if eval_result.passing:
correct += 1
# 计算分数
score = correct / total if total > 0 else 0
return RunResult(score=score, params=params_dict, question_answers=question_answers)
# 创建 ParamTuner 实例
param_tuner = ParamTuner(
param_fn=lambda params_dict: objective_function(params_dict, documents, eval_questions),
param_dict=param_dict,
show_progress=True,
)
# 调用 tune 方法
results = param_tuner.tune()
best_result = results.best_run_result
best_top_k = best_result.params["top_k"]
best_chunk_size = best_result.params["chunk_size"]
best_temperature = best_result.params["temperature"]
print(f"得分: {best_result.score}")
print(f"Top-k: {best_top_k}")
print(f"文本块大小: {best_chunk_size}")
print(f"温度: {best_temperature}")
# 使用最佳参数再次运行查询引擎,并打印问题与答案
best_vector_index = _build_index(best_chunk_size, documents)
best_query_engine = best_vector_index.as_query_engine(
similarity_top_k=best_top_k, temperature=best_temperature
)
best_question_answers = []
for question in eval_questions:
response = best_query_engine.query(question)
if response is not None:
best_question_answers.append((question, response.response))
# 打印最佳参数下的问题与答案
for i, (question, answer) in enumerate(best_question_answers, start=1):
print(f"最佳参数 - 问题 {i}: {question}\n答案: {answer}\n")
-81
View File
@@ -1,81 +0,0 @@
from app.observability import init_observability
from app.settings import init_settings
from dotenv import load_dotenv
import nest_asyncio
nest_asyncio.apply()
load_dotenv()
from llama_index.core.node_parser import SentenceSplitter
from llama_index.core import (
VectorStoreIndex,
SimpleDirectoryReader,
Response,
)
from llama_index.core.evaluation import (
FaithfulnessEvaluator,
DatasetGenerator,
CorrectnessEvaluator,
SemanticSimilarityEvaluator,)
init_settings()
init_observability()
faith_evaluator_qwen = FaithfulnessEvaluator() #诚实度评测
corr_evaluator_qwen = CorrectnessEvaluator() #准确率评测
Seman_evaluator_qwen = SemanticSimilarityEvaluator()#嵌入相似度评估
documents = SimpleDirectoryReader("D:/LLM_model/text2sql/zjdataai-app-test/backend/data-test").load_data()
splitter = SentenceSplitter(chunk_size=512)
vector_index = VectorStoreIndex.from_documents(
documents, transformations=[splitter],
)
# # 运行评估
# query_engine = vector_index.as_query_engine()
# response_vector = query_engine.query("工程监理费的金额是多少?")
# eval_result = evaluator_qwen.evaluate_response(response=response_vector)
# print(response_vector)
# print(eval_result)
question_generator = DatasetGenerator.from_documents(documents)
eval_questions = question_generator.generate_questions_from_nodes(5)
print(eval_questions)
import asyncio
async def evaluate_query_engine_async(query_engine, questions):
c = [query_engine.aquery(q) for q in questions]
gathering_future = asyncio.gather(*c)
results = await gathering_future
#print(results)
total_correct = 0
for r in results:
eval_result = (
1 if faith_evaluator_qwen.evaluate_response(response=r).passing else 0
)
total_correct += eval_result
return total_correct, len(results)
def evaluate_query_engine(query_engine, questions):
loop = asyncio.get_event_loop()
correct, total = loop.run_until_complete(evaluate_query_engine_async(query_engine, questions))
return correct, total
# 使用 evaluate_query_engine 函数
vector_query_engine = vector_index.as_query_engine()
correct, total = evaluate_query_engine(vector_query_engine, eval_questions[:5])
print(f"score: {correct}/{total}")
-202
View File
@@ -1,202 +0,0 @@
[
{
"question": "人工费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "临时设施费的费率是多少?",
"answer": "费率是6.3500000000"
},
{
"question": "乙供装置性材料费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "直接费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "甲供装置性材料费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "直接费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "夜间施工增加费的费率是多少?",
"answer": "费率是0E-10"
},
{
"question": "装置性材料费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "冬雨季施工增加费的费率是多少?",
"answer": "费率是3.5700000000"
},
{
"question": "材料费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "机械价差的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "规费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "直接工程费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "安全文明施工费的费率是多少?",
"answer": "费率是3.5500000000"
},
{
"question": "企业管理费的费率是多少?",
"answer": "费率是35.7600000000"
},
{
"question": "税金的费率是多少?",
"answer": "费率是9.0000000000"
},
{
"question": "直接费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "安全文明施工费的费率是多少?",
"answer": "费率是3.5500000000"
},
{
"question": "合计的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "税金的费率是多少?",
"answer": "费率是9.0000000000"
},
{
"question": "安全文明施工费的费率是多少?",
"answer": "费率是3.5500000000"
},
{
"question": "直接工程费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "税金的费率是多少?",
"answer": "费率是9.0000000000"
},
{
"question": "社会保险费的费率是多少?",
"answer": "费率是15.0000000000"
},
{
"question": "间接费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "合计的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "临时设施费的费率是多少?",
"answer": "费率是0E-10"
},
{
"question": "利润的费率是多少?",
"answer": "费率是5.2400000000"
},
{
"question": "税金的费率是多少?",
"answer": "费率是9.0000000000"
},
{
"question": "社会保险费的费率是多少?",
"answer": "费率是15.0000000000"
},
{
"question": "直接工程费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "乙供设备不含税价的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "企业管理费的费率是多少?",
"answer": "费率是17.1300000000"
},
{
"question": "企业管理费的费率是多少?",
"answer": "费率是35.7600000000"
},
{
"question": "夜间施工增加费的费率是多少?",
"answer": "费率是0E-10"
},
{
"question": "直接费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "夜间施工增加费的费率是多少?",
"answer": "费率是0E-10"
},
{
"question": "甲供设备含税价的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "施工机械使用费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "安全文明施工费的费率是多少?",
"answer": "费率是3.5500000000"
},
{
"question": "定额直接费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "主材费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "直接费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "施工企业配合调试费的费率是多少?",
"answer": "费率是0E-10"
},
{
"question": "施工机械使用费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "临时设施费的费率是多少?",
"answer": "费率是6.3500000000"
},
{
"question": "施工工具用具使用费的费率是多少?",
"answer": "费率是3.8200000000"
},
{
"question": "措施费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "材料价差的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "措施费的费率是多少?",
"answer": "费率是100.0000000000"
}
]
-202
View File
@@ -1,202 +0,0 @@
[
{
"question": "前期工作管理费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "特种设备安全监测费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "工程监理费的金额是多少?",
"answer": "金额是131009.9200000000"
},
{
"question": "水土保持方案编审费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "生产准备费的金额是多少?",
"answer": "金额是472373669.4635599852"
},
{
"question": "电力工程技术经济标准编制费的金额是多少?",
"answer": "金额是84352440.9756360054"
},
{
"question": "项目建设技术服务费的金额是多少?",
"answer": "金额是16855957065.4302005768"
},
{
"question": "工程保险费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "其他的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "施工图文件评审费的金额是多少?",
"answer": "金额是24940.0000000000"
},
{
"question": "节能评估费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "桩基检测费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "项目前期工作费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "其他的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "项目法人管理费的金额是多少?",
"answer": "金额是986923559.4149370193"
},
{
"question": "专业爆破服务费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "节能评估费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "用地预审费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "设备材料监造费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "环境监测及环境保护验收费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "环境监测及环境保护验收费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "设备材料监造费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "勘察费的金额是多少?",
"answer": "金额是12122154260.0000000000"
},
{
"question": "项目法人管理费的金额是多少?",
"answer": "金额是986923559.4149370193"
},
{
"question": "社会稳定风险评估费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "勘察费的金额是多少?",
"answer": "金额是12122154260.0000000000"
},
{
"question": "环境影响评价费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "水土保持方案编审费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "使用林地可行性研究费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "环境监测及环境保护验收费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "桩基检测费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "设计费的金额是多少?",
"answer": "金额是4042055949.4299998283"
},
{
"question": "环境监测及环境保护验收费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "建设场地征用及清理费的金额是多少?",
"answer": "金额是16831284.2287110016"
},
{
"question": "施工图文件评审费的金额是多少?",
"answer": "金额是24940.0000000000"
},
{
"question": "项目后评价费的金额是多少?",
"answer": "金额是421762204.8781780005"
},
{
"question": "水土保持方案编审费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "勘察设计费的金额是多少?",
"answer": "金额是16164210209.4300003052"
},
{
"question": "前期工作管理费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "节能评估费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "初步设计文件评审费的金额是多少?",
"answer": "金额是18560.0000000000"
},
{
"question": "特种设备安全监测费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "初步设计文件评审费的金额是多少?",
"answer": "金额是18560.0000000000"
},
{
"question": "桩基检测费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "矿产压覆评估费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "设计费的金额是多少?",
"answer": "金额是4042055949.4299998283"
},
{
"question": "水土保持方案编审费用的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "电力工程技术经济标准编制费的金额是多少?",
"answer": "金额是84352440.9756360054"
},
{
"question": "桩基检测费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "矿产压覆评估费用的金额是多少?",
"answer": "金额是0E-10"
}
]
-202
View File
@@ -1,202 +0,0 @@
[
{
"question": "新增项目名称的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "预制基础的合价是多少?",
"answer": "合价是40567.2639480000"
},
{
"question": "绝缘子串及金具安装的合价是多少?",
"answer": "合价是2897171.9878110001"
},
{
"question": "杆塔工程材料工地运输的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "基础防护的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "护坡、挡土墙及排洪沟土石方工程的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "新增项目名称的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "(1)拆除后能利用的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "地基处理的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "灌注桩基础的合价是多少?",
"answer": "合价是43466660.0544390008"
},
{
"question": "(1)拆除后能利用的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "悬垂绝缘子串及金具安装的合价是多少?",
"answer": "合价是1251465.0340440001"
},
{
"question": "护坡、挡土墙及排洪沟土石方工程的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "附件安装工程的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "导地线跨越架设的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "辅助工程的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "新增项目名称的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "绝缘子串及金具安装的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "护坡、挡土墙及排洪沟砌筑的合价是多少?",
"answer": "合价是709931.9013930000"
},
{
"question": "锚杆基础的合价是多少?",
"answer": "合价是15344967.9002950005"
},
{
"question": "建筑工程的合价是多少?",
"answer": "合价是25411.2790780000"
},
{
"question": "辅助工程的合价是多少?",
"answer": "合价是1046253.4135240000"
},
{
"question": "导地线跨越架设的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "电缆工程的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "输、送电线路试运的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "基础土石方工程的合价是多少?",
"answer": "合价是32872843180.7429008484"
},
{
"question": "基础永久性围堰的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "基础永久性围堰的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "混凝土及钢筋混凝土结构的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "输、送电线路试运的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "混合结构的合价是多少?",
"answer": "合价是16967.5193850000"
},
{
"question": "杆塔组立的合价是多少?",
"answer": "合价是2253906.0859830002"
},
{
"question": "附件安装工程的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "接地工程材料工地运输的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "新增项目名称的合价是多少?",
"answer": "合价是27148.0310160000"
},
{
"question": "导地线架设的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "护坡、挡土墙及排洪沟的合价是多少?",
"answer": "合价是709931.9013930000"
},
{
"question": "(1)拆除后能利用的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "基础永久性围堰砌筑的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "(2)拆除后不能利用的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "安装工程的合价是多少?",
"answer": "合价是65324.9496330000"
},
{
"question": "尖峰、施工基面土石方工程的合价是多少?",
"answer": "合价是325205.4178770000"
},
{
"question": "架线工程的合价是多少?",
"answer": "合价是4844399648.0778598785"
},
{
"question": "杆塔组立的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "架线工程材料工地运输的合价是多少?",
"answer": "合价是2088570123.2409000397"
},
{
"question": "导地线架设的合价是多少?",
"answer": "合价是0E-10"
},
{
"question": "耐张绝缘子串及金具安装的合价是多少?",
"answer": "合价是1645706.9537680000"
},
{
"question": "架线工程材料工地运输的合价是多少?",
"answer": "合价是2088570123.2409000397"
},
{
"question": "其他基础的合价是多少?",
"answer": "合价是3839666.7656879998"
},
{
"question": "架线工程材料工地运输的合价是多少?",
"answer": "合价是0E-10"
}
]
@@ -1,202 +0,0 @@
[
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是440877984.9458540082"
},
{
"question": "线路取费表(拆除)的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是1086586.9018659999"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表(拆除)的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是51486.7898090000"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是3321.8139230000"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是78005.0340730000"
},
{
"question": "的直接费是多少?",
"answer": "直接费是3535892767.0972299576"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是24045.2334060000"
},
{
"question": "的直接费是多少?",
"answer": "直接费是336253.7482950000"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是142270.1346780000"
},
{
"question": "的直接费是多少?",
"answer": "直接费是61049.8665780000"
},
{
"question": "线路取费表(拆除)的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是933061.7795919999"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "的直接费是多少?",
"answer": "直接费是182949.5997350000"
},
{
"question": "的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表(余物清理)的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表(拆除)的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是21220645.1637400016"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是933061.7795919999"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是2501470269.7231497765"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是51486.7898090000"
},
{
"question": "的直接费是多少?",
"answer": "直接费是55265.9111100000"
},
{
"question": "的直接费是多少?",
"answer": "直接费是442897633.6273120046"
},
{
"question": "线路取费表(拆除)的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "的直接费是多少?",
"answer": "直接费是1057484.3306960000"
},
{
"question": "的直接费是多少?",
"answer": "直接费是442897633.6273120046"
},
{
"question": "的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是21220645.1637400016"
},
{
"question": "线路取费表(余物清理)的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "的直接费是多少?",
"answer": "直接费是336253.7482950000"
},
{
"question": "的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "的直接费是多少?",
"answer": "直接费是61049.8665780000"
},
{
"question": "线路取费表(余物清理)(1)的直接费是多少?",
"answer": "直接费是61049.8665780000"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是24045.2334060000"
},
{
"question": "线路取费表(拆除)的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表(拆除)的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表(余物清理)的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表(拆除)的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表(拆除)的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是659466.5955000001"
},
{
"question": "线路取费表(拆除)的直接费是多少?",
"answer": "直接费是0E-10"
},
{
"question": "线路取费表的直接费是多少?",
"answer": "直接费是2501470269.7231497765"
}
]
-202
View File
@@ -1,202 +0,0 @@
[
{
"question": "降阻剂_数量的属性值是多少?",
"answer": "属性值是f"
},
{
"question": "导线2_单位单价的属性值是多少?",
"answer": "属性值是9"
},
{
"question": "导线_单公里用量的属性值是多少?",
"answer": "属性值是36"
},
{
"question": "线路参数_导地线防震措施的属性值是多少?",
"answer": "属性值是457"
},
{
"question": "合成绝缘子_数量的属性值是多少?",
"answer": "属性值是5"
},
{
"question": "基础垫层的属性值是多少?",
"answer": "属性值是"
},
{
"question": "其中:基础护壁用量的属性值是多少?",
"answer": "属性值是74394.212"
},
{
"question": "铺石加混凝土的属性值是多少?",
"answer": "属性值是0.0"
},
{
"question": "导线用量(西北)的属性值是多少?",
"answer": "属性值是-795976.0855"
},
{
"question": "导线单公里用量(西北)的属性值是多少?",
"answer": "属性值是-159195.2171"
},
{
"question": "灰土垫层单公里用量(西北)的属性值是多少?",
"answer": "属性值是8.0"
},
{
"question": "地线瓷绝缘子单公里用量(西北)的属性值是多少?",
"answer": "属性值是738.253"
},
{
"question": "地形条件_高山的属性值是多少?",
"answer": "属性值是7"
},
{
"question": "流砂坑比例的属性值是多少?",
"answer": "属性值是0.001"
},
{
"question": "碎石_数量的属性值是多少?",
"answer": "属性值是12"
},
{
"question": "线路参数_导地线防震措施的属性值是多少?",
"answer": "属性值是457"
},
{
"question": "灰土垫层的属性值是多少?",
"answer": "属性值是40.0"
},
{
"question": "交叉跨越_弱电线路的属性值是多少?",
"answer": "属性值是45"
},
{
"question": "地线1_根数的属性值是多少?",
"answer": "属性值是12"
},
{
"question": "土质比例_岩石(人凿)的属性值是多少?",
"answer": "属性值是49"
},
{
"question": "耐张混凝土杆基数的属性值是多少?",
"answer": "属性值是26.0"
},
{
"question": "设计单位的属性值是多少?",
"answer": "属性值是3"
},
{
"question": "接地钢的属性值是多少?",
"answer": "属性值是"
},
{
"question": "间隔棒_单公里用量的属性值是多少?",
"answer": "属性值是r"
},
{
"question": "导线其中:跳线和导线弧垂单公里用量(西北)的属性值是多少?",
"answer": "属性值是159203.0171"
},
{
"question": "桩基础的属性值是多少?",
"answer": "属性值是310.0"
},
{
"question": "降阻剂的属性值是多少?",
"answer": "属性值是"
},
{
"question": "可抵扣增值税(万元)的属性值是多少?",
"answer": "属性值是2005241.808822"
},
{
"question": "主要技术经济指标2的属性值是多少?",
"answer": "属性值是"
},
{
"question": "合成绝缘子_数量的属性值是多少?",
"answer": "属性值是5"
},
{
"question": "土质比例_水坑的属性值是多少?",
"answer": "属性值是47"
},
{
"question": "基础_插入式的属性值是多少?",
"answer": "属性值是3"
},
{
"question": "耐张角钢塔比例的属性值是多少?",
"answer": "属性值是250%"
},
{
"question": "地线的属性值是多少?",
"answer": "属性值是"
},
{
"question": "回路数的属性值是多少?",
"answer": "属性值是三回"
},
{
"question": "导线其中:跳线和导线弧垂用量的属性值是多少?",
"answer": "属性值是796015.0855"
},
{
"question": "OPGW用量(西北)的属性值是多少?",
"answer": "属性值是2904.737"
},
{
"question": "现浇混凝土_单公里用量的属性值是多少?",
"answer": "属性值是22"
},
{
"question": "架线工程费用(万元)(含价差)的属性值是多少?",
"answer": "属性值是3203726.0"
},
{
"question": "耐张钢管塔比例的属性值是多少?",
"answer": "属性值是300%"
},
{
"question": "单公里土石方量_基面的属性值是多少?",
"answer": "属性值是8*8"
},
{
"question": "地线2的属性值是多少?",
"answer": "属性值是"
},
{
"question": "降阻剂的属性值是多少?",
"answer": "属性值是"
},
{
"question": "土质比例的属性值是多少?",
"answer": "属性值是"
},
{
"question": "地线1_单位单价的属性值是多少?",
"answer": "属性值是113"
},
{
"question": "绝缘子串型式_悬垂串的属性值是多少?",
"answer": "属性值是48"
},
{
"question": "基坑土石方量(西北)的属性值是多少?",
"answer": "属性值是405403506.156"
},
{
"question": "基坑坚土的属性值是多少?",
"answer": "属性值是25585167.713"
},
{
"question": "基坑普通土的属性值是多少?",
"answer": "属性值是313873965.334"
},
{
"question": "瓷绝缘子单公里用量(西北)的属性值是多少?",
"answer": "属性值是201.0"
}
]
-202
View File
@@ -1,202 +0,0 @@
[
{
"question": "电杆坑、塔坑、拉线坑人工挖方(或爆破)及回填 水坑 坑深2.0m以内的编码是多少?",
"answer": "编码是YX2-72"
},
{
"question": "钢筋加工及制作的编码是多少?",
"answer": "编码是YX3-43"
},
{
"question": "船舶运输 线材 每件重400kg以内 运输的编码是多少?",
"answer": "编码是YX1-132"
},
{
"question": "船舶运输 钢管塔材 运输的编码是多少?",
"answer": "编码是YX1-152"
},
{
"question": "碎石的编码是多少?",
"answer": "编码是C10020103"
},
{
"question": "混凝土(保护帽)的编码是多少?",
"answer": "编码是ZH1001"
},
{
"question": "船舶运输 金具、绝缘子、零星钢材 运输的编码是多少?",
"answer": "编码是YX1-144"
},
{
"question": "人力运输 混凝土杆 每件重500kg以内的编码是多少?",
"answer": "编码是YX1-1"
},
{
"question": "船舶运输 线材 每件重1000kg以内 运输的编码是多少?",
"answer": "编码是YX1-136"
},
{
"question": "混凝土搅拌及浇制 每基基础联系梁混凝土量20m³以内的编码是多少?",
"answer": "编码是YX3-69"
},
{
"question": "索道运输 循环式 塔材 荷载1t以内 装卸的编码是多少?",
"answer": "编码是YX1-185"
},
{
"question": "人力运输 混凝土预制品 每件重100kg以内的编码是多少?",
"answer": "编码是YX1-6"
},
{
"question": "船舶运输 混凝土杆 每件重1500kg以上 运输的编码是多少?",
"answer": "编码是YX1-118"
},
{
"question": "碎石的编码是多少?",
"answer": "编码是C10020103"
},
{
"question": "电杆坑、塔坑、拉线坑人工挖方(或爆破)及回填 泥水 坑深8.0m以上的编码是多少?",
"answer": "编码是YX2-55"
},
{
"question": "机械施工土方 场地平整的编码是多少?",
"answer": "编码是GT1-1"
},
{
"question": "汽车运输 混凝土预制品 每件重100kg以内 装卸的编码是多少?",
"answer": "编码是YX1-69"
},
{
"question": "汽车运输 其他建筑安装材料 运输的编码是多少?",
"answer": "编码是YX1-108"
},
{
"question": "钻孔灌注桩基础 混凝土搅拌及浇制 孔深10m以内的编码是多少?",
"answer": "编码是YX3-171"
},
{
"question": "线路复测及分坑 直线双杆及拉线塔的编码是多少?",
"answer": "编码是YX2-3"
},
{
"question": "氧化锌避雷器安装 35kV的编码是多少?",
"answer": "编码是YX7-32"
},
{
"question": "混凝土(保护帽)的编码是多少?",
"answer": "编码是ZH1002"
},
{
"question": "汽车运输 其他建筑安装材料 装卸的编码是多少?",
"answer": "编码是YX1-107"
},
{
"question": "船舶运输 混凝土杆 每件重500kg以内 装卸的编码是多少?",
"answer": "编码是YX1-109"
},
{
"question": "混凝土(保护帽)的编码是多少?",
"answer": "编码是ZH1001"
},
{
"question": "人力运输 混凝土杆 每件重500kg以内的编码是多少?",
"answer": "编码是YX1-1"
},
{
"question": "人力运输 混凝土杆 每件重500kg以内的编码是多少?",
"answer": "编码是YX1-1"
},
{
"question": "普通硅酸盐水泥的编码是多少?",
"answer": "编码是C09010102"
},
{
"question": "拖拉机运输 钢管塔材 运输的编码是多少?",
"answer": "编码是YX1-44"
},
{
"question": "尖峰及施工基面挖方(或爆破) 普通土的编码是多少?",
"answer": "编码是YX2-226"
},
{
"question": "汽车运输 角钢塔材 装卸的编码是多少?",
"answer": "编码是YX1-103"
},
{
"question": "接地槽挖方(或爆破)及回填 普通土的编码是多少?",
"answer": "编码是YX2-213"
},
{
"question": "水的编码是多少?",
"answer": "编码是C21010101"
},
{
"question": "直线(直线换位、直线转角)杆塔绝缘子串悬挂安装 35kV 针式单联串(悬垂串)的编码是多少?",
"answer": "编码是YX6-21"
},
{
"question": "直线(直线换位、直线转角)杆塔绝缘子串悬挂安装 35kV I型双联串(悬垂串)的编码是多少?",
"answer": "编码是YX6-22"
},
{
"question": "钻孔灌注桩基础 机械推钻成孔 砂砾石 孔深20m以内 孔径1.0m以内的编码是多少?",
"answer": "编码是YX3-117"
},
{
"question": "线路复测及分坑 直线自立塔的编码是多少?",
"answer": "编码是YX2-6"
},
{
"question": "钻孔灌注桩基础 凿桩头 桩径0.8m以上的编码是多少?",
"answer": "编码是YX3-180"
},
{
"question": "线路复测及分坑 耐张(转角)单杆的编码是多少?",
"answer": "编码是YX2-2"
},
{
"question": "中砂的编码是多少?",
"answer": "编码是C10010101"
},
{
"question": "人力运输 混凝土杆 每件重500kg以内的编码是多少?",
"answer": "编码是YX1-1"
},
{
"question": "带电跨越电力线 被跨线电压等级 35kV的编码是多少?",
"answer": "编码是YX5-186"
},
{
"question": "人工挖土方 普土 深2m以内的编码是多少?",
"answer": "编码是YT1-1"
},
{
"question": "混凝土杆的编码是多少?",
"answer": "编码是"
},
{
"question": "接地模块安装的编码是多少?",
"answer": "编码是YX3-213"
},
{
"question": "拖拉机运输 线材 每件重400kg以内 运输的编码是多少?",
"answer": "编码是YX1-34"
},
{
"question": "拖拉机运输 其他建筑安装材料 装卸的编码是多少?",
"answer": "编码是YX1-45"
},
{
"question": "普通硅酸盐水泥的编码是多少?",
"answer": "编码是C09010102"
},
{
"question": "船舶运输 线材 每件重4000kg以内 装卸的编码是多少?",
"answer": "编码是YX1-139"
},
{
"question": "水的编码是多少?",
"answer": "编码是C21010101"
}
]
-202
View File
@@ -1,202 +0,0 @@
[
{
"question": "架空输电线路本体工程的金额是多少?",
"answer": "金额是55105688268.5176010132"
},
{
"question": "价差预备费的金额是多少?",
"answer": "金额是22731130869.6655998230"
},
{
"question": "工程静态投资的金额是多少?",
"answer": "金额是715035853336.3909912109"
},
{
"question": "工程动态投资的金额是多少?",
"answer": "金额是776282009093.5660400391"
},
{
"question": "其中:工程建设检测费的金额是多少?",
"answer": "金额是185575370.1463980079"
},
{
"question": "工程静态投资的金额是多少?",
"answer": "金额是715035853336.3909912109"
},
{
"question": "建设期贷款利息的金额是多少?",
"answer": "金额是38515024887.5095977783"
},
{
"question": "特殊项目的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "动态费用的金额是多少?",
"answer": "金额是61246155757.1752014160"
},
{
"question": "动态费用的金额是多少?",
"answer": "金额是61246155757.1752014160"
},
{
"question": "小计的金额是多少?",
"answer": "金额是458257942570.3129882812"
},
{
"question": "其他费用的金额是多少?",
"answer": "金额是210942912572.8689880371"
},
{
"question": "基本预备费的金额是多少?",
"answer": "金额是14020310849.7332000732"
},
{
"question": "其中:水土保持监测及验收费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "其中:工程建设检测费的金额是多少?",
"answer": "金额是185575370.1463980079"
},
{
"question": "其中:特种设备安全监测费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "工程静态投资的金额是多少?",
"answer": "金额是715035853336.3909912109"
},
{
"question": "其中:水土保持监测及验收费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "架空输电线路本体工程的金额是多少?",
"answer": "金额是55105688268.5176010132"
},
{
"question": "基本预备费的金额是多少?",
"answer": "金额是14020310849.7332000732"
},
{
"question": "其中:水土保持监测及验收费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "小计的金额是多少?",
"answer": "金额是458257942570.3129882812"
},
{
"question": "编制基准期价差的金额是多少?",
"answer": "金额是29246752707.1180000305"
},
{
"question": "其中:水土保持监测及验收费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "小计的金额是多少?",
"answer": "金额是458257942570.3129882812"
},
{
"question": "其他费用的金额是多少?",
"answer": "金额是210942912572.8689880371"
},
{
"question": "特殊项目的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "编制基准期价差的金额是多少?",
"answer": "金额是29246752707.1180000305"
},
{
"question": "特殊项目的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "小计的金额是多少?",
"answer": "金额是458257942570.3129882812"
},
{
"question": "工程动态投资的金额是多少?",
"answer": "金额是776282009093.5660400391"
},
{
"question": "其中:建设场地征用及清理费的金额是多少?",
"answer": "金额是16831284.2287110016"
},
{
"question": "其中:可抵扣增值税额的金额是多少?",
"answer": "金额是20069645492.2888984680"
},
{
"question": "小计的金额是多少?",
"answer": "金额是458257942570.3129882812"
},
{
"question": "动态费用的金额是多少?",
"answer": "金额是61246155757.1752014160"
},
{
"question": "建设期贷款利息的金额是多少?",
"answer": "金额是38515024887.5095977783"
},
{
"question": "工程静态投资的金额是多少?",
"answer": "金额是715035853336.3909912109"
},
{
"question": "其中:建设场地征用及清理费的金额是多少?",
"answer": "金额是16831284.2287110016"
},
{
"question": "建设期贷款利息的金额是多少?",
"answer": "金额是38515024887.5095977783"
},
{
"question": "工程动态投资的金额是多少?",
"answer": "金额是776282009093.5660400391"
},
{
"question": "架空输电线路本体工程的金额是多少?",
"answer": "金额是55105688268.5176010132"
},
{
"question": "其中:工程建设检测费的金额是多少?",
"answer": "金额是185575370.1463980079"
},
{
"question": "其中:水土保持监测及验收费的金额是多少?",
"answer": "金额是0E-10"
},
{
"question": "工程动态投资的金额是多少?",
"answer": "金额是776282009093.5660400391"
},
{
"question": "其中:可抵扣增值税额的金额是多少?",
"answer": "金额是20069645492.2888984680"
},
{
"question": "价差预备费的金额是多少?",
"answer": "金额是22731130869.6655998230"
},
{
"question": "一般线路本体工程的金额是多少?",
"answer": "金额是55105688268.5176010132"
},
{
"question": "其中:工程建设检测费的金额是多少?",
"answer": "金额是185575370.1463980079"
},
{
"question": "基本预备费的金额是多少?",
"answer": "金额是14020310849.7332000732"
},
{
"question": "设备购置费的金额是多少?",
"answer": "金额是2567934636.3574500084"
}
]
-118
View File
@@ -1,118 +0,0 @@
import json
from dotenv import load_dotenv
import asyncio
import nest_asyncio
nest_asyncio.apply()
from llama_index.core.prompts import (
ChatMessage,
ChatPromptTemplate,
MessageRole
)
DEFAULT_SYSTEM_TEMPLATE = """
您是一个问答聊天机器人的专业评估系统。
您将获得以下信息:
- 用户查询,
- 生成的回答,
也可能提供一个参考答案作为评估的依据。
您的任务是判断生成回答的相关性和正确性。
输出一个代表全面评估的单一分数。
您必须在一行中仅返回该分数。
不要以其他任何格式返回答案。
在单独的一行提供给定分数的理由。
请遵循以下评分指南:
- 您的分数必须在1到5之间,其中1是最差,5是最好的。
-如果生成的回答与用户查询不相关,您应该给出1分。
-如果生成的回答相关但包含错误,您应该给出2到3分之间的分数。
-如果生成的回答相关且完全正确,您应该给出4到5分之间的分数。
示例响应:
4.0
生成的回答与参考答案的指标完全相同,但不够精炼。
"""
DEFAULT_USER_TEMPLATE = """
## User Query
{query}
## Reference Answer
{reference_answer}
## Generated Answer
{generated_answer}
"""
DEFAULT_EVAL_TEMPLATE = ChatPromptTemplate(
message_templates=[
ChatMessage(role=MessageRole.SYSTEM, content=DEFAULT_SYSTEM_TEMPLATE),
ChatMessage(role=MessageRole.USER, content=DEFAULT_USER_TEMPLATE),
]
)
from app.api.routers.models import ChatData, Message
from llama_index.core.chat_engine.types import BaseChatEngine, NodeWithScore
from llama_index.core.vector_stores.types import MetadataFilter, MetadataFilters
from llama_index.core.evaluation import CorrectnessEvaluator
from app.engine import get_chat_engine
from app.api.routers.chat import generate_filters
from app.engine.index import get_index
from app.observability import init_observability
from app.settings import init_settings
load_dotenv()
init_settings()
init_observability()
index = get_index()
# 初始化聊天引擎和评估器
chat_engine = get_chat_engine()
corr_evaluator_qwen = CorrectnessEvaluator()
# 加载本地问题回答文件
file_path = 'D:/LLM_model/text2sql/zjdataai-app-test/backend/unit_test/test.json'
output_file_path = file_path.replace('.json', '_test.json')
with open(file_path, 'r', encoding='utf-8') as f:
data = json.load(f)
# 异步函数用于评估查询
async def evaluate_query(question, answer, index, output_file):
response = await chat_engine.astream_chat(question)
content_str = str(response.sources[0])
result = corr_evaluator_qwen.evaluate(
query=question,
response=content_str,
reference=answer,
)
result_dict = {
"编号": index,
"问题": question,
"答案": answer,
"回答": result.response,
"得分(0~5)": result.score,
"评价": result.feedback
}
with open(output_file, 'a', encoding='utf-8') as f:
f.write(json.dumps(result_dict, ensure_ascii=False, indent=4))
f.write(',')
# 主异步函数
async def main():
for index, item in enumerate(data, start=1):
await evaluate_query(item['question'], item['answer'], index, output_file_path)
# 运行主协程
asyncio.run(main())
-10
View File
@@ -1,10 +0,0 @@
[
{
"question": "人工费的费率是多少?",
"answer": "费率是100.0000000000"
},
{
"question": "临时设施费的费率是多少?",
"answer": "费率是6.3500000000"
}
]
-18
View File
@@ -1,18 +0,0 @@
{
"编号": 1,
"问题": "人工费的费率是多少?",
"答案": "费率是100.0000000000",
"实际回答": "人工费的费率是100.0。",
"得分": 4.5,
"评价": "生成的答案与参考答案一致,只是表达方式略有不同,但完全正确且相关。"
}
********************
{
"编号": 2,
"问题": "临时设施费的费率是多少?",
"答案": "费率是6.3500000000",
"实际回答": "在新的上下文中,临时设施费的计算应当基于取费定额人工费和取费定额机械费,但当前费率被设置为0.0,这意味着需要重新确认或调整费率。在没有具体费率的情况下,不能直接给出确定的费率值。如果需要计算临时设施费,应根据实际的取费定额人工费和取费定额机械费进行计算。参考之前的6.35%的费率可能是一个起点,但请注意,这需要根据项目的具体情况和最新的费用标准进行调整。",
"得分": 1.0,
"评价": "生成的答案与用户查询不相关,用户询问的是临时设施费的费率,而生成的答案提供的是一个关于如何计算临时设施费的解释,且提到了一个不相关的0.0费率,这与用户的问题不符。同时,即使提到了6.35%的费率,也没有明确指出这就是用户想要的答案,反而强调了需要根据项目具体情况调整,这增加了用户的困惑。"
}
********************