-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy patheasyllm-webui.py
374 lines (291 loc) · 17.8 KB
/
easyllm-webui.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
"""本文件为整个项目的主文件，并使用gradio搭建界面"""
import asyncio
import logging
import gradio as gr
import os
import time
import shutil
from langchain_community.document_loaders import TextLoader
from langchain_openai import OpenAIEmbeddings
from langchain.text_splitter import CharacterTextSplitter
from langchain_community.vectorstores import FAISS
from langchain.embeddings.huggingface import HuggingFaceEmbeddings
from modules import utils
from modules.NLG import *
from modules.Search import *
# 设置项目目录
PROJECT_DIR = utils.Configs["settings"]["project_dir"]
VECTOR_STORES_DIR = os.path.join(PROJECT_DIR, "kb")
VECTOR_FILES_DIR = os.path.join(PROJECT_DIR, "file")
embedding_model_path = utils.Configs["settings"]["embedding_model_path"]
nlgService = ChatGLM(utils.Configs["ChatGLM"])
BingService = Search(utils.Configs["Bing Search"])
# 获取项目目录中的所有 FAISS 库
def list_vector_store():
return [f.replace(".faiss", "") for f in os.listdir(VECTOR_STORES_DIR) if f.endswith(".faiss")]
# 初始化下拉列表选项
vector_stores = list_vector_store()
with gr.Blocks(theme=gr.themes.Soft(), title="EasyLLM, a simple chatbot based on LLM",
css="./assets/css/EasyLLMStyle.css", js="./assets/js/EasyLLMStyle.js"
) as demo:
with gr.Row(elem_id="baseContainer"):
with gr.Column(min_width=280, elem_id="sideBar"):
nlgSwitch = gr.Dropdown([i.name for i in NLGEnum], value=nlgService.type.name, interactive=True,
label="选择NLG模型", elem_id="nlgSwitch")
with gr.Column(scale=5, elem_id="chatPanel"):
botComponent = gr.Chatbot(label=nlgService.type.name, avatar_images=utils.getAvatars(), elem_id="chatbot")
with gr.Row(elem_id="inputPanel"):
with gr.Row():
knowledgebaseSwitch = gr.components.Checkbox(label="知识库")
internetSwitch = gr.components.Checkbox(label="互联网")
docSwitch = gr.components.Checkbox(label="文档对话")
textInput = gr.Textbox(placeholder="点击输入", show_label=False, scale=4, elem_id="textInput")
submitButton = gr.Button(value="发送", size="sm", min_width=80, elem_id="submitButton")
clearButton = gr.Button(value="清除", size="sm", min_width=80, elem_id="cleanButton")
with gr.Row(elem_id="KnowledgeBasePanel"):
with gr.Column(min_width=280, elem_id="sideBar"):
select_vector_store = gr.Dropdown(choices=vector_stores, label="选择知识库",placeholder="请选择知识库",
elem_id="vectorStoreSwitch")
with gr.Row(elem_id="inputPanel"):
vector_store_name = gr.Textbox(lines=1, placeholder="输入新建知识库名称",
label="新建知识库")
vector_store_file = gr.File(label="上传文件(支持pdf、txt）", file_types=['text'])
with gr.Row(elem_id="inputPanel"):
build_button = gr.Button(value="知识库创建", size="sm", min_width=80, elem_id="submitButton")
load_button = gr.Button(value="文档加载", size="sm", min_width=80, elem_id="submitButton")
with gr.Column(scale=5, elem_id="searchPanel"):
with gr.Row(elem_id="inputPanel"):
query = gr.Textbox(lines=5, placeholder="输入要查询的句子", label="查询知识库",
scale=4, elem_id="queryBox")
search_button = gr.Button(value="查询", size="sm", min_width=80, elem_id="submitButton")
with gr.Row():
similar_sentences_output = gr.Textbox(label="查询结果", elem_id="similarSentencesBox")
execution_time_output = gr.Textbox(label="查询时间", elem_id="executionTimeBox")
def cleanAllContent(message, chatHistory):
"""
清除全部消息
"""
return "", []
def textChat(message: str, chatHistory: list, internetSwitch: bool, knowledgebaseSwitch: bool, docSwitch:bool, select_vector_store: str):
"""
与聊天机器人进行文本聊天
:param message: str 用户输入的消息
:param chatHistory: [[str, str]...] 分别为用户输入和机器人回复(先前的)
"""
if internetSwitch:
print("Bing search starting...")
internetmessage = BingService.search_with_bing(message)
botMessage = nlgService.continuedQuery(internetmessage, chatHistory)
chatHistory.append((message, botMessage))
elif knowledgebaseSwitch:
print("Knowledge base search starting...")
knowledgebasesearch, _ = search_similar_sentences(message, select_vector_store)
KB_prompt = """
你是大型语言人工智能助手。你将接收到一个用户问题，请针对该问题提供清晰、简洁且准确的答案。你将获得一系列与问题相关的上下文信息，如果适用，请使用上下文信息。
你的答案必须正确、准确，并且以专家的角度用客观、专业的语气撰写。请将回答限制在1024个字符内。不要提供与问题无关的信息，也不要重复。如果给定的上下文没有提供足够的信息，请在相关主题后注明“信息缺失”。
除了代码、特定的名称和引用外，你的答案必须使用与问题相同的语言撰写，一定要用中文回答中文问题。
以下是上下文信息集：
{context}
请记住，不要盲目地逐字重复上下文内容。以下是用户的问题：
"""
knowledgebasemessage = KB_prompt.format(context = knowledgebasesearch) + message
botMessage = nlgService.continuedQuery(knowledgebasemessage, chatHistory)
chatHistory.append((message, botMessage))
elif docSwitch:
print("document search starting...")
docsearch, _ = search_similar_sentences(message, select_vector_store)
KB_prompt = """
你是大型语言人工智能助手。你将接收到一个用户问题，请针对该问题提供清晰、简洁且准确的答案。你将获得一系列与问题相关的上下文信息，如果适用，请使用上下文信息。
你的答案必须正确、准确，并且以专家的角度用客观、专业的语气撰写。请将回答限制在1024个字符内。不要提供与问题无关的信息，也不要重复。如果给定的上下文没有提供足够的信息，请在相关主题后注明“信息缺失”。
除了代码、特定的名称和引用外，你的答案必须使用与问题相同的语言撰写，一定要用中文回答中文问题。
以下是上下文信息集：
{context}
请记住，不要盲目地逐字重复上下文内容。以下是用户的问题：
"""
docmessage = KB_prompt.format(context=docsearch) + message
botMessage = nlgService.continuedQuery(docmessage, chatHistory)
chatHistory.append((message, botMessage))
else:
botMessage = nlgService.continuedQuery(message, chatHistory)
chatHistory.append((message, botMessage))
return "", chatHistory, internetSwitch, knowledgebaseSwitch, docSwitch, select_vector_store
def switchNLG(selectService: str):
"""
切换NLG模型
:param selectService: str NLG模型名称
:return: str NLG模型名称
"""
global nlgService, nlgSwitch
currentService = nlgService.type.name # 当前的NLG模型
if selectService == currentService:
return currentService
else: # 尝试切换模型
try:
if selectService == NLGEnum.Azure.name:
tempService = Azure(utils.Configs["Azure OpenAI"])
elif selectService == NLGEnum.ChatGLM.name:
tempService = ChatGLM(utils.Configs["ChatGLM"])
else: # 未知的模型选择，不执行切换
gr.Warning(f"未知的NLG模型，将不进行切换，当前：{currentService}")
return currentService
nlgService = tempService
gr.Info(f"模型切换成功，当前：{nlgService.type.name}")
return nlgService.type.name
except Exception:
gr.Warning("模型切换失败，请检查网络连接或模型配置")
return currentService
async def create_and_save_vector_store(vector_store_name, file):
model_name = "/root/projects/gpt2_cn/LLMs/text2vec-large-chinese"
model_kwargs = {"device": "cpu"}
hf = HuggingFaceEmbeddings(
model_name=model_name, model_kwargs=model_kwargs
)
text_splitter = CharacterTextSplitter(
separator="\n",
chunk_size=100,
chunk_overlap=0,
length_function=len,
is_separator_regex=False,
)
file_name = file.name
filename = os.path.splitext(os.path.basename(file_name))[0]
file_type = os.path.splitext(os.path.basename(file_name))[1]
save_path = os.path.join(VECTOR_FILES_DIR, filename)
# 保存文件
shutil.copyfile(file.name, save_path)
documents = []
texts = None
try:
if file_type == ".pdf":
from langchain_community.document_loaders import PyPDFLoader
loader = PyPDFLoader(save_path)
documents = loader.load_and_split()
else:
logging.debug("Loading text file...")
from langchain.document_loaders import TextLoader
loader = TextLoader(save_path, "utf8")
texts = loader.load()
documents = text_splitter.split_documents(texts)
except Exception as e:
import traceback
logging.error(f"Error loading file: {filename}")
traceback.print_exc()
db = FAISS.from_documents(documents, hf)
db.save_local(os.path.join(VECTOR_STORES_DIR, f"{vector_store_name}.faiss"))
# 异步更新下拉列表选项
async def update_dropdown():
await asyncio.sleep(0.1) # 等待一小段时间确保文件已保存
select_vector_store.choices = list_vector_store()
gr.Info(f"新知识库创建成功：{vector_store_name}")
await update_dropdown()
return db
def construct_index(file):
model_name = "/root/projects/gpt2_cn/LLMs/text2vec-large-chinese"
model_kwargs = {"device": "cpu"}
hf = HuggingFaceEmbeddings(
model_name=model_name, model_kwargs=model_kwargs
)
text_splitter = CharacterTextSplitter(
separator="\n",
chunk_size=100,
chunk_overlap=0,
length_function=len,
is_separator_regex=False,
)
file_name = file.name
filename = os.path.splitext(os.path.basename(file_name))[0]
file_type = os.path.splitext(os.path.basename(file_name))[1]
save_path = os.path.join(VECTOR_FILES_DIR, filename)
# 保存文件
shutil.copyfile(file.name, save_path)
documents = []
texts = None
try:
if file_type == ".pdf":
from langchain_community.document_loaders import PyPDFLoader
loader = PyPDFLoader(save_path)
documents = loader.load_and_split()
elif file_type == ".docx":
logging.debug("Loading Word...")
from langchain.document_loaders import UnstructuredWordDocumentLoader
loader = UnstructuredWordDocumentLoader(save_path)
texts = loader.load()
documents = text_splitter.split_documents(texts)
elif file_type == ".pptx":
logging.debug("Loading PowerPoint...")
from langchain.document_loaders import UnstructuredPowerPointLoader
loader = UnstructuredPowerPointLoader(save_path)
texts = loader.load()
documents = text_splitter.split_documents(texts)
elif file_type == ".epub":
logging.debug("Loading EPUB...")
from langchain.document_loaders import UnstructuredEPubLoader
loader = UnstructuredEPubLoader(save_path)
texts = loader.load()
documents = text_splitter.split_documents(texts)
else:
logging.debug("Loading text file...")
from langchain.document_loaders import TextLoader
loader = TextLoader(save_path, "utf8")
texts = loader.load()
documents = text_splitter.split_documents(texts)
except Exception as e:
import traceback
logging.error(f"Error loading file: {filename}")
traceback.print_exc()
db = FAISS.from_documents(documents, hf)
db.save_local(os.path.join(VECTOR_STORES_DIR, f"{filename}.faiss"))
gr.Info(f"临时文档加载完成: {filename} ")
return filename
def load_vector_store(vector_store_name):
model_name = "/root/projects/gpt2_cn/LLMs/text2vec-large-chinese"
model_kwargs = {"device": "cpu"}
hf = HuggingFaceEmbeddings(
model_name=model_name, model_kwargs=model_kwargs
)
db = FAISS.load_local(os.path.join(VECTOR_STORES_DIR, f"{vector_store_name}.faiss"), hf)
# print(db)
gr.Info(f"知识库切换成功，当前：{vector_store_name}")
return db
def search_similar_sentences(query, vector_store_name):
print(f"Selected vector store name: {vector_store_name}")
start_time = time.time()
model_name = "/root/projects/gpt2_cn/LLMs/text2vec-large-chinese"
model_kwargs = {"device": "cpu"}
hf = HuggingFaceEmbeddings(
model_name=model_name, model_kwargs=model_kwargs
)
embeddings = hf.embed_query(query)
# print("query:", query)
# print("embeddings:", embeddings)
vector_store = load_vector_store(vector_store_name)
docs = vector_store.similarity_search_by_vector(embeddings, k=3)
end_time = time.time()
similar_sentences = [doc.page_content for doc in docs]
execution_time = end_time - start_time
return similar_sentences, execution_time
def vector_store_builder_and_search(query, vector_store_name, raw_documents=None):
if raw_documents:
create_and_save_vector_store(vector_store_name, raw_documents)
vector_store = load_vector_store(vector_store_name)
similar_sentences, execution_time = search_similar_sentences(query, vector_store)
return similar_sentences, execution_time
def list_vector_store():
# 列出项目目录中的所有 FAISS 库
vector_stores = [f.replace(".faiss", "") for f in os.listdir(VECTOR_STORES_DIR) if f.endswith(".faiss")]
return vector_stores
# 按钮绑定事件
clearButton.click(cleanAllContent, [textInput, botComponent], [textInput, botComponent])
submitButton.click(textChat, [textInput, botComponent, internetSwitch, knowledgebaseSwitch, docSwitch, select_vector_store], [textInput, botComponent, internetSwitch, knowledgebaseSwitch, docSwitch, select_vector_store])
textInput.submit(textChat, [textInput, botComponent, internetSwitch, knowledgebaseSwitch, docSwitch, select_vector_store], [textInput, botComponent, internetSwitch, knowledgebaseSwitch, docSwitch, select_vector_store])
# 切换模型
nlgSwitch.change(switchNLG, [nlgSwitch], [nlgSwitch])
select_vector_store.change(load_vector_store, [select_vector_store])
build_button.click(create_and_save_vector_store, [vector_store_name, vector_store_file])
load_button.click(construct_index, [vector_store_file], [select_vector_store])
query.submit(search_similar_sentences, [query, select_vector_store],
[similar_sentences_output, execution_time_output])
search_button.click(search_similar_sentences, [query, select_vector_store],
[similar_sentences_output, execution_time_output])
if __name__ == "__main__":
demo.launch()