forked from chatchat-space/Langchain-Chatchat
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathscnu_webui.py
457 lines (407 loc) · 25.9 KB
/
scnu_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
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
import gradio as gr
import os
import shutil
# from chains.local_doc_qa import LocalDocQA
from chains.scnu_local_doc_qa import LocalDocQA
# from configs.model_config import *
from configs.scnu_model_config import *
import nltk
nltk.data.path = [NLTK_DATA_PATH] + nltk.data.path
def get_vs_list():
lst_default = ["SCNU_KB_v4"]
if not os.path.exists(VS_ROOT_PATH):
return lst_default
lst = os.listdir(VS_ROOT_PATH)
if not lst:
return lst_default
lst.sort()
return lst_default + lst
vs_list = get_vs_list()
embedding_model_dict_list = list(embedding_model_dict.keys())
llm_model_dict_list = list(llm_model_dict.keys())
local_doc_qa = LocalDocQA()
flag_csv_logger = gr.CSVLogger()
def get_answer(query, vs_path , history, mode = "知识库问答", score_threshold=VECTOR_SEARCH_SCORE_THRESHOLD,
vector_search_top_k=VECTOR_SEARCH_TOP_K, chunk_conent: bool = True,
chunk_size=CHUNK_SIZE, streaming: bool = STREAMING):
if mode == "知识库问答" and os.path.exists(vs_path):
for resp, history in local_doc_qa.get_knowledge_based_answer(
query=query, vs_path=vs_path, chat_history=history, streaming=streaming):
source = "\n\n"
source += "".join(
[f"""<details> <summary>出处 [{i + 1}] {os.path.split(doc.metadata["source"])[-1]}</summary>\n"""
f"""相关度：{round((doc.metadata['score']) / 1100, 4)}\n\n""" # 相似度
f"""{doc.page_content}\n""" # 这个page_content是被chunk过的，CHUNK_SIZE默认是250，在config.py
f"""</details>"""
for i, doc in
enumerate(resp["source_documents"])])
history[-1][-1] += source
yield history, ""
elif mode == "知识库测试":
if os.path.exists(vs_path):
resp, prompt = local_doc_qa.get_knowledge_based_conent_test(query=query, vs_path=vs_path,
score_threshold=score_threshold,
vector_search_top_k=vector_search_top_k,
chunk_conent=chunk_conent,
chunk_size=chunk_size)
if not resp["source_documents"]:
yield history + [[query,
"根据您的设定，没有匹配到任何内容，请确认您设置的知识相关度 Score 阈值是否过小或其他参数是否正确。"]], ""
else:
source = "\n".join(
[
f"""<details open> <summary>【知识相关度 Score】：{doc.metadata["score"]} - 【出处{i + 1}】： {os.path.split(doc.metadata["source"])[-1]} </summary>\n"""
f"""{doc.page_content}\n"""
f"""</details>"""
for i, doc in
enumerate(resp["source_documents"])])
history.append([query, "以下内容为知识库中满足设置条件的匹配结果：\n\n" + source])
yield history, ""
else:
yield history + [[query,
"请选择知识库后进行测试，当前未选择知识库。"]], ""
else:
for resp, history in local_doc_qa.llm._call(query, history, streaming=streaming):
history[-1][-1] = resp + (
"\n\n当前知识库为空，如需基于知识库进行问答，请先加载知识库后，再进行提问。" if mode == "知识库问答" else "")
yield history, ""
logger.info(f"flagging: username={FLAG_USER_NAME},query={query},vs_path={vs_path},mode={mode},history={history}")
# flag_csv_logger.flag([query, vs_path, history, mode], username=FLAG_USER_NAME)
def init_model():
try:
local_doc_qa.init_cfg()
local_doc_qa.llm._call("你好")
reply = """模型已成功加载，可以开始对话"""
logger.info(reply)
return reply
except Exception as e:
logger.error(e)
reply = """模型未成功加载，请到页面左上角"模型配置"选项卡中重新选择后点击"加载模型"按钮"""
if str(e) == "Unknown platform: darwin":
logger.info("该报错可能因为您使用的是 macOS 操作系统，需先下载模型至本地后执行 Web UI，具体方法请参考项目 README 中本地部署方法及常见问题："
" https://github.com/imClumsyPanda/langchain-ChatGLM")
else:
logger.info(reply)
return reply
def reinit_model(llm_model, embedding_model, llm_history_len, use_ptuning_v2, use_lora, top_k, history):
try:
local_doc_qa.init_cfg(llm_model=llm_model,
embedding_model=embedding_model,
llm_history_len=llm_history_len,
use_ptuning_v2=use_ptuning_v2,
use_lora=use_lora,
top_k=top_k, )
model_status = """模型已成功重新加载，可以开始对话"""
logger.info(model_status)
except Exception as e:
logger.error(e)
model_status = """模型未成功重新加载，请到页面左上角"模型配置"选项卡中重新选择后点击"加载模型"按钮"""
logger.info(model_status)
return history + [[None, model_status]]
def get_vector_store(vs_id, files, sentence_size, history, one_conent, one_content_segmentation):
vs_path = os.path.join(VS_ROOT_PATH, vs_id)
# filelist = []
# if not os.path.exists(os.path.join(UPLOAD_ROOT_PATH, vs_id)):
# os.makedirs(os.path.join(UPLOAD_ROOT_PATH, vs_id))
# if local_doc_qa.llm and local_doc_qa.embeddings:
# if isinstance(files, list):
# for file in files:
# filename = os.path.split(file.name)[-1]
# shutil.move(file.name, os.path.join(UPLOAD_ROOT_PATH, vs_id, filename))
# filelist.append(os.path.join(UPLOAD_ROOT_PATH, vs_id, filename))
vs_path, _ = local_doc_qa.init_knowledge_vector_store(vs_path, sentence_size)
# else:
# vs_path, loaded_files = local_doc_qa.one_knowledge_add(vs_path, files, one_conent, one_content_segmentation,
# sentence_size)
# if len(loaded_files):
# file_status = f"已添加 {'、'.join([os.path.split(i)[-1] for i in loaded_files])} 内容至知识库，并已加载知识库，请开始提问"
# else:
# file_status = "文件未成功加载，请重新上传文件"
# else:
# file_status = "模型未完成加载，请先在加载模型后再导入文件"
# vs_path = None
logger.info(file_status)
return vs_path, None, history + [[None, file_status]]
# def change_vs_name_input(vs_id, history):
# if vs_id == "新建知识库":
# return gr.update(visible=True), gr.update(visible=True), gr.update(visible=False), None, history
# else:
# file_status = f"已加载知识库{vs_id}，请开始提问"
# return gr.update(visible=False), gr.update(visible=False), gr.update(visible=True), os.path.join(VS_ROOT_PATH,
# vs_id), history + [
# [None, file_status]]
knowledge_base_test_mode_info = ("【注意】\n\n"
"1. 您已进入知识库测试模式，您输入的任何对话内容都将用于进行知识库查询，"
"并仅输出知识库匹配出的内容及相似度分值和及输入的文本源路径，查询的内容并不会进入模型查询。\n\n"
"2. 知识相关度 Score 经测试，建议设置为 500 或更低，具体设置情况请结合实际使用调整。"
"""3. 使用"添加单条数据"添加文本至知识库时，内容如未分段，则内容越多越会稀释各查询内容与之关联的score阈值。\n\n"""
"4. 单条内容长度建议设置在100-150左右。\n\n"
"5. 本界面用于知识入库及知识匹配相关参数设定，但当前版本中，"
"本界面中修改的参数并不会直接修改对话界面中参数，仍需前往`configs/model_config.py`修改后生效。"
"相关参数将在后续版本中支持本界面直接修改。")
# def change_mode(mode, history):
# if mode == "知识库问答":
# return gr.update(visible=True), gr.update(visible=False), history
# # + [[None, "【注意】：您已进入知识库问答模式，您输入的任何查询都将进行知识库查询，然后会自动整理知识库关联内容进入模型查询！！！"]]
# elif mode == "知识库测试":
# return gr.update(visible=True), gr.update(visible=True), [[None,
# knowledge_base_test_mode_info]]
# else:
# return gr.update(visible=False), gr.update(visible=False), history
# def change_chunk_conent(mode, label_conent, history):
# conent = ""
# if "chunk_conent" in label_conent:
# conent = "搜索结果上下文关联"
# elif "one_content_segmentation" in label_conent: # 这里没用上，可以先留着
# conent = "内容分段入库"
# if mode:
# return gr.update(visible=True), history + [[None, f"【已开启{conent}】"]]
# else:
# return gr.update(visible=False), history + [[None, f"【已关闭{conent}】"]]
# def add_vs_name(vs_name, vs_list, chatbot):
# if vs_name in vs_list:
# vs_status = "与已有知识库名称冲突，请重新选择其他名称后提交"
# chatbot = chatbot + [[None, vs_status]]
# return gr.update(visible=True), vs_list, gr.update(visible=True), gr.update(visible=True), gr.update(
# visible=False), chatbot
# else:
# vs_status = f"""已新增知识库"{vs_name}",将在上传文件并载入成功后进行存储。请在开始对话前，先完成文件上传。 """
# chatbot = chatbot + [[None, vs_status]]
# return gr.update(visible=True, choices=[vs_name] + vs_list, value=vs_name), [vs_name] + vs_list, gr.update(
# visible=False), gr.update(visible=False), gr.update(visible=True), chatbot
block_css = """.importantButton {
background: linear-gradient(45deg, #7e0570,#5d1c99, #6e00ff) !important;
border: none !important;
}
.importantButton:hover {
background: linear-gradient(45deg, #ff00e0,#8500ff, #6e00ff) !important;
border: none !important;
}"""
webui_title = """
# 🎉SCNU_QA_BOT 🎉
🚩本项目基于[langchain-ChatGLM](https://github.com/imClumsyPanda/langchain-ChatGLM)开发
# 模型说明v4.0🎈update:5.20
- 经收集了2014年2023.5.17的文章2200余篇，已全部爬取，知识库仅做学术研究
# 重要声明🎇
1. 模型生成信息不代表作者立场，请谨慎甄别有害的虚假信息
2. 我们坚定认为香港、澳门、台湾是中华人民共和国的一部分。坚决维护中国主权完整和用户中国共产党的领导
3. 本项目属于学术研究性质，不参与商用
# 联系开发者😎
@ [banskyc](https://github.com/bansky-cl)
# 历史更新⏰
- update:5.17 模型说明v3.0，已经收集了2017.2.23到2023.5.13的文章约1230篇，目前文章爬取进度64%
- update:5.13 模型说明v2.0，知识库来源是SCNU的官方公众号《晚安华师》，收集了2019.12.25到2023.5.13的文章约750篇，目前文章爬取进度40%
"""
default_vs = vs_list[0] if len(vs_list) > 1 else "为空"
init_message = f"""欢迎使用 SCNU-QA Web UI！
当前知识库{default_vs}, 可开始提问，尽量提问与华师相关内容
"""
model_status = init_model()
# for file in files:
# local_doc_qa.get_knowledge_based_answer(query=file.replace('.txt',''),vs_path=vs_path,streaming=True)
with gr.Blocks(css=block_css) as demo:
vs_path, file_status, model_status, vs_list = gr.State(
os.path.join(VS_ROOT_PATH, vs_list[0]) if len(vs_list) > 1 else ""), gr.State(""), gr.State(
model_status), gr.State(vs_list)
# files = os.listdir('content/0518_9k3')
# for file in files:
# print(file)
# print(type(file))
# q = file.replace('.txt','')
# print(q)
# print(type(q))
# local_doc_qa.get_knowledge_based_answer(query=q, vs_path=vs_path, streaming=True)
gr.Markdown(webui_title)
# with gr.Tab("对话"):
with gr.Row():
with gr.Column(scale=10):
chatbot = gr.Chatbot([[None, init_message], [None, model_status.value]],
elem_id="chat-box",
show_label=False).style(height=750)
query = gr.Textbox(show_label=False,
placeholder="请输入提问内容，按回车进行提交").style(container=False)
# with gr.Column(scale=5):
# mode = gr.Radio(["LLM 对话", "知识库问答"],
# label="请选择使用模式",
# value="知识库问答")
# knowledge_set = gr.Accordion("知识库设定", visible=False)
# vs_setting = gr.Accordion("配置知识库")
# mode.change(fn=change_mode,
# inputs=[mode, chatbot],
# outputs=[vs_setting, knowledge_set, chatbot])
# with vs_setting:
# select_vs = gr.Dropdown(vs_list.value,
# label="请选择要加载的知识库",
# interactive=True,
# value=vs_list.value[0] if len(vs_list.value) > 0 else None
# )
# vs_name = gr.Textbox(label="请输入新建知识库名称，当前知识库命名暂不支持中文",
# lines=1,
# interactive=True,
# visible=True)
# vs_add = gr.Button(value="添加至知识库选项", visible=True)
# file2vs = gr.Column(visible=False)
# with file2vs:
# # load_vs = gr.Button("加载知识库")
# gr.Markdown("向知识库中添加文件")
# sentence_size = gr.Number(value=SENTENCE_SIZE, precision=0,
# label="文本入库分句长度限制",
# interactive=True, visible=True)
# with gr.Tab("上传文件"):
# files = gr.File(label="添加文件",
# file_types=['.txt', '.md', '.docx', '.pdf'],
# file_count="multiple",
# show_label=False)
# load_file_button = gr.Button("上传文件并加载知识库")
# with gr.Tab("上传文件夹"):
# folder_files = gr.File(label="添加文件",
# # file_types=['.txt', '.md', '.docx', '.pdf'],
# file_count="directory",
# show_label=False)
# load_folder_button = gr.Button("上传文件夹并加载知识库")
# vs_add.click(fn=add_vs_name,
# inputs=[vs_name, vs_list, chatbot],
# outputs=[select_vs, vs_list, vs_name, vs_add, file2vs, chatbot])
# select_vs.change(fn=change_vs_name_input,
# inputs=[select_vs, chatbot],
# outputs=[vs_name, vs_add, file2vs, vs_path, chatbot])
# load_file_button.click(get_vector_store,
# show_progress=True,
# inputs=[select_vs, files, sentence_size, chatbot, vs_add, vs_add],
# outputs=[vs_path, files, chatbot], )
# load_folder_button.click(get_vector_store,
# show_progress=True,
# inputs=[select_vs, folder_files, sentence_size, chatbot, vs_add,
# vs_add],
# outputs=[vs_path, folder_files, chatbot], )
# flag_csv_logger.setup([query, vs_path, chatbot], "flagged")
query.submit(get_answer,[ query, vs_path, chatbot],[chatbot, query])
# with gr.Tab("知识库测试 Beta"):
# with gr.Row():
# with gr.Column(scale=10):
# chatbot = gr.Chatbot([[None, knowledge_base_test_mode_info]],
# elem_id="chat-box",
# show_label=False).style(height=750)
# query = gr.Textbox(show_label=False,
# placeholder="请输入提问内容，按回车进行提交").style(container=False)
# with gr.Column(scale=5):
# mode = gr.Radio(["知识库测试"], # "知识库问答",
# label="请选择使用模式",
# value="知识库测试",
# visible=False)
# knowledge_set = gr.Accordion("知识库设定", visible=True)
# vs_setting = gr.Accordion("配置知识库", visible=True)
# mode.change(fn=change_mode,
# inputs=[mode, chatbot],
# outputs=[vs_setting, knowledge_set, chatbot])
# with knowledge_set:
# score_threshold = gr.Number(value=VECTOR_SEARCH_SCORE_THRESHOLD,
# label="知识相关度 Score 阈值，分值越低匹配度越高",
# precision=0,
# interactive=True)
# vector_search_top_k = gr.Number(value=VECTOR_SEARCH_TOP_K, precision=0,
# label="获取知识库内容条数", interactive=True)
# chunk_conent = gr.Checkbox(value=False,
# label="是否启用上下文关联",
# interactive=True)
# chunk_sizes = gr.Number(value=CHUNK_SIZE, precision=0,
# label="匹配单段内容的连接上下文后最大长度",
# interactive=True, visible=False)
# chunk_conent.change(fn=change_chunk_conent,
# inputs=[chunk_conent, gr.Textbox(value="chunk_conent", visible=False), chatbot],
# outputs=[chunk_sizes, chatbot])
# with vs_setting:
# select_vs = gr.Dropdown(vs_list.value,
# label="请选择要加载的知识库",
# interactive=True,
# value=vs_list.value[0] if len(vs_list.value) > 0 else None)
# vs_name = gr.Textbox(label="请输入新建知识库名称，当前知识库命名暂不支持中文",
# lines=1,
# interactive=True,
# visible=True)
# vs_add = gr.Button(value="添加至知识库选项", visible=True)
# file2vs = gr.Column(visible=False)
# with file2vs:
# # load_vs = gr.Button("加载知识库")
# gr.Markdown("向知识库中添加单条内容或文件")
# sentence_size = gr.Number(value=SENTENCE_SIZE, precision=0,
# label="文本入库分句长度限制",
# interactive=True, visible=True)
# with gr.Tab("上传文件"):
# files = gr.File(label="添加文件",
# file_types=['.txt', '.md', '.docx', '.pdf'],
# file_count="multiple",
# show_label=False
# )
# load_file_button = gr.Button("上传文件并加载知识库")
# with gr.Tab("上传文件夹"):
# folder_files = gr.File(label="添加文件",
# # file_types=['.txt', '.md', '.docx', '.pdf'],
# file_count="directory",
# show_label=False)
# load_folder_button = gr.Button("上传文件夹并加载知识库")
# with gr.Tab("添加单条内容"):
# one_title = gr.Textbox(label="标题", placeholder="请输入要添加单条段落的标题", lines=1)
# one_conent = gr.Textbox(label="内容", placeholder="请输入要添加单条段落的内容", lines=5)
# one_content_segmentation = gr.Checkbox(value=True, label="禁止内容分句入库",
# interactive=True)
# load_conent_button = gr.Button("添加内容并加载知识库")
# # 将上传的文件保存到content文件夹下,并更新下拉框
# vs_add.click(fn=add_vs_name,
# inputs=[vs_name, vs_list, chatbot],
# outputs=[select_vs, vs_list, vs_name, vs_add, file2vs, chatbot])
# select_vs.change(fn=change_vs_name_input,
# inputs=[select_vs, chatbot],
# outputs=[vs_name, vs_add, file2vs, vs_path, chatbot])
# load_file_button.click(get_vector_store,
# show_progress=True,
# inputs=[select_vs, files, sentence_size, chatbot, vs_add, vs_add],
# outputs=[vs_path, files, chatbot], )
# load_folder_button.click(get_vector_store,
# show_progress=True,
# inputs=[select_vs, folder_files, sentence_size, chatbot, vs_add,
# vs_add],
# outputs=[vs_path, folder_files, chatbot], )
# load_conent_button.click(get_vector_store,
# show_progress=True,
# inputs=[select_vs, one_title, sentence_size, chatbot,
# one_conent, one_content_segmentation],
# outputs=[vs_path, files, chatbot], )
# flag_csv_logger.setup([query, vs_path, chatbot, mode], "flagged")
# query.submit(get_answer,
# [query, vs_path, chatbot, mode, score_threshold, vector_search_top_k, chunk_conent,
# chunk_sizes],
# [chatbot, query])
# with gr.Tab("模型配置"):
# llm_model = gr.Radio(llm_model_dict_list,
# label="LLM 模型",
# value=LLM_MODEL,
# interactive=True)
# llm_history_len = gr.Slider(0, 10,
# value=LLM_HISTORY_LEN,
# step=1,
# label="LLM 对话轮数",
# interactive=True)
# use_ptuning_v2 = gr.Checkbox(USE_PTUNING_V2,
# label="使用p-tuning-v2微调过的模型",
# interactive=True)
# use_lora = gr.Checkbox(USE_LORA,
# label="使用lora微调的权重",
# interactive=True)
# embedding_model = gr.Radio(embedding_model_dict_list,
# label="Embedding 模型",
# value=EMBEDDING_MODEL,
# interactive=True)
# top_k = gr.Slider(1, 20, value=VECTOR_SEARCH_TOP_K, step=1,
# label="向量匹配 top k", interactive=True)
# load_model_button = gr.Button("重新加载模型")
# load_model_button.click(reinit_model, show_progress=True,
# inputs=[llm_model, embedding_model, llm_history_len, use_ptuning_v2, use_lora,
# top_k, chatbot], outputs=chatbot)
(demo
.queue(concurrency_count=3)
.launch(server_name='0.0.0.0',
server_port=7860,
show_api=False,
share=True,
inbrowser=False))