Spaces:
Sleeping
Sleeping
File size: 11,253 Bytes
57cf043 |
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 |
import os
import re
from logging import Logger
from typing import List, Union
from openai import OpenAI
from common.configuration import FilterChunks, LLMConfiguration, SummaryChunks
from components.nmd.aggregate_answers import preprocessed_chunks
class LLMChunkSearch:
def __init__(self, config: LLMConfiguration, prompt: str, logger: Logger):
self.config = config
self.logger = logger
self.prompt = prompt
self.pattern = r'\d+'
self.pattern_list = [
r'\[\d+\]',
r'Ответ: [1-9]',
r'Ответ [1-9]',
r'Ответ[1-9]',
r'Ответ:[1-9]',
r'Ответ: \[\d+\]',
]
# Initialize OpenAI client
if self.config.base_url is not None:
self.client = OpenAI(
base_url=self.config.base_url,
api_key=os.getenv(self.config.api_key_env)
)
else:
self.client = None
def llm_chunk_search(self, query: str, answer_chunks: SummaryChunks, prompt: str):
"""
Args:
query: User query
answer_chunks: Retrieved chunks to process
prompt: System prompt template
Returns:
Tuple containing processed chunks, LLM response, prompt used, and token count
"""
text_chunks = preprocessed_chunks(
answer_chunks, self.config.base_url, self.logger
)
self.logger.info('Searching LLM Chunks')
if self.client is None:
return (
text_chunks,
self.__postprocessing_answer_llm(answer_chunks),
prompt,
0
)
llm_prompt = prompt.format(query=query, answer=text_chunks)
for i in range(5):
try:
response = self.client.chat.completions.create(
model=self.config.model,
messages=[
{"role": "system", "content": prompt},
{"role": "user", "content": query}
],
temperature=self.config.temperature,
top_p=self.config.top_p,
frequency_penalty=self.config.frequency_penalty,
presence_penalty=self.config.presence_penalty,
seed=self.config.seed
)
answer_llm = response.choices[0].message.content
count_tokens = response.usage.total_tokens
self.logger.info(f'Answer LLM {answer_llm}')
# Process the response
if re.search('%%', answer_llm):
index = re.search('%%', answer_llm).span()[1]
answer_llm = answer_llm[index:]
if re.search('Конец ответа', answer_llm):
index = re.search('Конец ответа', answer_llm).span()[1]
answer_llm = answer_llm[:index]
return text_chunks, answer_llm, llm_prompt, count_tokens
except Exception as e:
self.logger.error(f"Attempt {i+1} failed: {str(e)}")
if i == 4:
self.logger.error("All attempts failed")
return (
text_chunks,
self.__postprocessing_answer_llm(answer_chunks),
llm_prompt,
0
)
@staticmethod
def __postprocessing_answer_llm(answer_chunks: Union[SummaryChunks, List]) -> str:
"""
Postprocess the answer chunks into a formatted string
Args:
answer_chunks: Chunks to process
Returns:
Formatted string response
"""
output_text = ''
if isinstance(answer_chunks, SummaryChunks):
if len(answer_chunks.doc_chunks) == 0:
# TODO: Протестировать как работает и исправить на уведомление о БД и ли
return 'БАЗА ДАННЫХ ПУСТА'
if answer_chunks.doc_chunks is not None:
doc = answer_chunks.doc_chunks[0]
output_text += f'Документ: [1]\n'
if doc.title != 'unknown':
output_text += f'Название документа: {doc.title}\n'
else:
output_text += f'Название документа: {doc.filename}\n'
for chunk in doc.chunks:
if len(chunk.other_info):
for i in chunk.other_info:
output_text += f'{i}'
else:
output_text += f'{chunk.text_answer}'
output_text += '\n\n'
else:
doc = answer_chunks.people_search[0]
output_text += (
f'Название документа: Информация о сотруднике {doc.person_name}\n'
)
if doc.organizatinal_structure is not None:
for organizatinal_structure in doc.organizatinal_structure:
output_text += '('
if organizatinal_structure.position != 'undefined':
output_text += (
f'Должность: {organizatinal_structure.position}\n'
)
if organizatinal_structure.leads is not None:
output_text += f'Руководит следующими сотрудниками:\n'
for lead in organizatinal_structure.leads:
if lead.person != "undefined":
output_text += f'{lead.person}\n'
if (
organizatinal_structure.subordinates.person_name
!= "undefined"
):
output_text += f'Руководителем {doc.person_name} является {organizatinal_structure.subordinates.person_name}\n'
output_text += ')'
if doc.business_processes is not None:
if len(doc.business_processes) >= 2:
output_text += f'Отвечает за Бизнес процессы:\n'
else:
output_text += f'Отвечает за Бизнес процесс: '
for process in doc.business_processes:
output_text += f'{process.processes_name}\n'
if doc.business_curator is not None:
output_text += 'Является Бизнес-куратором (РОКС НН):\n'
for curator in doc.business_curator:
output_text += f'{curator.company_name}'
if doc.groups is not None:
if len(doc.groups) >= 2:
output_text += 'Входит в состав групп:\n'
else:
output_text += 'Входит в состав группы:\n'
for group in doc.groups:
if 'Члены' in group.position_in_group:
output_text += f'{group.group_name}. Должность внутри группы: {group.position_in_group.replace("Члены", "Член")}\n'
else:
output_text += f'{group.group_name}. Должность внутри группы: {group.position_in_group}\n'
output_text += f'\\\n\n'
else:
if isinstance(answer_chunks[0], FilterChunks):
doc = answer_chunks[0]
output_text += f'Документ: [1]\n'
if doc.title != 'unknown':
output_text += f'Название документа: {doc.title}\n'
for chunk in doc.chunks:
if len(chunk.other_info):
for i in chunk.other_info:
output_text += f'{i}'
else:
output_text += f'{chunk.text_answer}'
output_text += '\n\n'
else:
doc = answer_chunks[0]
output_text += f'Информация о сотруднике {doc.person_name}\n'
if doc.organizatinal_structure is not None:
for organizatinal_structure in doc.organizatinal_structure:
output_text += (
f'Должность: {organizatinal_structure.position}\n'
)
if organizatinal_structure.leads is not None:
output_text += f'Руководит следующими сотрудниками:\n'
for lead in organizatinal_structure.leads:
if lead.person != "undefined":
output_text += f'{lead.person}\n'
if (
organizatinal_structure.subordinates.person_name
!= "undefined"
):
output_text += f'Руководителем {doc.person_name} является {organizatinal_structure.subordinates.person_name}\n'
if doc.business_processes is not None:
if len(doc.business_processes) >= 2:
output_text += f'Отвечает за Бизнес процессы:\n'
else:
output_text += f'Отвечает за Бизнес процесс: '
for process in doc.business_processes:
output_text += f'{process.processes_name}\n'
if doc.business_curator is not None:
output_text += 'Является Бизнес-куратором (РОКС НН):\n'
for curator in doc.business_curator:
output_text += f'{curator.company_name}'
if doc.groups is not None:
if len(doc.groups) >= 2:
output_text += 'Входит в состав групп:\n'
else:
output_text += 'Входит в состав группы:\n'
for group in doc.groups:
if 'Члены' in group.position_in_group:
output_text += f'{group.group_name}. Должность внутри группы: {group.position_in_group.replace("Члены", "Член")}\n'
else:
output_text += f'{group.group_name}. Должность внутри группы: {group.position_in_group}\n'
output_text += f'\\\n\n'
return output_text
|