Spaces:
Sleeping
Sleeping
File size: 11,909 Bytes
6a422c8 |
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 |
import copy
import os
import re
import json
from config_utils import get_user_cfg_file
from modelscope_agent.prompt.prompt import (KNOWLEDGE_INTRODUCTION_PROMPT,
KNOWLEDGE_PROMPT, LengthConstraint,
PromptGenerator, build_raw_prompt)
from modelscope.utils.config import Config
DEFAULT_SYSTEM_TEMPLATE = """
# 工具
## 你拥有如下工具:
<tool_list>
## 当你需要调用工具时,请在你的回复中穿插如下的工具调用命令,可以根据需求调用零次或多次:
工具调用
Action: 工具的名称,必须是<tool_name_list>之一
Action Input: 工具的输入
Observation: <result>工具返回的结果</result>
Answer: 根据Observation总结本次工具调用返回的结果,如果结果中出现url,请不要展示出。
```
[链接](url)
```
# 指令
"""
DEFAULT_SYSTEM_TEMPLATE_WITHOUT_TOOL = """
# 指令
"""
DEFAULT_INSTRUCTION_TEMPLATE = ''
DEFAULT_USER_TEMPLATE = """(你正在扮演<role_name>,你可以使用工具:<tool_name_list><knowledge_note>)<file_names><user_input>"""
DEFAULT_USER_TEMPLATE_WITHOUT_TOOL = """(你正在扮演<role_name><knowledge_note>) <file_names><user_input>"""
DEFAULT_EXEC_TEMPLATE = """Observation: <result><exec_result></result>\nAnswer:"""
TOOL_DESC = (
'{name_for_model}: {name_for_human} API。 {description_for_model} 输入参数: {parameters}'
)
class CustomPromptGenerator(PromptGenerator):
def __init__(self,
system_template=DEFAULT_SYSTEM_TEMPLATE,
instruction_template=DEFAULT_INSTRUCTION_TEMPLATE,
user_template=DEFAULT_USER_TEMPLATE,
exec_template=DEFAULT_EXEC_TEMPLATE,
assistant_template='',
sep='\n\n',
llm=None,
length_constraint=LengthConstraint(),
**kwargs):
super().__init__(
system_template=system_template,
instruction_template=instruction_template,
user_template=user_template,
exec_template=exec_template,
assistant_template=assistant_template,
sep=sep,
llm=llm,
length_constraint=length_constraint)
# hack here for special prompt, such as add an addition round before user input
self.add_addition_round = kwargs.get('add_addition_round', False)
self.addition_assistant_reply = kwargs.get('addition_assistant_reply',
'')
builder_cfg_file = get_user_cfg_file(
uuid_str=kwargs.get('uuid_str', ''))
builder_cfg = Config.from_file(builder_cfg_file)
self.builder_cfg = builder_cfg
self.knowledge_file_name = kwargs.get('knowledge_file_name', '')
self.llm = llm
self.prompt_preprocessor = build_raw_prompt(llm.model_id)
self.length_constraint = length_constraint
self._parse_length_restriction()
def _parse_length_restriction(self):
constraint = self.llm.cfg.get('length_constraint', None)
# if isinstance(constraint, Config):
# constraint = constraint.to_dict()
self.length_constraint.update(constraint)
def _update_user_prompt_without_knowledge(self, task, tool_list, **kwargs):
if len(tool_list) > 0:
# user input
user_input = self.user_template.replace('<role_name>',
self.builder_cfg.name)
user_input = user_input.replace(
'<tool_name_list>',
','.join([tool.name for tool in tool_list]))
else:
self.user_template = DEFAULT_USER_TEMPLATE_WITHOUT_TOOL
user_input = self.user_template.replace('<user_input>', task)
user_input = user_input.replace('<role_name>',
self.builder_cfg.name)
user_input = user_input.replace('<user_input>', task)
if 'append_files' in kwargs:
append_files = kwargs.get('append_files', [])
if len(append_files) > 0:
file_names = ','.join(
[os.path.basename(path) for path in append_files])
user_input = user_input.replace('<file_names>',
f'[上传文件{file_names}]')
else:
user_input = user_input.replace('<file_names>', '')
else:
user_input = user_input.replace('<file_names>', '')
return user_input
def init_prompt(self, task, tool_list, knowledge_list, **kwargs):
if len(self.history) == 0:
self.history.append({
'role': 'system',
'content': 'You are a helpful assistant.'
})
if len(tool_list) > 0:
prompt = f'{self.system_template}\n{self.instruction_template}'
# get tool description str
tool_str = self.get_tool_str(tool_list)
prompt = prompt.replace('<tool_list>', tool_str)
tool_name_str = self.get_tool_name_str(tool_list)
prompt = prompt.replace('<tool_name_list>', tool_name_str)
else:
self.system_template = DEFAULT_SYSTEM_TEMPLATE_WITHOUT_TOOL
prompt = f'{self.system_template}\n{self.instruction_template}'
user_input = self._update_user_prompt_without_knowledge(
task, tool_list, **kwargs)
if len(knowledge_list) > 0:
user_input = user_input.replace('<knowledge_note>',
',请查看前面的知识库')
else:
user_input = user_input.replace('<knowledge_note>', '')
self.system_prompt = copy.deepcopy(prompt)
# build history
if self.add_addition_round:
self.history.append({
'role': 'user',
'content': self.system_prompt
})
self.history.append({
'role': 'assistant',
'content': self.addition_assistant_reply
})
self.history.append({'role': 'user', 'content': user_input})
self.history.append({
'role': 'assistant',
'content': self.assistant_template
})
else:
self.history.append({
'role': 'user',
'content': self.system_prompt + user_input
})
self.history.append({
'role': 'assistant',
'content': self.assistant_template
})
self.function_calls = self.get_function_list(tool_list)
else:
user_input = self._update_user_prompt_without_knowledge(
task, tool_list, **kwargs)
if len(knowledge_list) > 0:
user_input = user_input.replace('<knowledge_note>',
',请查看前面的知识库')
else:
user_input = user_input.replace('<knowledge_note>', '')
self.history.append({'role': 'user', 'content': user_input})
self.history.append({
'role': 'assistant',
'content': self.assistant_template
})
if len(knowledge_list) > 0:
knowledge_str = self.get_knowledge_str(
knowledge_list,
file_name=self.knowledge_file_name,
only_content=True)
self.update_knowledge_str(knowledge_str)
def update_knowledge_str(self, knowledge_str):
"""If knowledge base information was not used previously, it will be added;
if knowledge base information was previously used, it will be replaced.
Args:
knowledge_str (str): knowledge str generated by get_knowledge_str
"""
knowledge_introduction = KNOWLEDGE_INTRODUCTION_PROMPT.replace(
'<file_name>', self.knowledge_file_name)
if len(knowledge_str) > self.length_constraint.knowledge:
# todo: use tokenizer to constrain length
knowledge_str = knowledge_str[-self.length_constraint.knowledge:]
knowledge_str = f'{KNOWLEDGE_PROMPT}{self.sep}{knowledge_introduction}{self.sep}{knowledge_str}'
for i in range(0, len(self.history)):
if self.history[i]['role'] == 'user':
content: str = self.history[i]['content']
start_pos = content.find(f'{KNOWLEDGE_PROMPT}{self.sep}')
end_pos = content.rfind('\n\n# 工具\n\n')
if start_pos >= 0 and end_pos >= 0: # replace knowledge
self.history[i]['content'] = content[
0:start_pos] + knowledge_str + content[end_pos:]
break
elif start_pos < 0 and end_pos == 0: # add knowledge
self.history[i]['content'] = knowledge_str + content
break
else:
continue
def get_tool_str(self, tool_list):
tool_texts = []
for tool in tool_list:
tool_texts.append(
TOOL_DESC.format(
name_for_model=tool.name,
name_for_human=tool.name,
description_for_model=tool.description,
parameters=json.dumps(tool.parameters,
ensure_ascii=False)))
# + ' ' + FORMAT_DESC['json'])
tool_str = '\n\n'.join(tool_texts)
return tool_str
def get_tool_name_str(self, tool_list):
tool_name = []
for tool in tool_list:
tool_name.append(tool.name)
tool_name_str = json.dumps(tool_name, ensure_ascii=False)
return tool_name_str
def _generate(self, llm_result, exec_result: str):
"""
generate next round prompt based on previous llm_result and exec_result and update history
"""
if len(llm_result) != 0:
self.history[-1]['content'] += f'{llm_result}'
if len(exec_result) != 0:
# handle image markdown wrapper
image_markdown_re = re.compile(
pattern=r'!\[IMAGEGEN\]\(([\s\S]+)\)')
match = image_markdown_re.search(exec_result)
if match is not None:
exec_result = match.group(1).rstrip()
exec_result = self.exec_template.replace('<exec_result>',
str(exec_result))
self.history[-1]['content'] += exec_result
# generate plate prompt here
self.prompt = self.prompt_preprocessor(self.history)
return self.prompt
def parse_role_config(config: dict):
prompt = '你扮演AI-Agent,'
# concat prompt
if 'name' in config and config['name']:
prompt += ('你的名字是' + config['name'] + '。')
if 'description' in config and config['description']:
prompt += config['description']
prompt += '\n你具有下列具体功能:'
if 'instruction' in config and config['instruction']:
if isinstance(config['instruction'], list):
for ins in config['instruction']:
prompt += ins
prompt += ';'
elif isinstance(config['instruction'], str):
prompt += config['instruction']
if prompt[-1] == ';':
prompt = prompt[:-1]
prompt += '\n下面你将开始扮演'
if 'name' in config and config['name']:
prompt += config['name']
prompt += ',明白了请说“好的。”,不要说其他的。'
return prompt
|