Spaces:
Sleeping
Sleeping
File size: 7,402 Bytes
a387a9a |
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 |
from abc import ABCMeta, abstractmethod
from functional import *
class ChoiceStrategy(metaclass=ABCMeta):
def __init__(self, choice):
self.choice = choice
self.delta = choice['delta']
@abstractmethod
def support(self):
pass
@abstractmethod
def execute(self, bot_backend: BotBackend, history: List, whether_exit: bool):
pass
class RoleChoiceStrategy(ChoiceStrategy):
def support(self):
return 'role' in self.delta
def execute(self, bot_backend: BotBackend, history: List, whether_exit: bool):
bot_backend.set_assistant_role_name(assistant_role_name=self.delta['role'])
return history, whether_exit
class ContentChoiceStrategy(ChoiceStrategy):
def support(self):
return 'content' in self.delta and self.delta['content'] is not None
def execute(self, bot_backend: BotBackend, history: List, whether_exit: bool):
# null value of content often occur in function call:
# {
# "role": "assistant",
# "content": null,
# "function_call": {
# "name": "python",
# "arguments": ""
# }
# }
bot_backend.add_content(content=self.delta.get('content', ''))
history[-1][1] = bot_backend.content
return history, whether_exit
class NameFunctionCallChoiceStrategy(ChoiceStrategy):
def support(self):
return 'function_call' in self.delta and 'name' in self.delta['function_call']
def execute(self, bot_backend: BotBackend, history: List, whether_exit: bool):
function_dict = bot_backend.jupyter_kernel.available_functions
bot_backend.set_function_name(function_name=self.delta['function_call']['name'])
bot_backend.copy_current_bot_history(bot_history=history)
if bot_backend.function_name not in function_dict:
history.append(
[
None,
f'GPT attempted to call a function that does '
f'not exist: {bot_backend.function_name}\n '
]
)
whether_exit = True
return history, whether_exit
class ArgumentsFunctionCallChoiceStrategy(ChoiceStrategy):
def support(self):
return 'function_call' in self.delta and 'arguments' in self.delta['function_call']
def execute(self, bot_backend: BotBackend, history: List, whether_exit: bool):
bot_backend.add_function_args_str(function_args_str=self.delta['function_call']['arguments'])
if bot_backend.function_name == 'python': # handle hallucinatory function calls
'''
In practice, we have noticed that GPT, especially GPT-3.5, may occasionally produce hallucinatory
function calls. These calls involve a non-existent function named `python` with arguments consisting
solely of raw code text (not a JSON format).
'''
temp_code_str = bot_backend.function_args_str
bot_backend.update_display_code_block(
display_code_block="\n🔴Working:\n```python\n{}\n```".format(temp_code_str)
)
history = copy.deepcopy(bot_backend.bot_history)
history[-1][1] += bot_backend.display_code_block
else:
temp_code_str = parse_json(function_args=bot_backend.function_args_str, finished=False)
if temp_code_str is not None:
bot_backend.update_display_code_block(
display_code_block="\n🔴Working:\n```python\n{}\n```".format(
temp_code_str
)
)
history = copy.deepcopy(bot_backend.bot_history)
history[-1][1] += bot_backend.display_code_block
return history, whether_exit
class FinishReasonChoiceStrategy(ChoiceStrategy):
def support(self):
return self.choice['finish_reason'] is not None
def execute(self, bot_backend: BotBackend, history: List, whether_exit: bool):
function_dict = bot_backend.jupyter_kernel.available_functions
if bot_backend.content:
bot_backend.add_gpt_response_content_message()
bot_backend.update_finish_reason(finish_reason=self.choice['finish_reason'])
if bot_backend.finish_reason == 'function_call':
try:
code_str = self.get_code_str(bot_backend)
bot_backend.update_display_code_block(
display_code_block="\n🟢Working:\n```python\n{}\n```".format(code_str)
)
history = copy.deepcopy(bot_backend.bot_history)
history[-1][1] += bot_backend.display_code_block
# function response
text_to_gpt, content_to_display = function_dict[
bot_backend.function_name
](code_str)
# add function call to conversion
bot_backend.add_function_call_response_message(function_response=text_to_gpt, save_tokens=True)
add_function_response_to_bot_history(
content_to_display=content_to_display, history=history, unique_id=bot_backend.unique_id
)
except json.JSONDecodeError:
history.append(
[None, f"GPT generate wrong function args: {bot_backend.function_args_str}"]
)
whether_exit = True
return history, whether_exit
except Exception as e:
history.append([None, f'Backend error: {e}'])
whether_exit = True
return history, whether_exit
bot_backend.reset_gpt_response_log_values(exclude=['finish_reason'])
return history, whether_exit
@staticmethod
def get_code_str(bot_backend):
if bot_backend.function_name == 'python':
code_str = bot_backend.function_args_str
else:
code_str = parse_json(function_args=bot_backend.function_args_str, finished=True)
if code_str is None:
raise json.JSONDecodeError
return code_str
class ChoiceHandler:
strategies = [
RoleChoiceStrategy, ContentChoiceStrategy, NameFunctionCallChoiceStrategy,
ArgumentsFunctionCallChoiceStrategy, FinishReasonChoiceStrategy
]
def __init__(self, choice):
self.choice = choice
def handle(self, bot_backend: BotBackend, history: List, whether_exit: bool):
for Strategy in self.strategies:
strategy_instance = Strategy(choice=self.choice)
if not strategy_instance.support():
continue
history, whether_exit = strategy_instance.execute(
bot_backend=bot_backend,
history=history,
whether_exit=whether_exit
)
return history, whether_exit
def parse_response(chunk, history, bot_backend: BotBackend):
"""
:return: history, whether_exit
"""
whether_exit = False
if chunk['choices']:
choice = chunk['choices'][0]
choice_handler = ChoiceHandler(choice=choice)
history, whether_exit = choice_handler.handle(
history=history,
bot_backend=bot_backend,
whether_exit=whether_exit
)
return history, whether_exit
|