forked from emcf/engshell
-
Notifications
You must be signed in to change notification settings - Fork 0
/
engshell.py
205 lines (190 loc) · 8.55 KB
/
engshell.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
import openai
import time
from colorama import Fore, Style
import os
import sys
from prompts import *
from keys import *
import subprocess
import io
import contextlib
import platform
import traceback
openai.api_key = OPENAI_KEY
MAX_PROMPT = 20480
CONTEXT_LEFT, CONTEXT_RIGHT = '{', '}'
engshell_PREVIX = lambda: Style.RESET_ALL + os.getcwd() + ' ' + Style.RESET_ALL + Fore.MAGENTA + "engshell" + Style.RESET_ALL + '>'
API_CALLS_PER_MIN = 50
VERBOSE = False
MAX_DEBUG_ATTEMPTS = 3
RETRY_ERRORS = ["The server had an error while processing your request. Sorry about that!"]
memory = []
def print_console_prompt():
print(engshell_PREVIX(), end="")
def print_status(status):
print_console_prompt()
print(Style.RESET_ALL + Fore.YELLOW + status + Style.RESET_ALL)
def print_success(status):
print_console_prompt()
print(Style.RESET_ALL + Fore.GREEN + status + Style.RESET_ALL)
def print_err(status):
print_console_prompt()
print(Style.RESET_ALL + Fore.RED + status + Style.RESET_ALL)
def print_code(status):
print_console_prompt()
print(Style.RESET_ALL + Fore.LIGHTBLACK_EX + status + Style.RESET_ALL)
def clean_code_string(response_content):
lines = response_content.split("\n")
for i, line in enumerate(lines):
if line.startswith("!pip"):
lines.pop(i)
response_content = "\n".join(lines)
split_response_content = response_content.split('```')
if len(split_response_content) > 1:
response_content = split_response_content[1]
for code_languge in ['python', 'bash']:
if response_content[:len(code_languge)]==code_languge: response_content = response_content[len(code_languge)+1:] # remove python+newline blocks
return response_content.replace('`','')
def clean_install_string(response_content):
split_response_content = response_content.split('`')
if len(split_response_content) > 1:
response_content = split_response_content[1]
return response_content.replace('`','').replace('!', '')
def summarize(text):
summarized = text
raise NotImplementedError("summarize(text) not yet implemented")
return summarized
def LLM(prompt, mode='text', gpt4 = False):
global memory
if len(prompt) > MAX_PROMPT:
print_status('prompt too large, summarizing...')
prompt = summarize(prompt)
time.sleep(1.0/API_CALLS_PER_MIN)
moderation_resp = openai.Moderation.create(input=prompt)
if moderation_resp.results[0].flagged:
raise ValueError(f'prompt ({prompt}) flagged by moderation endpoint')
time.sleep(1.0/API_CALLS_PER_MIN)
if mode == 'text':
messages=[
{"role": "system", "content": LLM_SYSTEM_CALIBRATION_MESSAGE},
{"role": "user", "content": prompt},
]
elif mode == 'code':
messages=memory
elif mode == 'debug':
messages=[
{"role": "system", "content": DEBUG_SYSTEM_CALIBRATION_MESSAGE},
{"role": "user", "content": prompt},
]
elif mode == 'install':
messages=[
{"role": "system", "content": INSTALL_SYSTEM_CALIBRATION_MESSAGE},
{"role": "user", "content": prompt},
]
response = openai.ChatCompletion.create(
model="gpt-4" if gpt4 else "gpt-3.5-turbo-0301",
messages=messages,
temperature = 0.0
)
response_content = response.choices[0].message.content
if mode == 'code' or mode == 'debug': response_content = clean_code_string(response_content)
elif mode == 'install': response_content = clean_install_string(response_content)
return response_content
def containerize_code(code_string):
code_string = code_string.replace('your_openai_api_key_here', OPENAI_KEY)
# uncomment this if you wish to easily use photos from Unsplash API
# code_string = code_string.replace('your_unsplash_access_key_here', UNSPLASH_ACCESS_KEY)
try:
output_buffer = io.StringIO()
with contextlib.redirect_stdout(output_buffer):
exec(code_string, globals())
except Exception as e:
exc_type, exc_value, exc_traceback = sys.exc_info()
tb = traceback.extract_tb(exc_traceback)
filename, line, func, text = tb[-1]
error_msg = f"{exc_type.__name__}: {str(e)}"
return False, f'Error: {error_msg}. Getting the error from function: {func} (line: {line})'
code_printout = output_buffer.getvalue()
return True, code_printout
def run_python(returned_code, debug = False, showcode = False, gpt4 = False):
print_status("compiling...")
if showcode:
print(returned_code, end = '' if returned_code[-1] == '\n' else '\n')
print_status("running...")
returned_code = clean_code_string(returned_code)
success, output = containerize_code(returned_code)
attempts = 0
should_debug = debug and (attempts < MAX_DEBUG_ATTEMPTS) and (not success)
should_install = (output is not None) and ('No module named' in output or 'ModuleNotFoundError' in output)
should_retry = should_debug or should_install or ((output is not None) and any([(err in output) for err in RETRY_ERRORS]))
while should_retry:
if should_install:
print_status('installing: ' + output)
prompt = INSTALL_USER_MESSAGE(output)
returned_command = LLM(prompt, mode='install', gpt4 = gpt4)
os.system(returned_command)
elif should_debug:
print_status('debugging: ' + output)
prompt = DEBUG_MESSAGE(returned_code, output)
returned_code = LLM(prompt, mode='debug', gpt4 = gpt4)
print_status('rerunning...')
returned_code = clean_code_string(returned_code)
success, output = containerize_code(returned_code)
attempts += 1
should_debug = debug and (attempts < MAX_DEBUG_ATTEMPTS) and (not success)
should_retry = should_debug or any([(err in output) for err in RETRY_ERRORS])
if not success:
raise ValueError(f"failed ({output})")
return output
def clear_memory():
global memory
memory = [
{"role": "system", "content": CODE_SYSTEM_CALIBRATION_MESSAGE},
{"role": "user", "content": CODE_USER_CALIBRATION_MESSAGE},
{"role": "assistant", "content": CODE_ASSISTANT_CALIBRATION_MESSAGE},
{"role": "system", "content": CONSOLE_OUTPUT_CALIBRATION_MESSAGE},
# uncomment these if you wish to easily use photos from Unsplash API
#{"role": "user", "content": CODE_USER_CALIBRATION_MESSAGE3},
#{"role": "assistant", "content": CODE_ASSISTANT_CALIBRATION_MESSAGE3},
#{"role": "system", "content": CONSOLE_OUTPUT_CALIBRATION_MESSAGE3},
]
if __name__ == "__main__":
if os.name == 'nt': os.system('')
always_showcode = '--showcode' in sys.argv
always_gpt4 = '--gpt4' in sys.argv
always_debug = '--debug' in sys.argv
always_llm = '--llm' in sys.argv
clear_memory()
while user_input := input(engshell_PREVIX()):
if user_input == 'clear':
clear_memory()
os.system("cls" if platform.system() == "Windows" else "clear")
continue
if ('--llm' in user_input) or always_llm: user_input += CONGNITIVE_USER_MESSAGE
debug = ('--debug' in user_input) or always_debug
showcode = ('--showcode' in user_input) or always_showcode
gpt4 = ('--gpt4' in user_input) or always_gpt4
user_input = user_input.replace('--llm','')
user_input = user_input.replace('--debug','')
user_input = user_input.replace('--showcode','')
user_input = user_input.replace('--gpt4','')
user_prompt = USER_MESSAGE(user_input, current_dir = os.getcwd())
memory.append({"role": "user", "content": user_prompt})
run_code = True
while run_code:
returned_code = LLM(user_prompt, mode='code', gpt4 = gpt4)
memory.append({"role": "assistant", "content": returned_code})
try:
console_output = run_python(returned_code, debug, showcode, gpt4)
#if len(console_output) > MAX_PROMPT:
# print_status('output too large, summarizing...')
# console_output = summarize(console_output)
if console_output.strip() == '': console_output = 'done executing.'
print_success(console_output)
run_code = False
except Exception as e:
error_message = str(e)
console_output = error_message
run_code = any([err in error_message for err in RETRY_ERRORS])
if len(console_output) < MAX_PROMPT:
memory.append({"role": "system", "content": console_output})