text stringlengths 1 93.6k |
|---|
import openai
|
import getpass
|
## Support
|
def parse_args():
|
parser = argparse.ArgumentParser(description='Bashir Command Line Arguments')
|
parser.add_argument('--os', default='Windows10', help='Operating system; eg "Windows10"')
|
parser.add_argument('--shell', default='cmd', help='Shell name; eg "cmd" or "PowerShell"')
|
args = parser.parse_args()
|
return args.os, args.shell
|
def save_script(content, comment=None):
|
fn = sdir + '\\' + str(uuid.uuid4()) + '.cmd'
|
comment_section = '' if comment is None else 'REM ' + comment + '\n'
|
content = '@echo off\n\n' + comment_section + content
|
with open(fn, 'w+') as f:
|
f.write(content)
|
return fn
|
## Body
|
# Context Settings: can be any OS and command line interpreter
|
operating_system, cli = parse_args()
|
# Confirmation state
|
CONFIRM_NONE, CONFIRM_ALL, CONFIRM_SUDO = 0, 1, 2
|
confirmation = CONFIRM_NONE
|
# Config langchain chat model
|
openai.api_key = os.getenv("OPENAI_API_KEY")
|
chat = ChatOpenAI(model_name="gpt-3.5-turbo",temperature=1.24)
|
system_message = SystemMessage(content=f"You are an advanced coding translator. You take natural language and convert whatever the user explains they want to do as a sequence of {cli} commands for {operating_system}. If a user asks how much space they have left on their hard drive, you output the proper commands needed to run in order to see that as a final output. Do not include explanations, only the translation into cmd commands. Always use command arguments that bypass user confirmation.")
|
# Prepare scripts folder
|
sdir = os.path.join(os.getcwd(), 'scripts')
|
if not os.path.exists(sdir):
|
os.makedirs(sdir)
|
print('\n--------------------------------------------\nBashir: LLM based Junior System Admin v0.1.0\n--------------------------------------------\n')
|
print('OS:', operating_system, '\nCLI:', cli, '\n')
|
import pexpect
|
import time
|
from pexpect.popen_spawn import PopenSpawn
|
# Spawn a new command prompt
|
child = PopenSpawn('cmd', encoding='utf-8')
|
def_prompt = 'Prompt> '
|
while True:
|
try:
|
print(def_prompt, end='', flush=True)
|
# Get command
|
command = input()
|
# If the user types 'exit', then end the loop and close
|
if command.strip() == 'exit':
|
break
|
# Perform call to LLM and get a bash script in return
|
response = chat([system_message, HumanMessage(content=command)])
|
bash_script = response.content
|
# Save the bash script
|
script_path = save_script(bash_script, comment=def_prompt + command.strip())
|
# Send command to command prompt
|
child.sendline(script_path)
|
time.sleep(0.5)
|
# Expect the command prompt again or asking for sudo pass
|
try:
|
child.expect(def_prompt, timeout=10)
|
except pexpect.TIMEOUT:
|
print(child.before.strip())
|
continue
|
except pexpect.EOF:
|
print(child.before.strip())
|
break
|
# Print the output of the command
|
print(child.before.strip())
|
except Exception as e:
|
print(str(e))
|
# Close the command prompt
|
child.close()
|
# <FILESEP>
|
# -*- coding: utf-8 -*-
|
translation_dict = {
|
"zh_CN": {
|
("*", "XiaoJuBao-HonkaiStarRail@何以千奈的橘子"): "小橘包-星穹铁道@何以千奈的橘子",
|
("*", "Shader blend file path"): "Shader blend 文件路径",
|
("*", "Input shader blend file path"): "输入 Shader blend 文件路径",
|
("*", "Texture file path"): "贴图文件路径",
|
("*", "Input texture file path"): "输入贴图文件路径",
|
("*", "Character JSON Path"): "角色配置文件路径",
|
("*", "Input character JSON Path"): "输入角色配置文件路径",
|
("Operator", "Batch apply materials"): "批量添加材质",
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.