gpt4free/etc/tool/translate_readme.py

88 lines
2.2 KiB
Python
Raw Normal View History

import sys
from pathlib import Path
import asyncio
sys.path.append(str(Path(__file__).parent.parent.parent))
import g4f
g4f.debug.logging = True
from g4f.debug import access_token
provider = g4f.Provider.OpenaiChat
iso = "GE"
language = "german"
translate_prompt = f"""
Translate this markdown document to {language}.
Don't translate or change inline code examples.
```md
"""
keep_note = "Keep this: [!Note] as [!Note].\n"
blacklist = [
'## ©️ Copyright',
'## 🚀 Providers and Models',
'## 🔗 Related GPT4Free Projects'
]
whitelist = [
"### Other",
"### Models"
]
def read_text(text):
start = end = 0
new = text.strip().split('\n')
for i, line in enumerate(new):
if line.startswith('```'):
if not start:
start = i + 1
end = i
return '\n'.join(new[start:end]).strip()
async def translate(text):
prompt = translate_prompt + text.strip() + '\n```'
if "[!Note]" in text:
prompt = keep_note + prompt
result = read_text(await provider.create_async(
model="",
messages=[{"role": "user", "content": prompt}],
access_token=access_token
))
if text.endswith("```") and not result.endswith("```"):
result += "\n```"
return result
async def translate_part(part, i):
blacklisted = False
for headline in blacklist:
if headline in part:
blacklisted = True
if blacklisted:
lines = part.split('\n')
lines[0] = await translate(lines[0])
part = '\n'.join(lines)
for trans in whitelist:
if trans in part:
part = part.replace(trans, await translate(trans))
else:
part = await translate(part)
print(f"[{i}] translated")
return part
async def translate_readme(readme) -> str:
parts = readme.split('\n## ')
print(f"{len(parts)} parts...")
parts = await asyncio.gather(
*[translate_part("## " + part, i) for i, part in enumerate(parts)]
)
return "\n\n".join(parts)
with open("README.md", "r") as fp:
readme = fp.read()
print("Translate readme...")
readme = asyncio.run(translate_readme(readme))
file = f"README-{iso}.md"
with open(file, "w") as fp:
fp.write(readme)
print(f'"{file}" saved')