forefront

This commit is contained in:
t.me/xtekky 2023-04-30 22:09:25 +01:00
parent 08abfe0ebb
commit b4aadbbac5
3 changed files with 232 additions and 0 deletions

View File

@ -0,0 +1,13 @@
### Example: `forefront` (use like openai pypi package) <a name="example-forefront"></a>
```python
from gpt4free import forefront
# create an account
token = forefront.Account.create(logging=False)
print(token)
# get a response
for response in forefront.StreamingCompletion.create(token=token,
prompt='hello world', model='gpt-4'):
print(response.completion.choices[0].text, end='')
print("")
```

View File

@ -0,0 +1,194 @@
from json import loads
from re import findall
from time import time, sleep
from typing import Generator, Optional
from uuid import uuid4
from fake_useragent import UserAgent
from requests import post
from pymailtm import MailTm, Message
from tls_client import Session
from .typing import ForeFrontResponse
class Account:
@staticmethod
def create(proxy: Optional[str] = None, logging: bool = False):
proxies = {'http': 'http://' + proxy, 'https': 'http://' + proxy} if proxy else False
start = time()
mail_client = MailTm().get_account()
mail_address = mail_client.address
client = Session(client_identifier='chrome110')
client.proxies = proxies
client.headers = {
'origin': 'https://accounts.forefront.ai',
'user-agent': UserAgent().random,
}
response = client.post(
'https://clerk.forefront.ai/v1/client/sign_ups?_clerk_js_version=4.38.4',
data={'email_address': mail_address},
)
try:
trace_token = response.json()['response']['id']
if logging:
print(trace_token)
except KeyError:
return 'Failed to create account!'
response = client.post(
f'https://clerk.forefront.ai/v1/client/sign_ups/{trace_token}/prepare_verification?_clerk_js_version=4.38.4',
data={
'strategy': 'email_link',
'redirect_url': 'https://accounts.forefront.ai/sign-up/verify'
},
)
if logging:
print(response.text)
if 'sign_up_attempt' not in response.text:
return 'Failed to create account!'
while True:
sleep(1)
new_message: Message = mail_client.wait_for_message()
if logging:
print(new_message.data['id'])
verification_url = findall(r'https:\/\/clerk\.forefront\.ai\/v1\/verify\?token=\w.+', new_message.text)[0]
if verification_url:
break
if logging:
print(verification_url)
response = client.get(verification_url)
response = client.get('https://clerk.forefront.ai/v1/client?_clerk_js_version=4.38.4')
token = response.json()['response']['sessions'][0]['last_active_token']['jwt']
with open('accounts.txt', 'a') as f:
f.write(f'{mail_address}:{token}\n')
if logging:
print(time() - start)
return token
class StreamingCompletion:
@staticmethod
def create(
token=None,
chat_id=None,
prompt='',
action_type='new',
default_persona='607e41fe-95be-497e-8e97-010a59b2e2c0', # default
model='gpt-4',
proxy=None
) -> Generator[ForeFrontResponse, None, None]:
if not token:
raise Exception('Token is required!')
if not chat_id:
chat_id = str(uuid4())
proxies = { 'http': 'http://' + proxy, 'https': 'http://' + proxy } if proxy else None
headers = {
'authority': 'chat-server.tenant-forefront-default.knative.chi.coreweave.com',
'accept': '*/*',
'accept-language': 'en,fr-FR;q=0.9,fr;q=0.8,es-ES;q=0.7,es;q=0.6,en-US;q=0.5,am;q=0.4,de;q=0.3',
'authorization': 'Bearer ' + token,
'cache-control': 'no-cache',
'content-type': 'application/json',
'origin': 'https://chat.forefront.ai',
'pragma': 'no-cache',
'referer': 'https://chat.forefront.ai/',
'sec-ch-ua': '"Chromium";v="112", "Google Chrome";v="112", "Not:A-Brand";v="99"',
'sec-ch-ua-mobile': '?0',
'sec-ch-ua-platform': '"macOS"',
'sec-fetch-dest': 'empty',
'sec-fetch-mode': 'cors',
'sec-fetch-site': 'cross-site',
'user-agent': UserAgent().random,
}
json_data = {
'text': prompt,
'action': action_type,
'parentId': chat_id,
'workspaceId': chat_id,
'messagePersona': default_persona,
'model': model,
}
for chunk in post(
'https://chat-server.tenant-forefront-default.knative.chi.coreweave.com/chat',
headers=headers,
proxies=proxies,
json=json_data,
stream=True,
).iter_lines():
if b'finish_reason":null' in chunk:
data = loads(chunk.decode('utf-8').split('data: ')[1])
token = data['choices'][0]['delta'].get('content')
if token is not None:
yield ForeFrontResponse(
**{
'id': chat_id,
'object': 'text_completion',
'created': int(time()),
'text': token,
'model': model,
'choices': [{'text': token, 'index': 0, 'logprobs': None, 'finish_reason': 'stop'}],
'usage': {
'prompt_tokens': len(prompt),
'completion_tokens': len(token),
'total_tokens': len(prompt) + len(token),
},
}
)
class Completion:
@staticmethod
def create(
token=None,
chat_id=None,
prompt='',
action_type='new',
default_persona='607e41fe-95be-497e-8e97-010a59b2e2c0', # default
model='gpt-4',
proxy=None
) -> ForeFrontResponse:
text = ''
final_response = None
for response in StreamingCompletion.create(
token=token,
chat_id=chat_id,
prompt=prompt,
action_type=action_type,
default_persona=default_persona,
model=model,
proxy=proxy
):
if response:
final_response = response
text += response.text
if final_response:
final_response.text = text
else:
raise Exception('Unable to get the response, Please try again')
return final_response

View File

@ -0,0 +1,25 @@
from typing import Any, List
from pydantic import BaseModel
class Choice(BaseModel):
text: str
index: int
logprobs: Any
finish_reason: str
class Usage(BaseModel):
prompt_tokens: int
completion_tokens: int
total_tokens: int
class ForeFrontResponse(BaseModel):
id: str
object: str
created: int
model: str
choices: List[Choice]
usage: Usage
text: str