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
|
from __future__ import annotations
import random
import json
from aiohttp import ClientSession, WSMsgType
from ..typing import AsyncResult, Messages
from .base_provider import AsyncGeneratorProvider
API_URL = "https://labs-api.perplexity.ai/socket.io/"
WS_URL = "wss://labs-api.perplexity.ai/socket.io/"
MODELS = ['pplx-7b-online', 'pplx-70b-online', 'pplx-7b-chat', 'pplx-70b-chat', 'mistral-7b-instruct',
'codellama-34b-instruct', 'llama-2-70b-chat', 'llava-7b-chat', 'mixtral-8x7b-instruct',
'mistral-medium', 'related']
DEFAULT_MODEL = MODELS[1]
MODEL_MAP = {
"mistralai/Mistral-7B-Instruct-v0.1": "mistral-7b-instruct",
"meta-llama/Llama-2-70b-chat-hf": "llama-2-70b-chat",
"mistralai/Mixtral-8x7B-Instruct-v0.1": "mixtral-8x7b-instruct",
}
class PerplexityLabs(AsyncGeneratorProvider):
url = "https://labs.perplexity.ai"
working = True
supports_gpt_35_turbo = True
@classmethod
async def create_async_generator(
cls,
model: str,
messages: Messages,
proxy: str = None,
**kwargs
) -> AsyncResult:
if not model:
model = DEFAULT_MODEL
elif model in MODEL_MAP:
model = MODEL_MAP[model]
elif model not in MODELS:
raise ValueError(f"Model is not supported: {model}")
headers = {
"User-Agent": "Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:121.0) Gecko/20100101 Firefox/121.0",
"Accept": "*/*",
"Accept-Language": "de,en-US;q=0.7,en;q=0.3",
"Accept-Encoding": "gzip, deflate, br",
"Origin": cls.url,
"Connection": "keep-alive",
"Referer": f"{cls.url}/",
"Sec-Fetch-Dest": "empty",
"Sec-Fetch-Mode": "cors",
"Sec-Fetch-Site": "same-site",
"TE": "trailers",
}
async with ClientSession(headers=headers) as session:
t = format(random.getrandbits(32), '08x')
async with session.get(
f"{API_URL}?EIO=4&transport=polling&t={t}",
proxy=proxy
) as response:
text = await response.text()
sid = json.loads(text[1:])['sid']
post_data = '40{"jwt":"anonymous-ask-user"}'
async with session.post(
f'{API_URL}?EIO=4&transport=polling&t={t}&sid={sid}',
data=post_data,
proxy=proxy
) as response:
assert await response.text() == 'OK'
async with session.ws_connect(f'{WS_URL}?EIO=4&transport=websocket&sid={sid}', autoping=False) as ws:
await ws.send_str('2probe')
assert(await ws.receive_str() == '3probe')
await ws.send_str('5')
assert(await ws.receive_str())
assert(await ws.receive_str() == '6')
message_data = {
'version': '2.2',
'source': 'default',
'model': model,
'messages': messages
}
await ws.send_str('42' + json.dumps(['perplexity_playground', message_data]))
last_message = 0
while True:
message = await ws.receive_str()
if message == '2':
await ws.send_str('3')
continue
try:
data = json.loads(message[2:])[1]
yield data["output"][last_message:]
last_message = len(data["output"])
if data["final"]:
break
except:
raise RuntimeError(f"Message: {message}")
|
