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
|
from __future__ import annotations
import json
import asyncio
from aiohttp import ClientSession, ContentTypeError
from ..typing import AsyncResult, Messages
from .base_provider import AsyncGeneratorProvider, ProviderModelMixin
from .helper import format_prompt
from ..image import ImageResponse
class ReplicateHome(AsyncGeneratorProvider, ProviderModelMixin):
url = "https://replicate.com"
api_endpoint = "https://homepage.replicate.com/api/prediction"
working = True
supports_stream = True
supports_system_message = True
supports_message_history = True
default_model = 'yorickvp/llava-13b'
image_models = [
'stability-ai/stable-diffusion-3',
'bytedance/sdxl-lightning-4step',
'playgroundai/playground-v2.5-1024px-aesthetic',
]
text_models = [
'meta/meta-llama-3-70b-instruct',
'mistralai/mixtral-8x7b-instruct-v0.1',
'google-deepmind/gemma-2b-it',
'yorickvp/llava-13b',
]
models = text_models + image_models
model_aliases = {
# image_models
"sd-3": "stability-ai/stable-diffusion-3",
"sdxl": "bytedance/sdxl-lightning-4step",
"playground-v2.5": "playgroundai/playground-v2.5-1024px-aesthetic",
# text_models
"gemma-2b": "google-deepmind/gemma-2b-it",
"llava-13b": "yorickvp/llava-13b",
}
model_versions = {
# image_models
'stability-ai/stable-diffusion-3': "527d2a6296facb8e47ba1eaf17f142c240c19a30894f437feee9b91cc29d8e4f",
'bytedance/sdxl-lightning-4step': "5f24084160c9089501c1b3545d9be3c27883ae2239b6f412990e82d4a6210f8f",
'playgroundai/playground-v2.5-1024px-aesthetic': "a45f82a1382bed5c7aeb861dac7c7d191b0fdf74d8d57c4a0e6ed7d4d0bf7d24",
# text_models
"google-deepmind/gemma-2b-it": "dff94eaf770e1fc211e425a50b51baa8e4cac6c39ef074681f9e39d778773626",
"yorickvp/llava-13b": "80537f9eead1a5bfa72d5ac6ea6414379be41d4d4f6679fd776e9535d1eb58bb",
}
@classmethod
def get_model(cls, model: str) -> str:
if model in cls.models:
return model
elif model in cls.model_aliases:
return cls.model_aliases[model]
else:
return cls.default_model
@classmethod
async def create_async_generator(
cls,
model: str,
messages: Messages,
proxy: str = None,
**kwargs
) -> AsyncResult:
model = cls.get_model(model)
headers = {
"accept": "*/*",
"accept-language": "en-US,en;q=0.9",
"cache-control": "no-cache",
"content-type": "application/json",
"origin": "https://replicate.com",
"pragma": "no-cache",
"priority": "u=1, i",
"referer": "https://replicate.com/",
"sec-ch-ua": '"Not;A=Brand";v="24", "Chromium";v="128"',
"sec-ch-ua-mobile": "?0",
"sec-ch-ua-platform": '"Linux"',
"sec-fetch-dest": "empty",
"sec-fetch-mode": "cors",
"sec-fetch-site": "same-site",
"user-agent": "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/128.0.0.0 Safari/537.36"
}
async with ClientSession(headers=headers) as session:
if model in cls.image_models:
prompt = messages[-1]['content'] if messages else ""
else:
prompt = format_prompt(messages)
data = {
"model": model,
"version": cls.model_versions[model],
"input": {"prompt": prompt},
}
async with session.post(cls.api_endpoint, json=data, proxy=proxy) as response:
response.raise_for_status()
result = await response.json()
prediction_id = result['id']
poll_url = f"https://homepage.replicate.com/api/poll?id={prediction_id}"
max_attempts = 30
delay = 5
for _ in range(max_attempts):
async with session.get(poll_url, proxy=proxy) as response:
response.raise_for_status()
try:
result = await response.json()
except ContentTypeError:
text = await response.text()
try:
result = json.loads(text)
except json.JSONDecodeError:
raise ValueError(f"Unexpected response format: {text}")
if result['status'] == 'succeeded':
if model in cls.image_models:
image_url = result['output'][0]
yield ImageResponse(image_url, "Generated image")
return
else:
for chunk in result['output']:
yield chunk
break
elif result['status'] == 'failed':
raise Exception(f"Prediction failed: {result.get('error')}")
await asyncio.sleep(delay)
if result['status'] != 'succeeded':
raise Exception("Prediction timed out")
|
