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
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
|
from __future__ import annotations
import asyncio
import random
from ..typing import Type, List, CreateResult, Messages, Iterator
from .types import BaseProvider, BaseRetryProvider
from .. import debug
from ..errors import RetryProviderError, RetryNoProviderError
class RetryProvider(BaseRetryProvider):
def __init__(
self,
providers: List[Type[BaseProvider]],
shuffle: bool = True,
single_provider_retry: bool = False,
max_retries: int = 3,
) -> None:
"""
Initialize the BaseRetryProvider.
Args:
providers (List[Type[BaseProvider]]): List of providers to use.
shuffle (bool): Whether to shuffle the providers list.
single_provider_retry (bool): Whether to retry a single provider if it fails.
max_retries (int): Maximum number of retries for a single provider.
"""
self.providers = providers
self.shuffle = shuffle
self.single_provider_retry = single_provider_retry
self.max_retries = max_retries
self.working = True
self.last_provider: Type[BaseProvider] = None
def create_completion(
self,
model: str,
messages: Messages,
stream: bool = False,
**kwargs,
) -> CreateResult:
"""
Create a completion using available providers, with an option to stream the response.
Args:
model (str): The model to be used for completion.
messages (Messages): The messages to be used for generating completion.
stream (bool, optional): Flag to indicate if the response should be streamed. Defaults to False.
Yields:
CreateResult: Tokens or results from the completion.
Raises:
Exception: Any exception encountered during the completion process.
"""
providers = [p for p in self.providers if stream and p.supports_stream] if stream else self.providers
if self.shuffle:
random.shuffle(providers)
exceptions = {}
started: bool = False
if self.single_provider_retry and len(providers) == 1:
provider = providers[0]
self.last_provider = provider
for attempt in range(self.max_retries):
try:
if debug.logging:
print(f"Using {provider.__name__} provider (attempt {attempt + 1})")
for token in provider.create_completion(model, messages, stream, **kwargs):
yield token
started = True
if started:
return
except Exception as e:
exceptions[provider.__name__] = e
if debug.logging:
print(f"{provider.__name__}: {e.__class__.__name__}: {e}")
if started:
raise e
else:
for provider in providers:
self.last_provider = provider
try:
if debug.logging:
print(f"Using {provider.__name__} provider")
for token in provider.create_completion(model, messages, stream, **kwargs):
yield token
started = True
if started:
return
except Exception as e:
exceptions[provider.__name__] = e
if debug.logging:
print(f"{provider.__name__}: {e.__class__.__name__}: {e}")
if started:
raise e
raise_exceptions(exceptions)
async def create_async(
self,
model: str,
messages: Messages,
**kwargs,
) -> str:
"""
Asynchronously create a completion using available providers.
Args:
model (str): The model to be used for completion.
messages (Messages): The messages to be used for generating completion.
Returns:
str: The result of the asynchronous completion.
Raises:
Exception: Any exception encountered during the asynchronous completion process.
"""
providers = self.providers
if self.shuffle:
random.shuffle(providers)
exceptions = {}
if self.single_provider_retry and len(providers) == 1:
provider = providers[0]
self.last_provider = provider
for attempt in range(self.max_retries):
try:
if debug.logging:
print(f"Using {provider.__name__} provider (attempt {attempt + 1})")
return await asyncio.wait_for(
provider.create_async(model, messages, **kwargs),
timeout=kwargs.get("timeout", 60),
)
except Exception as e:
exceptions[provider.__name__] = e
if debug.logging:
print(f"{provider.__name__}: {e.__class__.__name__}: {e}")
else:
for provider in providers:
self.last_provider = provider
try:
if debug.logging:
print(f"Using {provider.__name__} provider")
return await asyncio.wait_for(
provider.create_async(model, messages, **kwargs),
timeout=kwargs.get("timeout", 60),
)
except Exception as e:
exceptions[provider.__name__] = e
if debug.logging:
print(f"{provider.__name__}: {e.__class__.__name__}: {e}")
raise_exceptions(exceptions)
class IterProvider(BaseRetryProvider):
__name__ = "IterProvider"
def __init__(
self,
providers: List[BaseProvider],
) -> None:
providers.reverse()
self.providers: List[BaseProvider] = providers
self.working: bool = True
self.last_provider: BaseProvider = None
def create_completion(
self,
model: str,
messages: Messages,
stream: bool = False,
**kwargs
) -> CreateResult:
exceptions: dict = {}
started: bool = False
for provider in self.iter_providers():
if stream and not provider.supports_stream:
continue
try:
for token in provider.create_completion(model, messages, stream, **kwargs):
yield token
started = True
if started:
return
except Exception as e:
exceptions[provider.__name__] = e
if debug.logging:
print(f"{provider.__name__}: {e.__class__.__name__}: {e}")
if started:
raise e
raise_exceptions(exceptions)
async def create_async(
self,
model: str,
messages: Messages,
**kwargs
) -> str:
exceptions: dict = {}
for provider in self.iter_providers():
try:
return await asyncio.wait_for(
provider.create_async(model, messages, **kwargs),
timeout=kwargs.get("timeout", 60)
)
except Exception as e:
exceptions[provider.__name__] = e
if debug.logging:
print(f"{provider.__name__}: {e.__class__.__name__}: {e}")
raise_exceptions(exceptions)
def iter_providers(self) -> Iterator[BaseProvider]:
used_provider = []
try:
while self.providers:
provider = self.providers.pop()
used_provider.append(provider)
self.last_provider = provider
if debug.logging:
print(f"Using {provider.__name__} provider")
yield provider
finally:
used_provider.reverse()
self.providers = [*used_provider, *self.providers]
def raise_exceptions(exceptions: dict) -> None:
"""
Raise a combined exception if any occurred during retries.
Raises:
RetryProviderError: If any provider encountered an exception.
RetryNoProviderError: If no provider is found.
"""
if exceptions:
raise RetryProviderError("RetryProvider failed:\n" + "\n".join([
f"{p}: {exception.__class__.__name__}: {exception}" for p, exception in exceptions.items()
]))
raise RetryNoProviderError("No provider found")
|
