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
|
import asyncio
import logging
import random
import time
from abc import abstractmethod
from enum import Enum
from typing import Any, Optional
from litellm import AuthenticationError
from core.base.abstractions import VectorQuantizationSettings
from ..abstractions import (
ChunkSearchResult,
EmbeddingPurpose,
default_embedding_prefixes,
)
from .base import Provider, ProviderConfig
logger = logging.getLogger()
class EmbeddingConfig(ProviderConfig):
provider: str
base_model: str
base_dimension: int | float
rerank_model: Optional[str] = None
rerank_url: Optional[str] = None
batch_size: int = 1
prefixes: Optional[dict[str, str]] = None
add_title_as_prefix: bool = True
concurrent_request_limit: int = 256
max_retries: int = 3
initial_backoff: float = 1
max_backoff: float = 64.0
quantization_settings: VectorQuantizationSettings = (
VectorQuantizationSettings()
)
## deprecated
rerank_dimension: Optional[int] = None
rerank_transformer_type: Optional[str] = None
def validate_config(self) -> None:
if self.provider not in self.supported_providers:
raise ValueError(f"Provider '{self.provider}' is not supported.")
@property
def supported_providers(self) -> list[str]:
return ["litellm", "openai", "ollama"]
class EmbeddingProvider(Provider):
class Step(Enum):
BASE = 1
RERANK = 2
def __init__(self, config: EmbeddingConfig):
if not isinstance(config, EmbeddingConfig):
raise ValueError(
"EmbeddingProvider must be initialized with a `EmbeddingConfig`."
)
logger.info(f"Initializing EmbeddingProvider with config {config}.")
super().__init__(config)
self.config: EmbeddingConfig = config
self.semaphore = asyncio.Semaphore(config.concurrent_request_limit)
self.current_requests = 0
async def _execute_with_backoff_async(self, task: dict[str, Any]):
retries = 0
backoff = self.config.initial_backoff
while retries < self.config.max_retries:
try:
async with self.semaphore:
return await self._execute_task(task)
except AuthenticationError:
raise
except Exception as e:
logger.warning(
f"Request failed (attempt {retries + 1}): {str(e)}"
)
retries += 1
if retries == self.config.max_retries:
raise
await asyncio.sleep(random.uniform(0, backoff))
backoff = min(backoff * 2, self.config.max_backoff)
def _execute_with_backoff_sync(self, task: dict[str, Any]):
retries = 0
backoff = self.config.initial_backoff
while retries < self.config.max_retries:
try:
return self._execute_task_sync(task)
except AuthenticationError:
raise
except Exception as e:
logger.warning(
f"Request failed (attempt {retries + 1}): {str(e)}"
)
retries += 1
if retries == self.config.max_retries:
raise
time.sleep(random.uniform(0, backoff))
backoff = min(backoff * 2, self.config.max_backoff)
@abstractmethod
async def _execute_task(self, task: dict[str, Any]):
pass
@abstractmethod
def _execute_task_sync(self, task: dict[str, Any]):
pass
async def async_get_embedding(
self,
text: str,
stage: Step = Step.BASE,
purpose: EmbeddingPurpose = EmbeddingPurpose.INDEX,
):
task = {
"text": text,
"stage": stage,
"purpose": purpose,
}
return await self._execute_with_backoff_async(task)
def get_embedding(
self,
text: str,
stage: Step = Step.BASE,
purpose: EmbeddingPurpose = EmbeddingPurpose.INDEX,
):
task = {
"text": text,
"stage": stage,
"purpose": purpose,
}
return self._execute_with_backoff_sync(task)
async def async_get_embeddings(
self,
texts: list[str],
stage: Step = Step.BASE,
purpose: EmbeddingPurpose = EmbeddingPurpose.INDEX,
):
task = {
"texts": texts,
"stage": stage,
"purpose": purpose,
}
return await self._execute_with_backoff_async(task)
def get_embeddings(
self,
texts: list[str],
stage: Step = Step.BASE,
purpose: EmbeddingPurpose = EmbeddingPurpose.INDEX,
) -> list[list[float]]:
task = {
"texts": texts,
"stage": stage,
"purpose": purpose,
}
return self._execute_with_backoff_sync(task)
@abstractmethod
def rerank(
self,
query: str,
results: list[ChunkSearchResult],
stage: Step = Step.RERANK,
limit: int = 10,
):
pass
@abstractmethod
async def arerank(
self,
query: str,
results: list[ChunkSearchResult],
stage: Step = Step.RERANK,
limit: int = 10,
):
pass
def set_prefixes(self, config_prefixes: dict[str, str], base_model: str):
self.prefixes = {}
for t, p in config_prefixes.items():
purpose = EmbeddingPurpose(t.lower())
self.prefixes[purpose] = p
if base_model in default_embedding_prefixes:
for t, p in default_embedding_prefixes[base_model].items():
if t not in self.prefixes:
self.prefixes[t] = p
|
