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
|
import logging
import os
from typing import Any
from ollama import AsyncClient, Client
from core.base import (
ChunkSearchResult,
EmbeddingConfig,
EmbeddingProvider,
EmbeddingPurpose,
R2RException,
)
logger = logging.getLogger()
class OllamaEmbeddingProvider(EmbeddingProvider):
def __init__(self, config: EmbeddingConfig):
super().__init__(config)
provider = config.provider
if not provider:
raise ValueError(
"Must set provider in order to initialize `OllamaEmbeddingProvider`."
)
if provider != "ollama":
raise ValueError(
"OllamaEmbeddingProvider must be initialized with provider `ollama`."
)
if config.rerank_model:
raise ValueError(
"OllamaEmbeddingProvider does not support separate reranking."
)
self.base_model = config.base_model
self.base_dimension = config.base_dimension
self.base_url = os.getenv("OLLAMA_API_BASE")
logger.info(
f"Using Ollama API base URL: {self.base_url or 'http://127.0.0.1:11434'}"
)
self.client = Client(host=self.base_url)
self.aclient = AsyncClient(host=self.base_url)
self.set_prefixes(config.prefixes or {}, self.base_model)
self.batch_size = config.batch_size or 32
def _get_embedding_kwargs(self, **kwargs):
embedding_kwargs = {
"model": self.base_model,
}
embedding_kwargs.update(kwargs)
return embedding_kwargs
async def _execute_task(self, task: dict[str, Any]) -> list[list[float]]:
texts = task["texts"]
purpose = task.get("purpose", EmbeddingPurpose.INDEX)
kwargs = self._get_embedding_kwargs(**task.get("kwargs", {}))
try:
embeddings = []
for i in range(0, len(texts), self.batch_size):
batch = texts[i : i + self.batch_size]
prefixed_batch = [
self.prefixes.get(purpose, "") + text for text in batch
]
response = await self.aclient.embed(
input=prefixed_batch, **kwargs
)
embeddings.extend(response["embeddings"])
return embeddings
except Exception as e:
error_msg = f"Error getting embeddings: {str(e)}"
logger.error(error_msg)
raise R2RException(error_msg, 400) from e
def _execute_task_sync(self, task: dict[str, Any]) -> list[list[float]]:
texts = task["texts"]
purpose = task.get("purpose", EmbeddingPurpose.INDEX)
kwargs = self._get_embedding_kwargs(**task.get("kwargs", {}))
try:
embeddings = []
for i in range(0, len(texts), self.batch_size):
batch = texts[i : i + self.batch_size]
prefixed_batch = [
self.prefixes.get(purpose, "") + text for text in batch
]
response = self.client.embed(input=prefixed_batch, **kwargs)
embeddings.extend(response["embeddings"])
return embeddings
except Exception as e:
error_msg = f"Error getting embeddings: {str(e)}"
logger.error(error_msg)
raise R2RException(error_msg, 400) from e
async def async_get_embedding(
self,
text: str,
stage: EmbeddingProvider.Step = EmbeddingProvider.Step.BASE,
purpose: EmbeddingPurpose = EmbeddingPurpose.INDEX,
**kwargs,
) -> list[float]:
if stage != EmbeddingProvider.Step.BASE:
raise ValueError(
"OllamaEmbeddingProvider only supports search stage."
)
task = {
"texts": [text],
"stage": stage,
"purpose": purpose,
"kwargs": kwargs,
}
result = await self._execute_with_backoff_async(task)
return result[0]
def get_embedding(
self,
text: str,
stage: EmbeddingProvider.Step = EmbeddingProvider.Step.BASE,
purpose: EmbeddingPurpose = EmbeddingPurpose.INDEX,
**kwargs,
) -> list[float]:
if stage != EmbeddingProvider.Step.BASE:
raise ValueError(
"OllamaEmbeddingProvider only supports search stage."
)
task = {
"texts": [text],
"stage": stage,
"purpose": purpose,
"kwargs": kwargs,
}
result = self._execute_with_backoff_sync(task)
return result[0]
async def async_get_embeddings(
self,
texts: list[str],
stage: EmbeddingProvider.Step = EmbeddingProvider.Step.BASE,
purpose: EmbeddingPurpose = EmbeddingPurpose.INDEX,
**kwargs,
) -> list[list[float]]:
if stage != EmbeddingProvider.Step.BASE:
raise ValueError(
"OllamaEmbeddingProvider only supports search stage."
)
task = {
"texts": texts,
"stage": stage,
"purpose": purpose,
"kwargs": kwargs,
}
return await self._execute_with_backoff_async(task)
def get_embeddings(
self,
texts: list[str],
stage: EmbeddingProvider.Step = EmbeddingProvider.Step.BASE,
purpose: EmbeddingPurpose = EmbeddingPurpose.INDEX,
**kwargs,
) -> list[list[float]]:
if stage != EmbeddingProvider.Step.BASE:
raise ValueError(
"OllamaEmbeddingProvider only supports search stage."
)
task = {
"texts": texts,
"stage": stage,
"purpose": purpose,
"kwargs": kwargs,
}
return self._execute_with_backoff_sync(task)
def rerank(
self,
query: str,
results: list[ChunkSearchResult],
stage: EmbeddingProvider.Step = EmbeddingProvider.Step.RERANK,
limit: int = 10,
) -> list[ChunkSearchResult]:
return results[:limit]
async def arerank(
self,
query: str,
results: list[ChunkSearchResult],
stage: EmbeddingProvider.Step = EmbeddingProvider.Step.RERANK,
limit: int = 10,
):
return results[:limit]
|
