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
|
import logging
import time
import uuid
from typing import Optional
from r2r.base import (
GenerationConfig,
KGSearchSettings,
KVLoggingSingleton,
R2RException,
RunManager,
VectorSearchSettings,
manage_run,
to_async_generator,
)
from r2r.pipes import EvalPipe
from r2r.telemetry.telemetry_decorator import telemetry_event
from ..abstractions import R2RPipelines, R2RProviders
from ..assembly.config import R2RConfig
from .base import Service
logger = logging.getLogger(__name__)
class RetrievalService(Service):
def __init__(
self,
config: R2RConfig,
providers: R2RProviders,
pipelines: R2RPipelines,
run_manager: RunManager,
logging_connection: KVLoggingSingleton,
):
super().__init__(
config, providers, pipelines, run_manager, logging_connection
)
@telemetry_event("Search")
async def search(
self,
query: str,
vector_search_settings: VectorSearchSettings = VectorSearchSettings(),
kg_search_settings: KGSearchSettings = KGSearchSettings(),
*args,
**kwargs,
):
async with manage_run(self.run_manager, "search_app") as run_id:
t0 = time.time()
if (
kg_search_settings.use_kg_search
and self.config.kg.provider is None
):
raise R2RException(
status_code=400,
message="Knowledge Graph search is not enabled in the configuration.",
)
if (
vector_search_settings.use_vector_search
and self.config.vector_database.provider is None
):
raise R2RException(
status_code=400,
message="Vector search is not enabled in the configuration.",
)
# TODO - Remove these transforms once we have a better way to handle this
for filter, value in vector_search_settings.search_filters.items():
if isinstance(value, uuid.UUID):
vector_search_settings.search_filters[filter] = str(value)
results = await self.pipelines.search_pipeline.run(
input=to_async_generator([query]),
vector_search_settings=vector_search_settings,
kg_search_settings=kg_search_settings,
run_manager=self.run_manager,
*args,
**kwargs,
)
t1 = time.time()
latency = f"{t1 - t0:.2f}"
await self.logging_connection.log(
log_id=run_id,
key="search_latency",
value=latency,
is_info_log=False,
)
return results.dict()
@telemetry_event("RAG")
async def rag(
self,
query: str,
rag_generation_config: GenerationConfig,
vector_search_settings: VectorSearchSettings = VectorSearchSettings(),
kg_search_settings: KGSearchSettings = KGSearchSettings(),
*args,
**kwargs,
):
async with manage_run(self.run_manager, "rag_app") as run_id:
try:
t0 = time.time()
# TODO - Remove these transforms once we have a better way to handle this
for (
filter,
value,
) in vector_search_settings.search_filters.items():
if isinstance(value, uuid.UUID):
vector_search_settings.search_filters[filter] = str(
value
)
if rag_generation_config.stream:
t1 = time.time()
latency = f"{t1 - t0:.2f}"
await self.logging_connection.log(
log_id=run_id,
key="rag_generation_latency",
value=latency,
is_info_log=False,
)
async def stream_response():
async with manage_run(self.run_manager, "arag"):
async for (
chunk
) in await self.pipelines.streaming_rag_pipeline.run(
input=to_async_generator([query]),
run_manager=self.run_manager,
vector_search_settings=vector_search_settings,
kg_search_settings=kg_search_settings,
rag_generation_config=rag_generation_config,
):
yield chunk
return stream_response()
results = await self.pipelines.rag_pipeline.run(
input=to_async_generator([query]),
run_manager=self.run_manager,
vector_search_settings=vector_search_settings,
kg_search_settings=kg_search_settings,
rag_generation_config=rag_generation_config,
*args,
**kwargs,
)
t1 = time.time()
latency = f"{t1 - t0:.2f}"
await self.logging_connection.log(
log_id=run_id,
key="rag_generation_latency",
value=latency,
is_info_log=False,
)
if len(results) == 0:
raise R2RException(
status_code=404, message="No results found"
)
if len(results) > 1:
logger.warning(
f"Multiple results found for query: {query}"
)
# unpack the first result
return results[0]
except Exception as e:
logger.error(f"Pipeline error: {str(e)}")
if "NoneType" in str(e):
raise R2RException(
status_code=502,
message="Ollama server not reachable or returned an invalid response",
)
raise R2RException(
status_code=500, message="Internal Server Error"
)
@telemetry_event("Evaluate")
async def evaluate(
self,
query: str,
context: str,
completion: str,
eval_generation_config: Optional[GenerationConfig],
*args,
**kwargs,
):
eval_payload = EvalPipe.EvalPayload(
query=query,
context=context,
completion=completion,
)
result = await self.eval_pipeline.run(
input=to_async_generator([eval_payload]),
run_manager=self.run_manager,
eval_generation_config=eval_generation_config,
)
return result
|
