-
Notifications
You must be signed in to change notification settings - Fork 13.5k
/
ecloud_vector_search.py
580 lines (496 loc) Β· 20.1 KB
/
ecloud_vector_search.py
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
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
import logging
import uuid
from typing import (
TYPE_CHECKING,
Any,
Callable,
Dict,
Iterable,
List,
Optional,
Tuple,
Union,
)
from langchain_core.documents import Document
from langchain_core.embeddings import Embeddings
from langchain_core.vectorstores import VectorStore
if TYPE_CHECKING:
from elasticsearch import Elasticsearch
logger = logging.getLogger(__name__)
class EcloudESVectorStore(VectorStore):
"""`ecloud Elasticsearch` vector store.
Example:
.. code-block:: python
from langchain.vectorstores import EcloudESVectorStore
from langchain.embeddings.openai import OpenAIEmbeddings
embeddings = OpenAIEmbeddings()
vectorstore = EcloudESVectorStore(
embedding=OpenAIEmbeddings(),
index_name="langchain-demo",
es_url="http://localhost:9200"
)
Args:
index_name: Name of the Elasticsearch index to create.
es_url: URL of the ecloud Elasticsearch instance to connect to.
user: Username to use when connecting to Elasticsearch.
password: Password to use when connecting to Elasticsearch.
"""
def __init__(
self,
index_name: str,
es_url: str,
user: Optional[str] = None,
password: Optional[str] = None,
embedding: Optional[Embeddings] = None,
**kwargs: Optional[dict],
) -> None:
self.embedding = embedding
self.index_name = index_name
self.text_field = kwargs.get("text_field", "text")
self.vector_field = kwargs.get("vector_field", "vector")
self.vector_type = kwargs.get("vector_type", "knn_dense_float_vector")
self.vector_params = kwargs.get("vector_params") or {}
self.model = self.vector_params.get("model", "")
self.index_settings = kwargs.get("index_settings") or {}
key_list = [
"text_field",
"vector_field",
"vector_type",
"vector_params",
"index_settings",
]
[kwargs.pop(key, None) for key in key_list]
if es_url is not None:
self.client = EcloudESVectorStore.es_client(
es_url=es_url, username=user, password=password, **kwargs
)
else:
raise ValueError("""Please specified a es connection url.""")
@property
def embeddings(self) -> Optional[Embeddings]:
return self.embedding
@staticmethod
def es_client(
*,
es_url: Optional[str] = None,
username: Optional[str] = None,
password: Optional[str] = None,
**kwargs: Optional[dict],
) -> "Elasticsearch":
try:
import elasticsearch
except ImportError:
raise ImportError(
"Could not import elasticsearch python package. "
"Please install it with `pip install elasticsearch`."
)
connection_params: Dict[str, Any] = {"hosts": [es_url]}
if username and password:
connection_params["http_auth"] = (username, password)
connection_params.update(kwargs)
es_client = elasticsearch.Elasticsearch(**connection_params)
try:
es_client.info()
except Exception as e:
logger.error(f"Error connecting to Elasticsearch: {e}")
raise e
return es_client
def _create_index_if_not_exists(self, dims_length: Optional[int] = None) -> None:
"""Create the index if it doesn't already exist.
Args:
dims_length: Length of the embedding vectors.
"""
if self.client.indices.exists(index=self.index_name):
logger.info(f"Index {self.index_name} already exists. Skipping creation.")
else:
if dims_length is None:
raise ValueError(
"Cannot create index without specifying dims_length "
+ "when the index doesn't already exist. "
)
indexMapping = self._index_mapping(dims_length=dims_length)
logger.debug(
f"Creating index {self.index_name} with mappings {indexMapping}"
)
self.client.indices.create(
index=self.index_name,
body={
"settings": {"index.knn": True, **self.index_settings},
"mappings": {"properties": indexMapping},
},
)
def _index_mapping(self, dims_length: Union[int, None]) -> Dict:
"""
Executes when the index is created.
Args:
dims_length: Numeric length of the embedding vectors,
or None if not using vector-based query.
index_params: The extra pamameters for creating index.
Returns:
Dict: The Elasticsearch settings and mappings for the strategy.
"""
model = self.vector_params.get("model", "")
if "lsh" == model:
mapping: Dict[Any, Any] = {
self.vector_field: {
"type": self.vector_type,
"knn": {
"dims": dims_length,
"model": "lsh",
"similarity": self.vector_params.get("similarity", "cosine"),
"L": self.vector_params.get("L", 99),
"k": self.vector_params.get("k", 1),
},
}
}
if mapping[self.vector_field]["knn"]["similarity"] == "l2":
mapping[self.vector_field]["knn"]["w"] = self.vector_params.get("w", 3)
return mapping
elif "permutation_lsh" == model:
return {
self.vector_field: {
"type": self.vector_type,
"knn": {
"dims": dims_length,
"model": "permutation_lsh",
"k": self.vector_params.get("k", 10),
"similarity": self.vector_params.get("similarity", "cosine"),
"repeating": self.vector_params.get("repeating", True),
},
}
}
else:
return {
self.vector_field: {
"type": self.vector_type,
"knn": {"dims": dims_length},
}
}
def delete(
self,
ids: Optional[List[str]] = None,
**kwargs: Any,
) -> Optional[bool]:
"""Delete documents from the index.
Args:
ids: List of ids of documents to delete
"""
try:
from elasticsearch.helpers import BulkIndexError, bulk
except ImportError:
raise ImportError(
"Could not import elasticsearch python package. "
"Please install it with `pip install elasticsearch`."
)
body = []
if ids is None:
raise ValueError("ids must be provided.")
for _id in ids:
body.append({"_op_type": "delete", "_index": self.index_name, "_id": _id})
if len(body) > 0:
try:
bulk(
self.client,
body,
refresh=kwargs.get("refresh_indices", True),
ignore_status=404,
)
logger.debug(f"Deleted {len(body)} texts from index")
return True
except BulkIndexError as e:
logger.error(f"Error deleting texts: {e}")
raise e
else:
logger.info("No documents to delete")
return False
def _query_body(
self,
query_vector: Union[List[float], None],
filter: Optional[dict] = None,
search_params: Dict = {},
) -> Dict:
query_vector_body = {
"field": search_params.get("vector_field", self.vector_field)
}
if self.vector_type == "knn_dense_float_vector":
query_vector_body["vec"] = {"values": query_vector}
specific_params = self.get_dense_specific_model_similarity_params(
search_params
)
query_vector_body.update(specific_params)
else:
query_vector_body["vec"] = {
"true_indices": query_vector,
"total_indices": len(query_vector) if query_vector is not None else 0,
}
specific_params = self.get_sparse_specific_model_similarity_params(
search_params
)
query_vector_body.update(specific_params)
query_vector_body = {"knn_nearest_neighbors": query_vector_body}
if filter is not None and len(filter) != 0:
query_vector_body = {
"function_score": {"query": filter, "functions": [query_vector_body]}
}
return {
"size": search_params.get("size", 4),
"query": query_vector_body,
}
@staticmethod
def get_dense_specific_model_similarity_params(
search_params: Dict[str, Any],
) -> Dict:
model = search_params.get("model", "exact")
similarity = search_params.get("similarity", "cosine")
specific_params = {"model": model, "similarity": similarity}
if not model == "exact":
if model not in ("lsh", "permutation_lsh"):
raise ValueError(
f"vector type knn_dense_float_vector doesn't support model {model}"
)
if similarity not in ("cosine", "l2"):
raise ValueError(f"model exact doesn't support similarity {similarity}")
specific_params["candidates"] = search_params.get(
"candidates", search_params.get("size", 4)
)
if model == "lsh" and similarity == "l2":
specific_params["probes"] = search_params.get("probes", 0)
else:
if similarity not in ("cosine", "l2"):
raise ValueError(f"model exact don't support similarity {similarity}")
return specific_params
@staticmethod
def get_sparse_specific_model_similarity_params(
search_params: Dict[str, Any],
) -> Dict:
model = search_params.get("model", "exact")
similarity = search_params.get("similarity", "hamming")
specific_params = {"model": model, "similarity": similarity}
if not model == "exact":
if model not in ("lsh",):
raise ValueError(
f"vector type knn_dense_float_vector doesn't support model {model}"
)
if similarity not in ("hamming", "jaccard"):
raise ValueError(f"model exact doesn't support similarity {similarity}")
specific_params["candidates"] = search_params.get(
"candidates", search_params.get("size", 4)
)
else:
if similarity not in ("hamming", "jaccard"):
raise ValueError(f"model exact don't support similarity {similarity}")
return specific_params
def _search(
self,
query: Optional[str] = None,
query_vector: Union[List[float], None] = None,
filter: Optional[dict] = None,
custom_query: Optional[Callable[[Dict, Union[str, None]], Dict]] = None,
search_params: Dict = {},
) -> List[Tuple[Document, float]]:
"""Return searched documents result from ecloud ES
Args:
query: Text to look up documents similar to.
query_vector: Embedding to look up documents similar to.
filter: Array of ecloud ElasticSearch filter clauses to apply to the query.
custom_query: Function to modify the query body before it is sent to ES.
Returns:
List of Documents most similar to the query and score for each
"""
if self.embedding and query is not None:
query_vector = self.embedding.embed_query(query)
query_body = self._query_body(
query_vector=query_vector, filter=filter, search_params=search_params
)
if custom_query is not None:
query_body = custom_query(query_body, query)
logger.debug(f"Calling custom_query, Query body now: {query_body}")
logger.debug(f"Query body: {query_body}")
# Perform the kNN search on the ES index and return the results.
response = self.client.search(index=self.index_name, body=query_body)
logger.debug(f"response={response}")
hits = [hit for hit in response["hits"]["hits"]]
docs_and_scores = [
(
Document(
page_content=hit["_source"][
search_params.get("text_field", self.text_field)
],
metadata=hit["_source"]["metadata"],
),
hit["_score"],
)
for hit in hits
]
return docs_and_scores
def similarity_search(
self,
query: str,
k: int = 4,
filter: Optional[dict] = None,
**kwargs: Any,
) -> List[Document]:
"""Return documents most similar to query.
Args:
query: Text to look up documents similar to.
k: Number of Documents to return. Defaults to 4.
filter: Array of Elasticsearch filter clauses to apply to the query.
Returns:
List of Documents most similar to the query,
in descending order of similarity.
"""
results = self.similarity_search_with_score(
query=query, k=k, filter=filter, **kwargs
)
return [doc for doc, _ in results]
def similarity_search_with_score(
self, query: str, k: int, filter: Optional[dict] = None, **kwargs: Any
) -> List[Tuple[Document, float]]:
"""Return documents most similar to query, along with scores.
Args:
query: Text to look up documents similar to.
size: Number of Documents to return. Defaults to 4.
filter: Array of Elasticsearch filter clauses to apply to the query.
Returns:
List of Documents most similar to the query and score for each
"""
search_params: Dict[str, Any] = kwargs.get("search_params") or {}
if len(search_params) == 0:
kwargs = {"search_params": {"size": k}}
elif search_params.get("size") is None:
search_params["size"] = k
kwargs["search_params"] = search_params
return self._search(query=query, filter=filter, **kwargs)
@classmethod
def from_documents(
cls,
documents: List[Document],
embedding: Optional[Embeddings] = None,
**kwargs: Any,
) -> "EcloudESVectorStore":
"""Construct EcloudESVectorStore wrapper from documents.
Args:
documents: List of documents to add to the Elasticsearch index.
embedding: Embedding function to use to embed the texts.
Do not provide if using a strategy
that doesn't require inference.
kwargs: create index key words arguments
"""
vectorStore = EcloudESVectorStore._es_vector_store(
embedding=embedding, **kwargs
)
# Encode the provided texts and add them to the newly created index.
vectorStore.add_documents(documents)
return vectorStore
@classmethod
def from_texts(
cls,
texts: List[str],
embedding: Optional[Embeddings] = None,
metadatas: Optional[List[Dict[str, Any]]] = None,
**kwargs: Any,
) -> "EcloudESVectorStore":
"""Construct EcloudESVectorStore wrapper from raw documents.
Args:
texts: List of texts to add to the Elasticsearch index.
embedding: Embedding function to use to embed the texts.
metadatas: Optional list of metadatas associated with the texts.
index_name: Name of the Elasticsearch index to create.
kwargs: create index key words arguments
"""
vectorStore = cls._es_vector_store(embedding=embedding, **kwargs)
# Encode the provided texts and add them to the newly created index.
vectorStore.add_texts(texts, metadatas=metadatas, **kwargs)
return vectorStore
def add_texts(
self,
texts: Iterable[str],
metadatas: Optional[List[Dict[Any, Any]]] = None,
**kwargs: Any,
) -> List[str]:
"""Run more texts through the embeddings and add to the vectorstore.
Args:
texts: Iterable of strings to add to the vectorstore.
metadatas: Optional list of metadatas associated with the texts.
Returns:
List of ids from adding the texts into the vectorstore.
"""
try:
from elasticsearch.helpers import BulkIndexError, bulk
except ImportError:
raise ImportError(
"Could not import elasticsearch python package. "
"Please install it with `pip install elasticsearch`."
)
embeddings = []
create_index_if_not_exists = kwargs.get("create_index_if_not_exists", True)
ids = kwargs.get("ids", [str(uuid.uuid4()) for _ in texts])
refresh_indices = kwargs.get("refresh_indices", False)
requests = []
if self.embedding is not None:
embeddings = self.embedding.embed_documents(list(texts))
dims_length = len(embeddings[0])
if create_index_if_not_exists:
self._create_index_if_not_exists(dims_length=dims_length)
for i, (text, vector) in enumerate(zip(texts, embeddings)):
metadata = metadatas[i] if metadatas else {}
doc = {
"_op_type": "index",
"_index": self.index_name,
self.text_field: text,
"metadata": metadata,
"_id": ids[i],
}
if self.vector_type == "knn_dense_float_vector":
doc[self.vector_field] = vector
elif self.vector_type == "knn_sparse_bool_vector":
doc[self.vector_field] = {
"true_indices": vector,
"total_indices": len(vector),
}
requests.append(doc)
else:
if create_index_if_not_exists:
self._create_index_if_not_exists()
for i, text in enumerate(texts):
metadata = metadatas[i] if metadatas else {}
requests.append(
{
"_op_type": "index",
"_index": self.index_name,
self.text_field: text,
"metadata": metadata,
"_id": ids[i],
}
)
if len(requests) > 0:
try:
success, failed = bulk(
self.client, requests, stats_only=True, refresh=refresh_indices
)
logger.debug(
f"Added {success} and failed to add {failed} texts to index"
)
logger.debug(f"added texts {ids} to index")
if refresh_indices:
self.client.indices.refresh(index=self.index_name)
return ids
except BulkIndexError as e:
logger.error(f"Error adding texts: {e}")
firstError = e.errors[0].get("index", {}).get("error", {})
logger.error(f"First error reason: {firstError.get('reason')}")
raise e
else:
logger.debug("No texts to add to index")
return []
@staticmethod
def _es_vector_store(
embedding: Optional[Embeddings] = None, **kwargs: Any
) -> "EcloudESVectorStore":
index_name = kwargs.get("index_name")
if index_name is None:
raise ValueError("Please provide an index_name.")
es_url = kwargs.get("es_url")
if es_url is None:
raise ValueError("Please provided a valid es connection url")
return EcloudESVectorStore(embedding=embedding, **kwargs)