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 | class AWSDocDbVectorStore(BasePydanticVectorStore):
"""AWS DocumentDB向量存储。
要使用，您应该同时安装：
- ``pymongo`` python包
- 与DocumentDB实例关联的连接字符串
有关更多详细信息，请参考官方的向量搜索文档：
https://docs.aws.amazon.com/documentdb/latest/developerguide/vector-search.html"""
stores_text: bool = True
flat_metadata: bool = True
_docdb_client: MongoClient = PrivateAttr()
_similarity_score: AWSDocDbVectorStoreSimilarityType = PrivateAttr()
_collection: Any = PrivateAttr()
_embedding_key: str = PrivateAttr()
_id_key: str = PrivateAttr()
_text_key: str = PrivateAttr()
_metadata_key: str = PrivateAttr()
_insert_kwargs: Dict = PrivateAttr()
_index_crud: DocDbIndex = PrivateAttr()
def __init__(
self,
docdb_client: Optional[Any] = None,
db_name: str = "default_db",
index_name: str = "default_index",
collection_name: str = "default_collection",
id_key: str = "id",
embedding_key: str = "embedding",
text_key: str = "text",
metadata_key: str = "metadata",
insert_kwargs: Optional[Dict] = None,
similarity_score="cosine",
**kwargs: Any,
) -> None:
"""初始化向量存储。
Args:
docdb_client：DocumentDB 客户端。
db_name：DocumentDB 数据库名称。
collection_name：DocumentDB 集合名称。
id_key：用作 id 的数据字段。
embedding_key：将包含每个文档嵌入的 DocumentDB 字段。
text_key：将包含每个文档文本的 DocumentDB 字段。
metadata_key：将包含每个文档元数据的 DocumentDB 字段。
insert_kwargs：`insert` 过程中使用的 kwargs。
"""
super().__init__()
if docdb_client is not None:
self._docdb_client = cast(MongoClient, docdb_client)
else:
raise ValueError("Must specify connection string to DocumentDB instance ")
self._similarity_score = similarity_score
self._collection = self._docdb_client[db_name][collection_name]
self._embedding_key = embedding_key
self._id_key = id_key
self._text_key = text_key
self._metadata_key = metadata_key
self._insert_kwargs = insert_kwargs or {}
self._index_crud = DocDbIndex(index_name, self._embedding_key, self._collection)
@classmethod
def class_name(cls) -> str:
return "AWSDocDbVectorStore"
def add(
self,
nodes: List[BaseNode],
**add_kwargs: Any,
) -> List[str]:
"""将节点添加到索引中。
Args:
节点：List[BaseNode]：带有嵌入的节点列表
Returns:
成功添加节点的id列表。
"""
ids = []
data_to_insert = []
for node in nodes:
metadata = node_to_metadata_dict(
node, remove_text=True, flat_metadata=self.flat_metadata
)
entry = {
self._id_key: node.node_id,
self._embedding_key: node.get_embedding(),
self._text_key: node.get_content(metadata_mode=MetadataMode.NONE) or "",
self._metadata_key: metadata,
}
data_to_insert.append(entry)
ids.append(node.node_id)
logger.debug("Inserting data into DocumentDB: %s", data_to_insert)
insert_result = self._collection.insert_many(
data_to_insert, **self._insert_kwargs
)
logger.debug("Result of insert: %s", insert_result)
return ids
def delete(self, ref_doc_id: str, **delete_kwargs: Any) -> None:
"""使用id删除节点。
Args:
ref_doc_id（str）：要删除的文档的doc_id。
"""
if ref_doc_id is None:
raise ValueError("No document id provided to delete.")
self._collection.delete_one({self._metadata_key + ".ref_doc_id": ref_doc_id})
@property
def client(self) -> Any:
"""返回 DocDB 客户端。"""
return self._docdb_client
def _query(
self, query: VectorStoreQuery, projection: Optional[Dict[str, int]] = None
) -> VectorStoreQueryResult:
params: Dict[str, Any] = {
"vector": query.query_embedding,
"path": self._embedding_key,
"similarity": self._similarity_score,
"k": query.similarity_top_k,
}
if query.filters:
params["filter"] = _to_mongodb_filter(query.filters)
if projection is None:
pipeline = [{"$search": {"vectorSearch": params}}]
else:
pipeline = [{"$search": {"vectorSearch": params}}, {"$project": projection}]
logger.debug("Running query pipeline: %s", pipeline)
cursor = self._collection.aggregate(pipeline) # type: ignore
top_k_nodes = []
top_k_ids = []
top_k_scores = []
for res in cursor:
text = res.pop(self._text_key)
vector = res.pop(self._embedding_key)
id = res.pop(self._id_key)
metadata_dict = res.pop(self._metadata_key)
score = similarity(query.query_embedding, vector, self._similarity_score)
try:
node = metadata_dict_to_node(metadata_dict)
node.set_content(text)
except Exception:
# NOTE: deprecated legacy logic for backward compatibility
metadata, node_info, relationships = legacy_metadata_dict_to_node(
metadata_dict
)
node = TextNode(
text=text,
id_=id,
metadata=metadata,
start_char_idx=node_info.get("start", None),
end_char_idx=node_info.get("end", None),
relationships=relationships,
)
top_k_ids.append(id)
top_k_nodes.append(node)
top_k_scores.append(score)
result = VectorStoreQueryResult(
nodes=top_k_nodes, similarities=top_k_scores, ids=top_k_ids
)
logger.debug("Result of query: %s", result)
return result
def query(
self,
query: VectorStoreQuery,
projection: Optional[Dict[str, int]] = None,
**kwargs: Any,
) -> VectorStoreQueryResult:
"""查询前k个最相似节点的索引。
Args:
query: 一个VectorStoreQuery对象。
projection: 一个指定搜索后返回哪些字段的字典。
Returns:
包含查询结果的VectorStoreQueryResult。
"""
return self._query(query, projection=projection)
def create_index(self, dimensions, similarity_score=None):
score = self._similarity_score
if similarity_score is not None:
score = similarity
return self._index_crud.create_index(dimensions, score)
def delete_index(self):
return self._index_crud.delete_index()
def __del__(self) -> None:
self.docdb_client.close()
|