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 | class FirestoreVectorStore(BasePydanticVectorStore):
"""Firestore 向量存储。"""
stores_text: bool = True
flat_metadata: bool = True
collection_name: str
batch_size: Optional[int] = DEFAULT_BATCH_SIZE
embedding_key: str = "embedding"
text_key: str = "text"
metadata_key: str = "metadata"
distance_strategy: DistanceMeasure = DistanceMeasure.COSINE
_client: Client
def __init__(
self,
client: Optional[Client] = None,
**kwargs: Any,
) -> None:
"""初始化参数。"""
super().__init__(**kwargs)
object.__setattr__(self, "_client", client_with_user_agent(client))
@classmethod
def class_name(cls) -> str:
return "FirestoreVectorStore"
@property
def client(self) -> Any:
return self._client
def add(
self,
nodes: List[BaseNode],
) -> List[str]:
"""向向量存储中添加节点。"""
ids = []
entries = []
for node in nodes:
node_id = node.node_id
metadata = node_to_metadata_dict(
node,
remove_text=not self.stores_text,
flat_metadata=self.flat_metadata,
)
entry = {
self.embedding_key: node.get_embedding(),
self.metadata_key: metadata,
}
ids.append(node_id)
entries.append(entry)
self._upsert_batch(entries, ids)
return ids
def delete(self, ref_doc_id: str, **delete_kwargs: Any) -> None:
"""使用ref_doc_id删除节点。"""
docs = (
self._client.collection(self.collection_name)
.where("metadata.ref_doc_id", "==", ref_doc_id)
.stream()
)
self._delete_batch([doc.id for doc in docs])
def query(self, query: VectorStoreQuery, **kwargs: Any) -> VectorStoreQueryResult:
"""查询向量存储。"""
if query.query_embedding is None:
raise ValueError("Query embedding is required.")
filters = _to_firestore_filter(query.filters) if query.filters else None
results = self._similarity_search(
query.query_embedding, query.similarity_top_k, filters=filters, **kwargs
)
top_k_ids = []
top_k_nodes = []
top_k_similarities = []
LOGGER.debug(f"Found {len(results)} results.")
for result in results:
# Convert the Firestore document to dict
result_dict = result.to_dict() or {}
metadata = result_dict.get(self.metadata_key) or {}
fir_vec: Optional[Vector] = result_dict.get(self.embedding_key)
if fir_vec is None:
raise ValueError(
"Embedding is missing in Firestore document.", result.id
)
embedding = list(fir_vec.to_map_value()["value"])
# Convert metadata to node, and add text if available
node = metadata_dict_to_node(metadata, text=result_dict.get(self.text_key))
# Keep track of the top k ids and nodes
top_k_ids.append(result.id)
top_k_nodes.append(node)
top_k_similarities.append(
similarity(
query.query_embedding,
embedding,
self._distance_to_similarity_mode(self.distance_strategy),
)
)
return VectorStoreQueryResult(
nodes=top_k_nodes, ids=top_k_ids, similarities=top_k_similarities
)
def _distance_to_similarity_mode(self, distance: DistanceMeasure) -> SimilarityMode:
"""将Firestore的距离测量转换为相似度模式。"""
return {
DistanceMeasure.COSINE: SimilarityMode.DEFAULT,
DistanceMeasure.EUCLIDEAN: SimilarityMode.EUCLIDEAN,
DistanceMeasure.DOT_PRODUCT: SimilarityMode.DOT_PRODUCT,
}.get(distance, SimilarityMode.DEFAULT)
def _delete_batch(self, ids: List[str]) -> None:
"""从Firestore中删除一批向量。"""
db_batch = self._client.batch()
for batch in more_itertools.chunked(ids, DEFAULT_BATCH_SIZE):
for doc_id in batch:
doc = self._client.collection(self.collection_name).document(doc_id)
db_batch.delete(doc)
db_batch.commit()
def _upsert_batch(self, entries: List[dict], ids: Optional[List[str]]) -> None:
"""将一批向量插入/更新到Firestore。"""
if ids and len(ids) != len(entries):
raise ValueError("Length of ids and entries should be the same.")
db_batch = self._client.batch()
for batch in more_itertools.chunked(entries, DEFAULT_BATCH_SIZE):
for i, entry in enumerate(batch):
# Convert the embedding array to a Firestore Vector
entry[self.embedding_key] = Vector(entry[self.embedding_key])
doc = self._client.collection(self.collection_name).document(
ids[i] if ids else None
)
db_batch.set(doc, entry, merge=True)
db_batch.commit()
def _similarity_search(
self,
query: List[float],
k: int,
filters: Union[BaseFilter, BaseCompositeFilter, None] = None,
) -> List[DocumentSnapshot]:
wfilters = None
collection = self._client.collection(self.collection_name)
if filters:
wfilters = collection.where(filter=filters)
results = (wfilters or collection).find_nearest(
vector_field=self.embedding_key,
query_vector=Vector(query),
distance_measure=self.distance_strategy,
limit=k,
)
return results.get()
|