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 | class VectorStoreIndex(BaseIndex[IndexDict]):
"""索引向量存储。
Args:
use_async (bool): 是否使用异步调用。默认为False。
show_progress (bool): 是否显示tqdm进度条。默认为False。
store_nodes_override (bool): 设置为True，始终将Node对象存储在索引存储和文档存储中，即使向量存储保留文本。默认为False。"""
index_struct_cls = IndexDict
def __init__(
self,
nodes: Optional[Sequence[BaseNode]] = None,
# vector store index params
use_async: bool = False,
store_nodes_override: bool = False,
embed_model: Optional[EmbedType] = None,
insert_batch_size: int = 2048,
# parent class params
objects: Optional[Sequence[IndexNode]] = None,
index_struct: Optional[IndexDict] = None,
storage_context: Optional[StorageContext] = None,
callback_manager: Optional[CallbackManager] = None,
transformations: Optional[List[TransformComponent]] = None,
show_progress: bool = False,
# deprecated
service_context: Optional[ServiceContext] = None,
**kwargs: Any,
) -> None:
"""初始化参数。"""
self._use_async = use_async
self._store_nodes_override = store_nodes_override
self._embed_model = (
resolve_embed_model(embed_model, callback_manager=callback_manager)
if embed_model
else embed_model_from_settings_or_context(Settings, service_context)
)
self._insert_batch_size = insert_batch_size
super().__init__(
nodes=nodes,
index_struct=index_struct,
service_context=service_context,
storage_context=storage_context,
show_progress=show_progress,
objects=objects,
callback_manager=callback_manager,
transformations=transformations,
**kwargs,
)
@classmethod
def from_vector_store(
cls,
vector_store: BasePydanticVectorStore,
embed_model: Optional[EmbedType] = None,
# deprecated
service_context: Optional[ServiceContext] = None,
**kwargs: Any,
) -> "VectorStoreIndex":
if not vector_store.stores_text:
raise ValueError(
"Cannot initialize from a vector store that does not store text."
)
kwargs.pop("storage_context", None)
storage_context = StorageContext.from_defaults(vector_store=vector_store)
return cls(
nodes=[],
embed_model=embed_model,
service_context=service_context,
storage_context=storage_context,
**kwargs,
)
@property
def vector_store(self) -> BasePydanticVectorStore:
return self._vector_store
def as_retriever(self, **kwargs: Any) -> BaseRetriever:
# NOTE: lazy import
from llama_index.core.indices.vector_store.retrievers import (
VectorIndexRetriever,
)
return VectorIndexRetriever(
self,
node_ids=list(self.index_struct.nodes_dict.values()),
callback_manager=self._callback_manager,
object_map=self._object_map,
**kwargs,
)
def _get_node_with_embedding(
self,
nodes: Sequence[BaseNode],
show_progress: bool = False,
) -> List[BaseNode]:
"""获取id、节点和嵌入的元组。
允许我们将这些节点存储在向量存储中。
嵌入是以批量方式调用的。
"""
id_to_embed_map = embed_nodes(
nodes, self._embed_model, show_progress=show_progress
)
results = []
for node in nodes:
embedding = id_to_embed_map[node.node_id]
result = node.copy()
result.embedding = embedding
results.append(result)
return results
async def _aget_node_with_embedding(
self,
nodes: Sequence[BaseNode],
show_progress: bool = False,
) -> List[BaseNode]:
"""异步获取id、节点和嵌入的元组。
允许我们将这些节点存储在向量存储中。
嵌入是批量调用的。
"""
id_to_embed_map = await async_embed_nodes(
nodes=nodes,
embed_model=self._embed_model,
show_progress=show_progress,
)
results = []
for node in nodes:
embedding = id_to_embed_map[node.node_id]
result = node.copy()
result.embedding = embedding
results.append(result)
return results
async def _async_add_nodes_to_index(
self,
index_struct: IndexDict,
nodes: Sequence[BaseNode],
show_progress: bool = False,
**insert_kwargs: Any,
) -> None:
"""异步地将节点添加到索引中。"""
if not nodes:
return
for nodes_batch in iter_batch(nodes, self._insert_batch_size):
nodes_batch = await self._aget_node_with_embedding(
nodes_batch, show_progress
)
new_ids = await self._vector_store.async_add(nodes_batch, **insert_kwargs)
# if the vector store doesn't store text, we need to add the nodes to the
# index struct and document store
if not self._vector_store.stores_text or self._store_nodes_override:
for node, new_id in zip(nodes_batch, new_ids):
# NOTE: remove embedding from node to avoid duplication
node_without_embedding = node.copy()
node_without_embedding.embedding = None
index_struct.add_node(node_without_embedding, text_id=new_id)
self._docstore.add_documents(
[node_without_embedding], allow_update=True
)
else:
# NOTE: if the vector store keeps text,
# we only need to add image and index nodes
for node, new_id in zip(nodes_batch, new_ids):
if isinstance(node, (ImageNode, IndexNode)):
# NOTE: remove embedding from node to avoid duplication
node_without_embedding = node.copy()
node_without_embedding.embedding = None
index_struct.add_node(node_without_embedding, text_id=new_id)
self._docstore.add_documents(
[node_without_embedding], allow_update=True
)
def _add_nodes_to_index(
self,
index_struct: IndexDict,
nodes: Sequence[BaseNode],
show_progress: bool = False,
**insert_kwargs: Any,
) -> None:
"""向索引添加文档。"""
if not nodes:
return
for nodes_batch in iter_batch(nodes, self._insert_batch_size):
nodes_batch = self._get_node_with_embedding(nodes_batch, show_progress)
new_ids = self._vector_store.add(nodes_batch, **insert_kwargs)
if not self._vector_store.stores_text or self._store_nodes_override:
# NOTE: if the vector store doesn't store text,
# we need to add the nodes to the index struct and document store
for node, new_id in zip(nodes_batch, new_ids):
# NOTE: remove embedding from node to avoid duplication
node_without_embedding = node.copy()
node_without_embedding.embedding = None
index_struct.add_node(node_without_embedding, text_id=new_id)
self._docstore.add_documents(
[node_without_embedding], allow_update=True
)
else:
# NOTE: if the vector store keeps text,
# we only need to add image and index nodes
for node, new_id in zip(nodes_batch, new_ids):
if isinstance(node, (ImageNode, IndexNode)):
# NOTE: remove embedding from node to avoid duplication
node_without_embedding = node.copy()
node_without_embedding.embedding = None
index_struct.add_node(node_without_embedding, text_id=new_id)
self._docstore.add_documents(
[node_without_embedding], allow_update=True
)
def _build_index_from_nodes(
self,
nodes: Sequence[BaseNode],
**insert_kwargs: Any,
) -> IndexDict:
"""从节点构建索引。"""
index_struct = self.index_struct_cls()
if self._use_async:
tasks = [
self._async_add_nodes_to_index(
index_struct,
nodes,
show_progress=self._show_progress,
**insert_kwargs,
)
]
run_async_tasks(tasks)
else:
self._add_nodes_to_index(
index_struct,
nodes,
show_progress=self._show_progress,
**insert_kwargs,
)
return index_struct
def build_index_from_nodes(
self,
nodes: Sequence[BaseNode],
**insert_kwargs: Any,
) -> IndexDict:
"""从节点构建索引。
注意：覆盖了BaseIndex.build_index_from_nodes。
如果向量存储不存储文本，则VectorStoreIndex仅在文档存储中存储节点。
"""
# raise an error if even one node has no content
if any(
node.get_content(metadata_mode=MetadataMode.EMBED) == "" for node in nodes
):
raise ValueError(
"Cannot build index from nodes with no content. "
"Please ensure all nodes have content."
)
return self._build_index_from_nodes(nodes, **insert_kwargs)
def _insert(self, nodes: Sequence[BaseNode], **insert_kwargs: Any) -> None:
"""插入一个文档。"""
self._add_nodes_to_index(self._index_struct, nodes, **insert_kwargs)
def insert_nodes(self, nodes: Sequence[BaseNode], **insert_kwargs: Any) -> None:
"""插入节点。
注意：覆盖了BaseIndex.insert_nodes。
VectorStoreIndex仅在向量存储不存储文本时才将节点存储在文档存储中。
"""
for node in nodes:
if isinstance(node, IndexNode):
try:
node.dict()
except ValueError:
self._object_map[node.index_id] = node.obj
node.obj = None
with self._callback_manager.as_trace("insert_nodes"):
self._insert(nodes, **insert_kwargs)
self._storage_context.index_store.add_index_struct(self._index_struct)
def _delete_node(self, node_id: str, **delete_kwargs: Any) -> None:
pass
def delete_nodes(
self,
node_ids: List[str],
delete_from_docstore: bool = False,
**delete_kwargs: Any,
) -> None:
"""从索引中删除节点列表。
Args:
node_ids（List[str]）：要删除的节点的node_ids列表
"""
# delete nodes from vector store
self._vector_store.delete_nodes(node_ids, **delete_kwargs)
# delete from docstore only if needed
if (
not self._vector_store.stores_text or self._store_nodes_override
) and delete_from_docstore:
for node_id in node_ids:
self._docstore.delete_document(node_id, raise_error=False)
def delete_ref_doc(
self, ref_doc_id: str, delete_from_docstore: bool = False, **delete_kwargs: Any
) -> None:
"""使用ref_doc_id删除文档及其节点。"""
self._vector_store.delete(ref_doc_id, **delete_kwargs)
# delete from index_struct only if needed
if not self._vector_store.stores_text or self._store_nodes_override:
ref_doc_info = self._docstore.get_ref_doc_info(ref_doc_id)
if ref_doc_info is not None:
for node_id in ref_doc_info.node_ids:
self._index_struct.delete(node_id)
self._vector_store.delete(node_id)
# delete from docstore only if needed
if (
not self._vector_store.stores_text or self._store_nodes_override
) and delete_from_docstore:
self._docstore.delete_ref_doc(ref_doc_id, raise_error=False)
self._storage_context.index_store.add_index_struct(self._index_struct)
@property
def ref_doc_info(self) -> Dict[str, RefDocInfo]:
"""获取已摄取文档及其节点和元数据的字典映射。"""
if not self._vector_store.stores_text or self._store_nodes_override:
node_doc_ids = list(self.index_struct.nodes_dict.values())
nodes = self.docstore.get_nodes(node_doc_ids)
all_ref_doc_info = {}
for node in nodes:
ref_node = node.source_node
if not ref_node:
continue
ref_doc_info = self.docstore.get_ref_doc_info(ref_node.node_id)
if not ref_doc_info:
continue
all_ref_doc_info[ref_node.node_id] = ref_doc_info
return all_ref_doc_info
else:
raise NotImplementedError(
"Vector store integrations that store text in the vector store are "
"not supported by ref_doc_info yet."
)
|