VectorStore -> BasePydanticVectorStore, also get/delete_nodes, clear()

docs/docs/examples/low_level/vector_store.ipynb

@@ -166,7 +166,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "from llama_index.core.vector_stores.types import VectorStore\n",
+    "from llama_index.core.vector_stores.types import BasePydanticVectorStore\n",
     "from llama_index.core.vector_stores import (\n",
     "    VectorStoreQuery,\n",
     "    VectorStoreQueryResult,\n",
@@ -176,7 +176,7 @@
     "import os\n",
     "\n",
     "\n",
-    "class BaseVectorStore(VectorStore):\n",
+    "class BaseVectorStore(BasePydanticVectorStore):\n",
     "    \"\"\"Simple custom Vector Store.\n",
     "\n",
     "    Stores documents in a simple in-memory dict.\n",
@@ -311,14 +311,14 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "from llama_index.core.bridge.pydantic import Field\n",
+    "\n",
+    "\n",
     "class VectorStore2(BaseVectorStore):\n",
     "    \"\"\"VectorStore2 (add/get/delete implemented).\"\"\"\n",
     "\n",
     "    stores_text: bool = True\n",
-    "\n",
-    "    def __init__(self) -> None:\n",
-    "        \"\"\"Init params.\"\"\"\n",
-    "        self.node_dict: Dict[str, BaseNode] = {}\n",
+    "    node_dict: Dict[str, BaseNode] = Field(default_factory=dict)\n",
     "\n",
     "    def get(self, text_id: str) -> List[float]:\n",
     "        \"\"\"Get embedding.\"\"\"\n",
@@ -469,6 +469,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "from typing import cast\n",
+    "\n",
+    "\n",
     "class VectorStore3A(VectorStore2):\n",
     "    \"\"\"Implements semantic/dense search.\"\"\"\n",
     "\n",
@@ -485,7 +488,7 @@
     "\n",
     "        similarities, node_ids = get_top_k_embeddings(\n",
     "            query_embedding,\n",
-    "            embeddings,\n",
+    "            doc_embeddings,\n",
     "            doc_ids,\n",
     "            similarity_top_k=query.similarity_top_k,\n",
     "        )\n",
@@ -891,9 +894,9 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "llama_index_v2",
+   "display_name": "venv",
    "language": "python",
-   "name": "llama_index_v2"
+   "name": "python3"
   },
   "language_info": {
    "codemirror_mode": {

llama-index-core/llama_index/core/indices/document_summary/base.py

@@ -40,7 +40,7 @@
 from llama_index.core.storage.docstore.types import RefDocInfo
 from llama_index.core.storage.storage_context import StorageContext
 from llama_index.core.utils import get_tqdm_iterable
-from llama_index.core.vector_stores.types import VectorStore
+from llama_index.core.vector_stores.types import BasePydanticVectorStore
 
 logger = logging.getLogger(__name__)
 
@@ -115,7 +115,7 @@ def __init__(
         )
 
     @property
-    def vector_store(self) -> VectorStore:
+    def vector_store(self) -> BasePydanticVectorStore:
         return self._vector_store
 
     def as_retriever(

llama-index-core/llama_index/core/indices/multi_modal/base.py

@@ -33,7 +33,7 @@
     DEFAULT_VECTOR_STORE,
     SimpleVectorStore,
 )
-from llama_index.core.vector_stores.types import VectorStore
+from llama_index.core.vector_stores.types import BasePydanticVectorStore
 
 logger = logging.getLogger(__name__)
 
@@ -63,7 +63,7 @@ def __init__(
         # Image-related kwargs
         # image_vector_store going to be deprecated. image_store can be passed from storage_context
         # keep image_vector_store here for backward compatibility
-        image_vector_store: Optional[VectorStore] = None,
+        image_vector_store: Optional[BasePydanticVectorStore] = None,
         image_embed_model: EmbedType = "clip:ViT-B/32",
         is_image_to_text: bool = False,
         # is_image_vector_store_empty is used to indicate whether image_vector_store is empty
@@ -112,7 +112,7 @@ def __init__(
         )
 
     @property
-    def image_vector_store(self) -> VectorStore:
+    def image_vector_store(self) -> BasePydanticVectorStore:
         return self._image_vector_store
 
     @property
@@ -164,12 +164,12 @@ def as_query_engine(
     @classmethod
     def from_vector_store(
         cls,
-        vector_store: VectorStore,
+        vector_store: BasePydanticVectorStore,
         embed_model: Optional[EmbedType] = None,
         # deprecated
         service_context: Optional[ServiceContext] = None,
         # Image-related kwargs
-        image_vector_store: Optional[VectorStore] = None,
+        image_vector_store: Optional[BasePydanticVectorStore] = None,
         image_embed_model: EmbedType = "clip",
         **kwargs: Any,
     ) -> "VectorStoreIndex":

llama-index-core/llama_index/core/indices/multi_modal/retriever.py

@@ -25,7 +25,7 @@
 )
 from llama_index.core.vector_stores.types import (
     MetadataFilters,
-    VectorStore,
+    BasePydanticVectorStore,
     VectorStoreQuery,
     VectorStoreQueryMode,
     VectorStoreQueryResult,
@@ -225,7 +225,7 @@ def _get_nodes_with_embeddings(
         self,
         query_bundle_with_embeddings: QueryBundle,
         similarity_top_k: int,
-        vector_store: VectorStore,
+        vector_store: BasePydanticVectorStore,
     ) -> List[NodeWithScore]:
         query = self._build_vector_store_query(
             query_bundle_with_embeddings, similarity_top_k
@@ -347,7 +347,7 @@ async def _aget_nodes_with_embeddings(
         self,
         query_bundle_with_embeddings: QueryBundle,
         similarity_top_k: int,
-        vector_store: VectorStore,
+        vector_store: BasePydanticVectorStore,
     ) -> List[NodeWithScore]:
         query = self._build_vector_store_query(
             query_bundle_with_embeddings, similarity_top_k

llama-index-core/llama_index/core/indices/vector_store/base.py

@@ -26,7 +26,7 @@
 from llama_index.core.storage.docstore.types import RefDocInfo
 from llama_index.core.storage.storage_context import StorageContext
 from llama_index.core.utils import iter_batch
-from llama_index.core.vector_stores.types import VectorStore
+from llama_index.core.vector_stores.types import BasePydanticVectorStore
 
 logger = logging.getLogger(__name__)
 
@@ -87,7 +87,7 @@ def __init__(
     @classmethod
     def from_vector_store(
         cls,
-        vector_store: VectorStore,
+        vector_store: BasePydanticVectorStore,
         embed_model: Optional[EmbedType] = None,
         # deprecated
         service_context: Optional[ServiceContext] = None,
@@ -110,7 +110,7 @@ def from_vector_store(
         )
 
     @property
-    def vector_store(self) -> VectorStore:
+    def vector_store(self) -> BasePydanticVectorStore:
         return self._vector_store
 
     def as_retriever(self, **kwargs: Any) -> BaseRetriever:

llama-index-core/llama_index/core/storage/storage_context.py

@@ -38,7 +38,6 @@
 )
 from llama_index.core.vector_stores.types import (
     BasePydanticVectorStore,
-    VectorStore,
 )
 
 DEFAULT_PERSIST_DIR = "./storage"
@@ -54,26 +53,24 @@ class StorageContext:
     indices, and vectors. It contains the following:
     - docstore: BaseDocumentStore
     - index_store: BaseIndexStore
-    - vector_store: VectorStore
+    - vector_store: BasePydanticVectorStore
     - graph_store: GraphStore
 
     """
 
     docstore: BaseDocumentStore
     index_store: BaseIndexStore
-    vector_stores: Dict[str, VectorStore]
+    vector_stores: Dict[str, BasePydanticVectorStore]
     graph_store: GraphStore
 
     @classmethod
     def from_defaults(
         cls,
         docstore: Optional[BaseDocumentStore] = None,
         index_store: Optional[BaseIndexStore] = None,
-        vector_store: Optional[Union[VectorStore, BasePydanticVectorStore]] = None,
-        image_store: Optional[VectorStore] = None,
-        vector_stores: Optional[
-            Dict[str, Union[VectorStore, BasePydanticVectorStore]]
-        ] = None,
+        vector_store: Optional[BasePydanticVectorStore] = None,
+        image_store: Optional[BasePydanticVectorStore] = None,
+        vector_stores: Optional[Dict[str, BasePydanticVectorStore]] = None,
         graph_store: Optional[GraphStore] = None,
         persist_dir: Optional[str] = None,
         fs: Optional[fsspec.AbstractFileSystem] = None,
@@ -83,9 +80,9 @@ def from_defaults(
         Args:
             docstore (Optional[BaseDocumentStore]): document store
             index_store (Optional[BaseIndexStore]): index store
-            vector_store (Optional[VectorStore]): vector store
+            vector_store (Optional[BasePydanticVectorStore]): vector store
             graph_store (Optional[GraphStore]): graph store
-            image_store (Optional[VectorStore]): image store
+            image_store (Optional[BasePydanticVectorStore]): image store
 
         """
         if persist_dir is None:
@@ -214,7 +211,7 @@ def from_dict(cls, save_dict: dict) -> "StorageContext":
         index_store = SimpleIndexStore.from_dict(save_dict[INDEX_STORE_KEY])
         graph_store = SimpleGraphStore.from_dict(save_dict[GRAPH_STORE_KEY])
 
-        vector_stores: Dict[str, VectorStore] = {}
+        vector_stores: Dict[str, BasePydanticVectorStore] = {}
         for key, vector_store_dict in save_dict[VECTOR_STORE_KEY].items():
             vector_stores[key] = SimpleVectorStore.from_dict(vector_store_dict)
 
@@ -226,10 +223,12 @@ def from_dict(cls, save_dict: dict) -> "StorageContext":
         )
 
     @property
-    def vector_store(self) -> VectorStore:
+    def vector_store(self) -> BasePydanticVectorStore:
         """Backwrds compatibility for vector_store property."""
         return self.vector_stores[DEFAULT_VECTOR_STORE]
 
-    def add_vector_store(self, vector_store: VectorStore, namespace: str) -> None:
+    def add_vector_store(
+        self, vector_store: BasePydanticVectorStore, namespace: str
+    ) -> None:
         """Add a vector store to the storage context."""
         self.vector_stores[namespace] = vector_store

llama-index-core/llama_index/core/vector_stores/simple.py

@@ -8,6 +8,7 @@
 
 import fsspec
 from dataclasses_json import DataClassJsonMixin
+from llama_index.core.bridge.pydantic import PrivateAttr
 from llama_index.core.indices.query.embedding_utils import (
     get_top_k_embeddings,
     get_top_k_embeddings_learner,
@@ -18,9 +19,9 @@
 from llama_index.core.vector_stores.types import (
     DEFAULT_PERSIST_DIR,
     DEFAULT_PERSIST_FNAME,
+    BasePydanticVectorStore,
     MetadataFilters,
     FilterCondition,
-    VectorStore,
     VectorStoreQuery,
     VectorStoreQueryMode,
     VectorStoreQueryResult,
@@ -95,7 +96,7 @@ class SimpleVectorStoreData(DataClassJsonMixin):
     metadata_dict: Dict[str, Any] = field(default_factory=dict)
 
 
-class SimpleVectorStore(VectorStore):
+class SimpleVectorStore(BasePydanticVectorStore):
     """Simple Vector Store.
 
     In this vector store, embeddings are stored within a simple, in-memory dictionary.
@@ -107,6 +108,8 @@ class SimpleVectorStore(VectorStore):
     """
 
     stores_text: bool = False
+    _data: SimpleVectorStoreData = PrivateAttr()
+    _fs: fsspec.AbstractFileSystem = PrivateAttr()
 
     def __init__(
         self,
@@ -115,6 +118,7 @@ def __init__(
         **kwargs: Any,
     ) -> None:
         """Initialize params."""
+        super().__init__()
         self._data = data or SimpleVectorStoreData()
         self._fs = fs or fsspec.filesystem("file")
 
@@ -142,11 +146,11 @@ def from_namespaced_persist_dir(
         cls,
         persist_dir: str = DEFAULT_PERSIST_DIR,
         fs: Optional[fsspec.AbstractFileSystem] = None,
-    ) -> Dict[str, VectorStore]:
+    ) -> Dict[str, BasePydanticVectorStore]:
         """Load from namespaced persist dir."""
         listing_fn = os.listdir if fs is None else fs.listdir
 
-        vector_stores: Dict[str, VectorStore] = {}
+        vector_stores: Dict[str, BasePydanticVectorStore] = {}
 
         try:
             for fname in listing_fn(persist_dir):
@@ -176,6 +180,11 @@ def from_namespaced_persist_dir(
 
         return vector_stores
 
+    @classmethod
+    def class_name(cls) -> str:
+        """Class name."""
+        return "SimpleVectorStore"
+
     @property
     def client(self) -> None:
         """Get client."""
@@ -185,6 +194,14 @@ def get(self, text_id: str) -> List[float]:
         """Get embedding."""
         return self._data.embedding_dict[text_id]
 
+    def get_nodes(
+        self,
+        node_ids: Optional[List[str]] = None,
+        filters: Optional[MetadataFilters] = None,
+    ) -> List[BaseNode]:
+        """Get nodes."""
+        raise NotImplementedError("SimpleVectorStore does not store nodes directly.")
+
     def add(
         self,
         nodes: List[BaseNode],
@@ -224,6 +241,37 @@ def delete(self, ref_doc_id: str, **delete_kwargs: Any) -> None:
             if self._data.metadata_dict is not None:
                 self._data.metadata_dict.pop(text_id, None)
 
+    def delete_nodes(
+        self,
+        node_ids: Optional[List[str]] = None,
+        filters: Optional[MetadataFilters] = None,
+        **delete_kwargs: Any,
+    ) -> None:
+        filter_fn = _build_metadata_filter_fn(
+            lambda node_id: self._data.metadata_dict[node_id], filters
+        )
+
+        if node_ids is not None:
+            node_id_set = set(node_ids)
+
+            def node_filter_fn(node_id: str) -> bool:
+                return node_id in node_id_set and filter_fn(node_id)
+
+        else:
+
+            def node_filter_fn(node_id: str) -> bool:
+                return filter_fn(node_id)
+
+        for node_id in list(self._data.embedding_dict.keys()):
+            if node_filter_fn(node_id):
+                del self._data.embedding_dict[node_id]
+                del self._data.text_id_to_ref_doc_id[node_id]
+                self._data.metadata_dict.pop(node_id, None)
+
+    def clear(self) -> None:
+        """Clear the store."""
+        self._data = SimpleVectorStoreData()
+
     def query(
         self,
         query: VectorStoreQuery,