feat: Add Vertex Vector Search Datapoints deletion, add autogenerated datapoint IDs to document metadata during ingestion

libs/vertexai/langchain_google_vertexai/vectorstores/_searcher.py

@@ -1,5 +1,5 @@
 from abc import ABC, abstractmethod
-from typing import Any, List, Tuple, Union
+from typing import Any, List, Sequence, Tuple, Union
 
 from google.cloud import storage  # type: ignore[attr-defined, unused-ignore]
 from google.cloud.aiplatform import telemetry
@@ -60,6 +60,11 @@ def add_to_index(
         """
         raise NotImplementedError()
 
+    @abstractmethod
+    def remove_datapoints(self, datapoint_ids: Sequence[str]) -> None:
+        """Remove datapoints from the index."""
+        raise NotImplementedError()
+
     def _postprocess_response(
         self, response: List[List[MatchNeighbor]]
     ) -> List[List[Tuple[str, float]]]:
@@ -136,6 +141,26 @@ def add_to_index(
                 is_complete_overwrite=is_complete_overwrite,
             )
 
+    def remove_datapoints(self, datapoint_ids: Sequence[str]) -> None:
+        """Remove datapoints from the index.
+
+        Args:
+            datapoint_ids: Sequence[str]
+                Required. The list of datapoint ids t
+        Raises:
+            ValueError: If the datapoint_ids sequence is empty.
+            RuntimeError: If there's an error while removing datapoints.
+        """
+        if not datapoint_ids:
+            raise ValueError("datapoint_ids must not be empty")
+
+        try:
+            self._index.remove_datapoints(datapoint_ids=datapoint_ids)
+        except Exception as e:
+            raise RuntimeError(
+                f"Error removing datapoints from the index: {str(e)}"
+            ) from e
+
     def find_neighbors(
         self,
         embeddings: List[List[float]],

libs/vertexai/langchain_google_vertexai/vectorstores/vectorstores.py

@@ -192,6 +192,10 @@ def add_texts(
                 f"{len(metadatas)} != {len(texts)}"
             )
 
+        # Add generated IDs to metadata
+        for metadata, id in zip(metadatas, ids):
+            metadata["doc_id"] = id
+
         documents = [
             Document(page_content=text, metadata=metadata)
             for text, metadata in zip(texts, metadatas)
@@ -207,6 +211,33 @@ def add_texts(
 
         return ids
 
+    def delete(self, ids: Optional[List[str]] = None, **kwargs: Any) -> Optional[bool]:
+        """
+        Delete by vector ID.
+        Args:
+            ids (Optional[List[str]]): List of ids to delete.
+            **kwargs (Any): Other keyword arguments (not used,
+            but included for compatibility).
+        Returns:
+            Optional[bool]: True if deletion is successful.
+        Raises:
+            ValueError: If ids is None or an empty list.
+            RuntimeError: If an error occurs during the deletion process.
+        """
+        if ids is None or len(ids) == 0:
+            raise ValueError("ids must be provided and cannot be an empty list")
+
+        try:
+            # Remove datapoints from the Vertex Vector Search Index
+            self._searcher.remove_datapoints(datapoint_ids=ids)
+            # Remove documents from the document storage
+            self._document_storage.mdelete(ids)
+
+            return True
+
+        except Exception as e:
+            raise RuntimeError(f"Error during deletion: {str(e)}") from e
+
     @classmethod
     def from_texts(
         cls: Type["_BaseVertexAIVectorStore"],