fix: Use Spanner's UPSERT over INSERT (#90)

Switch to UPSERT. This provides for a nice experience for DELETE-then-INSERT use cases in Spanner when the primary key stays the same. Otherwise the DELETE is not pushed before INSERT and the client gets a 409 Row [abc] already exists. The way LangChain integration is designed, there is no easy way to wrap them all in a TX with additional args for such visibility. 

Co-authored-by: goodvibes <goodvibes@google.com>

Author: aperepel

samples/langchain_quick_start.ipynb

@@ -407,7 +407,7 @@
     "for i in range(1, len(rows), batch_size):\n",
     "    with database.batch() as batch:\n",
     "        batch_rows = rows[i : i + batch_size]\n",
-    "        batch.insert(\n",
+    "        batch.insert_or_update(\n",
     "            table=table_name,\n",
     "            columns=[\n",
     "                \"show_id\",\n",
@@ -676,7 +676,7 @@
     "for i in range(0, len(vector_store_rows), batch_size):\n",
     "    with database.batch() as batch:\n",
     "        batch_rows = vector_store_rows[i : i + batch_size]\n",
-    "        batch.insert(\n",
+    "        batch.insert_or_update(\n",
     "            table=sample_vector_table_name,\n",
     "            columns=[\n",
     "                \"langchain_id\",\n",

src/langchain_google_spanner/chat_message_history.py

@@ -212,7 +212,7 @@ def messages(self) -> List[BaseMessage]:  # type: ignore
     def add_message(self, message: BaseMessage) -> None:
         """Append the message to the record in Cloud Spanner"""
         with self.database.batch() as batch:
-            batch.insert(
+            batch.insert_or_update(
                 table=self.table_name,
                 columns=("session_id", "created_at", "message"),
                 values=[

src/langchain_google_spanner/loader.py

@@ -318,7 +318,7 @@ def add_documents(self, documents: List[Document]):
 
         for values_batch in _batch(values, MUTATION_BATCH_SIZE):
             with db.batch() as batch:
-                batch.insert(
+                batch.insert_or_update(
                     table=self.table_name,
                     columns=self._table_fields,
                     values=values_batch,

src/langchain_google_spanner/vector_store.py

@@ -706,7 +706,7 @@ def add_texts(
 
     def _insert_data(self, records, columns_to_insert):
         with self._database.batch() as batch:
-            batch.insert(
+            batch.insert_or_update(
                 table=self._table_name,
                 columns=columns_to_insert,
                 values=records,