Permit updates in indexing (#14482)

libs/langchain/langchain/indexes/_api.py

@@ -195,6 +195,7 @@ def index(
     cleanup: Literal["incremental", "full", None] = None,
     source_id_key: Union[str, Callable[[Document], str], None] = None,
     cleanup_batch_size: int = 1_000,
+    force_update: bool = False,
 ) -> IndexingResult:
     """Index data from the loader into the vector store.
 
@@ -233,6 +234,8 @@ def index(
         source_id_key: Optional key that helps identify the original source
             of the document.
         cleanup_batch_size: Batch size to use when cleaning up documents.
+        force_update: Force update documents even if they are present in the
+            record manager. Useful if you are re-indexing with updated embeddings.
 
     Returns:
         Indexing result which contains information about how many documents
@@ -308,8 +311,12 @@ def index(
         uids = []
         docs_to_index = []
         uids_to_refresh = []
+        seen_docs: Set[str] = set()
         for hashed_doc, doc_exists in zip(hashed_docs, exists_batch):
             if doc_exists:
+                if force_update:
+                    seen_docs.add(hashed_doc.uid)
+                else:
                     uids_to_refresh.append(hashed_doc.uid)
                     continue
             uids.append(hashed_doc.uid)
@@ -324,7 +331,8 @@ def index(
         # First write to vector store
         if docs_to_index:
             vector_store.add_documents(docs_to_index, ids=uids)
-            num_added += len(docs_to_index)
+            num_added += len(docs_to_index) - len(seen_docs)
+            num_updated += len(seen_docs)
 
         # And only then update the record store.
         # Update ALL records, even if they already exist since we want to refresh
@@ -391,6 +399,7 @@ async def aindex(
     cleanup: Literal["incremental", "full", None] = None,
     source_id_key: Union[str, Callable[[Document], str], None] = None,
     cleanup_batch_size: int = 1_000,
+    force_update: bool = False,
 ) -> IndexingResult:
     """Index data from the loader into the vector store.
 
@@ -429,6 +438,8 @@ async def aindex(
         source_id_key: Optional key that helps identify the original source
             of the document.
         cleanup_batch_size: Batch size to use when cleaning up documents.
+        force_update: Force update documents even if they are present in the
+            record manager. Useful if you are re-indexing with updated embeddings.
 
     Returns:
         Indexing result which contains information about how many documents
@@ -508,9 +519,12 @@ async def aindex(
         uids: list[str] = []
         docs_to_index: list[Document] = []
         uids_to_refresh = []
-
+        seen_docs: Set[str] = set()
         for hashed_doc, doc_exists in zip(hashed_docs, exists_batch):
             if doc_exists:
+                if force_update:
+                    seen_docs.add(hashed_doc.uid)
+                else:
                     uids_to_refresh.append(hashed_doc.uid)
                     continue
             uids.append(hashed_doc.uid)
@@ -525,7 +539,8 @@ async def aindex(
         # First write to vector store
         if docs_to_index:
             await vector_store.aadd_documents(docs_to_index, ids=uids)
-            num_added += len(docs_to_index)
+            num_added += len(docs_to_index) - len(seen_docs)
+            num_updated += len(seen_docs)
 
         # And only then update the record store.
         # Update ALL records, even if they already exist since we want to refresh

libs/langchain/tests/unit_tests/indexes/test_indexing.py

@@ -58,9 +58,10 @@ class ToyLoader(BaseLoader):
 class InMemoryVectorStore(VectorStore):
     """In-memory implementation of VectorStore using a dictionary."""
 
-    def __init__(self) -> None:
+    def __init__(self, permit_upserts: bool = False) -> None:
         """Vector store interface for testing things in memory."""
         self.store: Dict[str, Document] = {}
+        self.permit_upserts = permit_upserts
 
     def delete(self, ids: Optional[Sequence[str]] = None, **kwargs: Any) -> None:
         """Delete the given documents from the store using their IDs."""
@@ -91,7 +92,7 @@ class InMemoryVectorStore(VectorStore):
             raise NotImplementedError("This is not implemented yet.")
 
         for _id, document in zip(ids, documents):
-            if _id in self.store:
+            if _id in self.store and not self.permit_upserts:
                 raise ValueError(
                     f"Document with uid {_id} already exists in the store."
                 )
@@ -115,7 +116,7 @@ class InMemoryVectorStore(VectorStore):
             raise NotImplementedError("This is not implemented yet.")
 
         for _id, document in zip(ids, documents):
-            if _id in self.store:
+            if _id in self.store and not self.permit_upserts:
                 raise ValueError(
                     f"Document with uid {_id} already exists in the store."
                 )
@@ -176,6 +177,12 @@ def vector_store() -> InMemoryVectorStore:
     return InMemoryVectorStore()
 
 
+@pytest.fixture
+def upserting_vector_store() -> InMemoryVectorStore:
+    """Vector store fixture."""
+    return InMemoryVectorStore(permit_upserts=True)
+
+
 def test_indexing_same_content(
     record_manager: SQLRecordManager, vector_store: InMemoryVectorStore
 ) -> None:
@@ -1074,6 +1081,101 @@ async def test_abatch() -> None:
     assert [batch async for batch in batches] == [[0, 1], [2, 3], [4]]
 
 
+def test_indexing_force_update(
+    record_manager: SQLRecordManager, upserting_vector_store: VectorStore
+) -> None:
+    """Test indexing with force update."""
+    docs = [
+        Document(
+            page_content="This is a test document.",
+            metadata={"source": "1"},
+        ),
+        Document(
+            page_content="This is another document.",
+            metadata={"source": "2"},
+        ),
+        Document(
+            page_content="This is a test document.",
+            metadata={"source": "1"},
+        ),
+    ]
+
+    assert index(docs, record_manager, upserting_vector_store, cleanup="full") == {
+        "num_added": 2,
+        "num_deleted": 0,
+        "num_skipped": 0,
+        "num_updated": 0,
+    }
+
+    assert index(docs, record_manager, upserting_vector_store, cleanup="full") == {
+        "num_added": 0,
+        "num_deleted": 0,
+        "num_skipped": 2,
+        "num_updated": 0,
+    }
+
+    assert index(
+        docs, record_manager, upserting_vector_store, cleanup="full", force_update=True
+    ) == {
+        "num_added": 0,
+        "num_deleted": 0,
+        "num_skipped": 0,
+        "num_updated": 2,
+    }
+
+
+@pytest.mark.requires("aiosqlite")
+async def test_aindexing_force_update(
+    arecord_manager: SQLRecordManager, upserting_vector_store: VectorStore
+) -> None:
+    """Test indexing with force update."""
+    docs = [
+        Document(
+            page_content="This is a test document.",
+            metadata={"source": "1"},
+        ),
+        Document(
+            page_content="This is another document.",
+            metadata={"source": "2"},
+        ),
+        Document(
+            page_content="This is a test document.",
+            metadata={"source": "1"},
+        ),
+    ]
+
+    assert await aindex(
+        docs, arecord_manager, upserting_vector_store, cleanup="full"
+    ) == {
+        "num_added": 2,
+        "num_deleted": 0,
+        "num_skipped": 0,
+        "num_updated": 0,
+    }
+
+    assert await aindex(
+        docs, arecord_manager, upserting_vector_store, cleanup="full"
+    ) == {
+        "num_added": 0,
+        "num_deleted": 0,
+        "num_skipped": 2,
+        "num_updated": 0,
+    }
+
+    assert await aindex(
+        docs,
+        arecord_manager,
+        upserting_vector_store,
+        cleanup="full",
+        force_update=True,
+    ) == {
+        "num_added": 0,
+        "num_deleted": 0,
+        "num_skipped": 0,
+        "num_updated": 2,
+    }
+
+
 def test_compatible_vectorstore_documentation() -> None:
     """Test which vectorstores are compatible with the indexing API.