fix: when using IVF* indexing, ensure the index is trained frist (#4311)

* add protection, in case we use IVF* indexing, we need to train the index first

Signed-off-by: Liu,Kaixuan <kaixuan.liu@intel.com>

* fix formatting issue

Signed-off-by: Liu,Kaixuan <kaixuan.liu@intel.com>

* just raising error, instead of silently training the index

* fixed mypy issue

* fixed error msg

---------

Signed-off-by: Liu,Kaixuan <kaixuan.liu@intel.com>
Co-authored-by: Mayank Jobanputra <mayankjobanputra@gmail.com>

haystack/document_stores/faiss.py

@@ -11,10 +11,9 @@ import numpy as np
 from tqdm.auto import tqdm
 
 try:
+    # These deps are optional, but get installed with the `faiss` extra
     import faiss
-    from haystack.document_stores.sql import (
+    from haystack.document_stores.sql import SQLDocumentStore  # type: ignore
-        SQLDocumentStore,
-    )  # its deps are optional, but get installed with the `faiss` extra
 except (ImportError, ModuleNotFoundError) as ie:
     from haystack.utils.import_utils import _optional_component_not_installed
 
@@ -276,6 +275,14 @@ class FAISSDocumentStore(SQLDocumentStore):
             ) as progress_bar:
                 for i in range(0, len(document_objects), batch_size):
                     if add_vectors:
+                        if not self.faiss_indexes[index].is_trained:
+                            raise ValueError(
+                                "FAISS index of type {} must be trained before adding vectors. Call `train_index()` "
+                                "method before adding the vectors. For details, refer to the documentation: "
+                                "[FAISSDocumentStore API](https://docs.haystack.deepset.ai/reference/document-store-api#faissdocumentstoretrain_index)."
+                                "".format(self.faiss_index_factory_str)
+                            )
+
                         embeddings = [doc.embedding for doc in document_objects[i : i + batch_size]]
                         embeddings_to_index = np.array(embeddings, dtype="float32")
 
@@ -339,6 +346,14 @@ class FAISSDocumentStore(SQLDocumentStore):
         if not self.faiss_indexes.get(index):
             raise ValueError("Couldn't find a FAISS index. Try to init the FAISSDocumentStore() again ...")
 
+        if not self.faiss_indexes[index].is_trained:
+            raise ValueError(
+                "FAISS index of type {} must be trained before adding vectors. Call `train_index()` "
+                "method before adding the vectors. For details, refer to the documentation: "
+                "[FAISSDocumentStore API](https://docs.haystack.deepset.ai/reference/document-store-api#faissdocumentstoretrain_index)."
+                "".format(self.faiss_index_factory_str)
+            )
+
         document_count = self.get_document_count(index=index)
         if document_count == 0:
             logger.warning("Calling DocumentStore.update_embeddings() on an empty index")

test/document_stores/test_faiss.py

@@ -129,6 +129,17 @@ class TestFAISSDocumentStore(DocumentStoreBaseTestAbstract):
         # Check that get_embedding_count works as expected
         assert document_store.get_embedding_count() == len(documents_with_embeddings)
 
+    @pytest.mark.integration
+    def test_write_docs_no_training(self, documents_with_embeddings, tmp_path, caplog):
+        document_store = FAISSDocumentStore(
+            sql_url=f"sqlite:///{tmp_path}/test_write_docs_no_training.db",
+            faiss_index_factory_str="IVF1,Flat",
+            isolation_level="AUTOCOMMIT",
+            return_embedding=True,
+        )
+        with pytest.raises(ValueError, match="must be trained before adding vectors"):
+            document_store.write_documents(documents_with_embeddings)
+
     @pytest.mark.integration
     def test_train_index_from_docs(self, documents_with_embeddings, tmp_path):
         document_store = FAISSDocumentStore(