Fix vectorstore tests

There were three different problems:

 * Elasticsearch automatically applied certain index_options that were
   missing
 * Two tests were using FakeEmbeddings instead of
   ConsistentFakeEmbeddings
 * One test used [1, ..., 1, i] with i going from 0 to 2. Making it
   going from 1 to 3 allowed the cosine similarity to work as expected.

Author: pquentin

test_elasticsearch/test_server/test_vectorstore/__init__.py

@@ -69,7 +69,7 @@ def embed_documents(self, texts: List[str]) -> List[List[float]]:
             if text not in self.known_texts:
                 self.known_texts.append(text)
             vector = [float(1.0)] * (self.dimensionality - 1) + [
-                float(self.known_texts.index(text))
+                float(self.known_texts.index(text) + 1)
             ]
             out_vectors.append(vector)
         return out_vectors

test_elasticsearch/test_server/test_vectorstore/test_vectorstore.py

@@ -73,15 +73,15 @@ def assert_query(query_body: dict, query: Optional[str]) -> dict:
                     "filter": [],
                     "k": 1,
                     "num_candidates": 50,
-                    "query_vector": [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0],
+                    "query_vector": [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 2.0],
                 }
             }
             return query_body
 
         store = VectorStore(
             index=index,
             retrieval_strategy=DenseVectorStrategy(),
-            embedding_service=FakeEmbeddings(),
+            embedding_service=ConsistentFakeEmbeddings(),
             client=sync_client,
         )
 
@@ -98,7 +98,7 @@ def test_search_without_metadata_async(
         store = VectorStore(
             index=index,
             retrieval_strategy=DenseVectorStrategy(),
-            embedding_service=FakeEmbeddings(),
+            embedding_service=ConsistentFakeEmbeddings(),
             client=sync_client,
         )
 
@@ -1030,6 +1030,11 @@ def test_metadata_mapping(self, sync_client: Elasticsearch, index: str) -> None:
             "type": "dense_vector",
             "dims": 10,
             "index": True,
+            "index_options": {
+                "ef_construction": 100,
+                "m": 16,
+                "type": "int8_hnsw",
+            },
             "similarity": "cosine",
         }