Fix tests, add parameter

Author: pamelafox

app/backend/app.py

@@ -458,6 +458,7 @@ async def setup_clients():
     AZURE_SEARCH_QUERY_REWRITING = os.getenv("AZURE_SEARCH_QUERY_REWRITING", "false").lower()
     # This defaults to the previous field name "embedding", for backwards compatibility
     AZURE_SEARCH_FIELD_NAME_EMBEDDING = os.getenv("AZURE_SEARCH_FIELD_NAME_EMBEDDING", "embedding")
+    AZURE_SEARCH_FIELD_NAME_IMAGE_EMBEDDING = os.getenv("AZURE_SEARCH_FIELD_NAME_IMAGE_EMBEDDING", "imageEmbedding")
 
     AZURE_SPEECH_SERVICE_ID = os.getenv("AZURE_SPEECH_SERVICE_ID")
     AZURE_SPEECH_SERVICE_LOCATION = os.getenv("AZURE_SPEECH_SERVICE_LOCATION")
@@ -574,7 +575,11 @@ async def setup_clients():
             disable_vectors=os.getenv("USE_VECTORS", "").lower() == "false",
         )
         ingester = UploadUserFileStrategy(
-            search_info=search_info, embeddings=text_embeddings_service, file_processors=file_processors
+            search_info=search_info,
+            embeddings=text_embeddings_service,
+            file_processors=file_processors,
+            search_field_name_embedding=AZURE_SEARCH_FIELD_NAME_EMBEDDING,
+            search_field_name_image_embedding=AZURE_SEARCH_FIELD_NAME_IMAGE_EMBEDDING,
         )
         current_app.config[CONFIG_INGESTER] = ingester
 

app/backend/prepdocs.py

@@ -432,8 +432,9 @@ async def main(strategy: Strategy, setup_index: bool = True):
             embeddings=openai_embeddings_service,
             image_embeddings=image_embeddings_service,
             search_analyzer_name=os.getenv("AZURE_SEARCH_ANALYZER_NAME"),
-            search_field_name_embedding=os.getenv("AZURE_SEARCH_FIELD_NAME_EMBEDDING"),
-            search_field_name_image_embedding=os.getenv("AZURE_SEARCH_FIELD_NAME_IMAGE_EMBEDDING"),
+            # Default to the previous field names for backward compatibility
+            search_field_name_embedding=os.getenv("AZURE_SEARCH_FIELD_NAME_EMBEDDING", "embedding"),
+            search_field_name_image_embedding=os.getenv("AZURE_SEARCH_FIELD_NAME_IMAGE_EMBEDDING", "imageEmbedding"),
             use_acls=use_acls,
             category=args.category,
             use_content_understanding=use_content_understanding,

app/backend/prepdocslib/filestrategy.py

@@ -136,12 +136,25 @@ def __init__(
         file_processors: dict[str, FileProcessor],
         embeddings: Optional[OpenAIEmbeddings] = None,
         image_embeddings: Optional[ImageEmbeddings] = None,
+        search_field_name_embedding: Optional[str] = None,
+        search_field_name_image_embedding: Optional[str] = None,
     ):
         self.file_processors = file_processors
         self.embeddings = embeddings
         self.image_embeddings = image_embeddings
         self.search_info = search_info
-        self.search_manager = SearchManager(self.search_info, None, True, False, self.embeddings)
+        self.search_manager = SearchManager(
+            search_info=self.search_info,
+            search_analyzer_name=None,
+            use_acls=True,
+            use_int_vectorization=False,
+            embeddings=self.embeddings,
+            field_name_embedding=search_field_name_embedding,
+            field_name_image_embedding=search_field_name_image_embedding,
+            search_images=False,
+        )
+        self.search_field_name_embedding = search_field_name_embedding
+        self.search_field_name_image_embedding = search_field_name_image_embedding
 
     async def add_file(self, file: File):
         if self.image_embeddings:

infra/main.bicep

@@ -28,6 +28,7 @@ param searchQueryLanguage string // Set in main.parameters.json
 param searchQuerySpeller string // Set in main.parameters.json
 param searchServiceSemanticRankerLevel string // Set in main.parameters.json
 param searchFieldNameEmbedding string // Set in main.parameters.json
+param searchFieldNameImageEmbedding string // Set in main.parameters.json
 var actualSearchServiceSemanticRankerLevel = (searchServiceSkuName == 'free')
   ? 'disabled'
   : searchServiceSemanticRankerLevel
@@ -392,6 +393,7 @@ var appEnvVariables = {
   AZURE_SEARCH_QUERY_LANGUAGE: searchQueryLanguage
   AZURE_SEARCH_QUERY_SPELLER: searchQuerySpeller
   AZURE_SEARCH_FIELD_NAME_EMBEDDING: searchFieldNameEmbedding
+  AZURE_SEARCH_FIELD_NAME_IMAGE_EMBEDDING: searchFieldNameImageEmbedding
   APPLICATIONINSIGHTS_CONNECTION_STRING: useApplicationInsights
     ? monitoring.outputs.applicationInsightsConnectionString
     : ''
@@ -1288,6 +1290,7 @@ output AZURE_SEARCH_SERVICE_RESOURCE_GROUP string = searchServiceResourceGroup.n
 output AZURE_SEARCH_SEMANTIC_RANKER string = actualSearchServiceSemanticRankerLevel
 output AZURE_SEARCH_SERVICE_ASSIGNED_USERID string = searchService.outputs.principalId
 output AZURE_SEARCH_FIELD_NAME_EMBEDDING string = searchFieldNameEmbedding
+output AZURE_SEARCH_FIELD_NAME_IMAGE_EMBEDDING string = searchFieldNameEmbedding
 
 output AZURE_COSMOSDB_ACCOUNT string = (useAuthentication && useChatHistoryCosmos) ? cosmosDb.outputs.name : ''
 output AZURE_CHAT_HISTORY_DATABASE string = chatHistoryDatabaseName

infra/main.parameters.json

@@ -86,6 +86,9 @@
     "searchFieldNameEmbedding": {
       "value": "${AZURE_SEARCH_FIELD_NAME_EMBEDDING=embedding3}"
     },
+    "searchFieldNameImageEmbedding": {
+      "value": "${AZURE_SEARCH_FIELD_NAME_IMAGE_EMBEDDING=imageEmbedding}"
+    },
     "storageAccountName": {
       "value": "${AZURE_STORAGE_ACCOUNT}"
     },

tests/conftest.py

@@ -246,12 +246,16 @@ def mock_blob_container_client(monkeypatch):
         "OPENAI_HOST": "openai",
         "OPENAI_API_KEY": "secretkey",
         "OPENAI_ORGANIZATION": "organization",
+        "AZURE_OPENAI_EMB_MODEL_NAME": "text-embedding-3-large",
+        "AZURE_OPENAI_EMB_DIMENSIONS": "3072",
     },
     {
         "OPENAI_HOST": "azure",
         "AZURE_OPENAI_SERVICE": "test-openai-service",
         "AZURE_OPENAI_CHATGPT_DEPLOYMENT": "test-chatgpt",
         "AZURE_OPENAI_EMB_DEPLOYMENT": "test-ada",
+        "AZURE_OPENAI_EMB_MODEL_NAME": "text-embedding-3-large",
+        "AZURE_OPENAI_EMB_DIMENSIONS": "3072",
         "USE_GPT4V": "true",
         "AZURE_OPENAI_GPT4V_MODEL": "gpt-4",
         "VISION_ENDPOINT": "https://testvision.cognitiveservices.azure.com/",
@@ -264,6 +268,8 @@ def mock_blob_container_client(monkeypatch):
         "AZURE_OPENAI_SERVICE": "test-openai-service",
         "AZURE_OPENAI_CHATGPT_DEPLOYMENT": "test-chatgpt",
         "AZURE_OPENAI_EMB_DEPLOYMENT": "test-ada",
+        "AZURE_OPENAI_EMB_MODEL_NAME": "text-embedding-3-large",
+        "AZURE_OPENAI_EMB_DIMENSIONS": "3072",
         "AZURE_USE_AUTHENTICATION": "true",
         "AZURE_USER_STORAGE_ACCOUNT": "test-user-storage-account",
         "AZURE_USER_STORAGE_CONTAINER": "test-user-storage-container",
@@ -280,6 +286,8 @@ def mock_blob_container_client(monkeypatch):
         "AZURE_OPENAI_SERVICE": "test-openai-service",
         "AZURE_OPENAI_CHATGPT_DEPLOYMENT": "test-chatgpt",
         "AZURE_OPENAI_EMB_DEPLOYMENT": "test-ada",
+        "AZURE_OPENAI_EMB_MODEL_NAME": "text-embedding-3-large",
+        "AZURE_OPENAI_EMB_DIMENSIONS": "3072",
         "AZURE_USE_AUTHENTICATION": "true",
         "AZURE_ENABLE_GLOBAL_DOCUMENT_ACCESS": "true",
         "AZURE_ENABLE_UNAUTHENTICATED_ACCESS": "true",

tests/test_app_config.py

@@ -16,6 +16,8 @@ def minimal_env(monkeypatch):
         monkeypatch.setenv("AZURE_SEARCH_SERVICE", "test-search-service")
         monkeypatch.setenv("AZURE_OPENAI_SERVICE", "test-openai-service")
         monkeypatch.setenv("AZURE_OPENAI_CHATGPT_MODEL", "gpt-4o-mini")
+        monkeypatch.setenv("AZURE_OPENAI_EMB_MODEL_NAME", "text-embedding-3-large")
+        monkeypatch.setenv("AZURE_OPENAI_EMB_DIMENSIONS", 3072)
         yield
 
 

tests/test_chatapproach.py

@@ -30,6 +30,7 @@ def chat_approach():
         embedding_deployment="embeddings",
         embedding_model=MOCK_EMBEDDING_MODEL_NAME,
         embedding_dimensions=MOCK_EMBEDDING_DIMENSIONS,
+        embedding_field="embedding3",
         sourcepage_field="",
         content_field="",
         query_language="en-us",
@@ -176,6 +177,7 @@ async def test_search_results_filtering_by_scores(
         embedding_deployment="embeddings",
         embedding_model=MOCK_EMBEDDING_MODEL_NAME,
         embedding_dimensions=MOCK_EMBEDDING_DIMENSIONS,
+        embedding_field="embedding3",
         sourcepage_field="",
         content_field="",
         query_language="en-us",
@@ -214,6 +216,7 @@ async def test_search_results_query_rewriting(monkeypatch):
         embedding_deployment="embeddings",
         embedding_model=MOCK_EMBEDDING_MODEL_NAME,
         embedding_dimensions=MOCK_EMBEDDING_DIMENSIONS,
+        embedding_field="embedding3",
         sourcepage_field="",
         content_field="",
         query_language="en-us",

tests/test_chatvisionapproach.py

@@ -60,6 +60,7 @@ def chat_approach(openai_client, mock_confidential_client_success):
         embedding_deployment="embeddings",
         embedding_model=MOCK_EMBEDDING_MODEL_NAME,
         embedding_dimensions=MOCK_EMBEDDING_DIMENSIONS,
+        embedding_field="embedding3",
         sourcepage_field="",
         content_field="",
         query_language="en-us",
@@ -149,4 +150,4 @@ async def test_compute_text_embedding(chat_approach, openai_client, mock_openai_
     assert isinstance(result, VectorizedQuery)
     assert result.vector == [0.0023064255, -0.009327292, -0.0028842222]
     assert result.k_nearest_neighbors == 50
-    assert result.fields == "embedding"
+    assert result.fields == "embedding3"

tests/test_searchmanager.py

@@ -50,11 +50,16 @@ async def mock_list_index_names(self):
     monkeypatch.setattr(SearchIndexClient, "create_index", mock_create_index)
     monkeypatch.setattr(SearchIndexClient, "list_index_names", mock_list_index_names)
 
-    manager = SearchManager(search_info)
+    manager = SearchManager(
+        search_info,
+        use_int_vectorization=False,
+        field_name_embedding="embedding",
+        field_name_image_embedding="imageEmbedding",
+    )
     await manager.create_index()
     assert len(indexes) == 1, "It should have created one index"
     assert indexes[0].name == "test"
-    assert len(indexes[0].fields) == 7
+    assert len(indexes[0].fields) == 6
 
 
 @pytest.mark.asyncio
@@ -71,11 +76,16 @@ async def mock_list_index_names(self):
     monkeypatch.setattr(SearchIndexClient, "create_index", mock_create_index)
     monkeypatch.setattr(SearchIndexClient, "list_index_names", mock_list_index_names)
 
-    manager = SearchManager(search_info, use_int_vectorization=True)
+    manager = SearchManager(
+        search_info,
+        use_int_vectorization=True,
+        field_name_embedding="embedding",
+        field_name_image_embedding="image_embedding",
+    )
     await manager.create_index()
     assert len(indexes) == 1, "It should have created one index"
     assert indexes[0].name == "test"
-    assert len(indexes[0].fields) == 8
+    assert len(indexes[0].fields) == 7
 
 
 @pytest.mark.asyncio
@@ -165,11 +175,13 @@ async def mock_list_index_names(self):
     manager = SearchManager(
         search_info,
         use_acls=True,
+        field_name_embedding="embedding",
+        field_name_image_embedding="image_embedding",
     )
     await manager.create_index()
     assert len(indexes) == 1, "It should have created one index"
     assert indexes[0].name == "test"
-    assert len(indexes[0].fields) == 9
+    assert len(indexes[0].fields) == 8
 
 
 @pytest.mark.asyncio
@@ -283,6 +295,8 @@ async def mock_upload_documents(self, documents):
     manager = SearchManager(
         search_info,
         embeddings=embeddings,
+        field_name_embedding="embedding3",
+        field_name_image_embedding="image_embedding",
     )
 
     test_io = io.BytesIO(b"test content")
@@ -303,7 +317,7 @@ async def mock_upload_documents(self, documents):
     )
 
     assert len(documents_uploaded) == 1, "It should have uploaded one document"
-    assert documents_uploaded[0]["embedding"] == [
+    assert documents_uploaded[0]["embedding3"] == [
         0.0023064255,
         -0.009327292,
         -0.0028842222,