Update tests

Author: pamelafox

app/backend/app.py

@@ -471,7 +471,6 @@ async def setup_clients():
     USE_CHAT_HISTORY_BROWSER = os.getenv("USE_CHAT_HISTORY_BROWSER", "").lower() == "true"
     USE_CHAT_HISTORY_COSMOS = os.getenv("USE_CHAT_HISTORY_COSMOS", "").lower() == "true"
     USE_AGENTIC_RETRIEVAL = os.getenv("USE_AGENTIC_RETRIEVAL", "").lower() == "true"
-    # TODO: ENABLE_AGENTIC_RETRIEVAL_SOURCE_DATA = os.getenv("ENABLE_AGENTIC_RETRIEVAL_SOURCE_DATA", "").lower() == "true"
 
     # WEBSITE_HOSTNAME is always set by App Service, RUNNING_IN_PRODUCTION is set in main.bicep
     RUNNING_ON_AZURE = os.getenv("WEBSITE_HOSTNAME") is not None or os.getenv("RUNNING_IN_PRODUCTION") is not None

app/backend/approaches/approach.py

@@ -273,15 +273,15 @@ async def run_agentic_retrieval(
         search_index_name: str,
         top: Optional[int] = None,
         filter_add_on: Optional[str] = None,
+        minimum_reranker_score: Optional[float] = None,
         results_merge_strategy: Optional[str] = None,
     ) -> tuple[KnowledgeAgentRetrievalResponse, list[Document]]:
         # STEP 1: Invoke agentic retrieval
         response = await agent_client.retrieve(
             retrieval_request=KnowledgeAgentRetrievalRequest(
                 messages=[
                     KnowledgeAgentMessage(
-                        role=str(msg["role"]),
-                        content=[KnowledgeAgentMessageTextContent(text=str(msg["content"]))],
+                        role=str(msg["role"]), content=[KnowledgeAgentMessageTextContent(text=str(msg["content"]))]
                     )
                     for msg in messages
                     if msg["role"] != "system"
@@ -341,6 +341,9 @@ async def run_agentic_retrieval(
                 if top and len(documents) >= top:
                     break
 
+        if minimum_reranker_score is not None:
+            documents = [doc for doc in documents if (doc.reranker_score or 0) >= minimum_reranker_score]
+
         if results_merge_strategy == "interleaved":
             documents = sorted(
                 documents,

app/backend/approaches/chatreadretrieveread.py

@@ -389,6 +389,7 @@ async def run_agentic_retrieval_approach(
         auth_claims: dict[str, Any],
     ):
         search_index_filter = self.build_filter(overrides, auth_claims)
+        minimum_reranker_score = overrides.get("minimum_reranker_score", 0)
         top = overrides.get("top", 3)
         results_merge_strategy = overrides.get("results_merge_strategy", "interleaved")
         send_text_sources = overrides.get("send_text_sources", True)
@@ -400,6 +401,7 @@ async def run_agentic_retrieval_approach(
             search_index_name=self.search_index_name,
             top=top,
             filter_add_on=search_index_filter,
+            minimum_reranker_score=minimum_reranker_score,
             results_merge_strategy=results_merge_strategy,
         )
 
@@ -417,6 +419,7 @@ async def run_agentic_retrieval_approach(
                     "Use agentic retrieval",
                     messages,
                     {
+                        "reranker_threshold": minimum_reranker_score,
                         "results_merge_strategy": results_merge_strategy,
                         "filter": search_index_filter,
                     },

app/backend/approaches/retrievethenread.py

@@ -224,6 +224,7 @@ async def run_agentic_retrieval_approach(
         overrides: dict[str, Any],
         auth_claims: dict[str, Any],
     ) -> ExtraInfo:
+        minimum_reranker_score = overrides.get("minimum_reranker_score", 0)
         search_index_filter = self.build_filter(overrides, auth_claims)
         top = overrides.get("top", 3)
         results_merge_strategy = overrides.get("results_merge_strategy", "interleaved")
@@ -236,6 +237,7 @@ async def run_agentic_retrieval_approach(
             search_index_name=self.search_index_name,
             top=top,
             filter_add_on=search_index_filter,
+            minimum_reranker_score=minimum_reranker_score,
             results_merge_strategy=results_merge_strategy,
         )
 
@@ -254,6 +256,7 @@ async def run_agentic_retrieval_approach(
                     "Use agentic retrieval",
                     messages,
                     {
+                        "reranker_threshold": minimum_reranker_score,
                         "results_merge_strategy": results_merge_strategy,
                         "filter": search_index_filter,
                     },

app/backend/prepdocslib/searchmanager.py

@@ -462,10 +462,7 @@ async def create_agent(self):
                         name=self.search_info.agent_name,
                         knowledge_sources=[
                             KnowledgeSourceReference(
-                                name=knowledge_source.name,
-                                include_reference_source_data=True,  # TODO: base on env var
-                                include_references=True,
-                                reranker_threshold=1.5,
+                                name=knowledge_source.name, include_references=True, include_reference_source_data=True
                             )
                         ],
                         models=[

docs/agentic_retrieval.md

@@ -34,27 +34,15 @@ See the agentic retrieval documentation.
    azd env set AZURE_OPENAI_SEARCHAGENT_MODEL_VERSION 2025-04-14
    ```
 
-3. **(Optional) Enable extra field hydration**
-
-   By default, agentic retrieval only returns fields included in the semantic configuration.
-
-   You can enable this optional feature below, to include all fields from the search index in the result.
-   ⚠️ This feature is currently only compatible with indexes set up with integrated vectorization,
-   or indexes that otherwise have an "id" field marked as filterable.
-
-   ```shell
-   azd env set ENABLE_AGENTIC_RETRIEVAL_SOURCE_DATA true
-   ```
-
-4. **Update the infrastructure and application:**
+3. **Update the infrastructure and application:**
 
    Execute `azd up` to provision the infrastructure changes (only the new model, if you ran `up` previously) and deploy the application code with the updated environment variables.
 
-5. **Try out the feature:**
+4. **Try out the feature:**
 
    Open the web app and start a new chat. Agentic retrieval will be used to find all sources.
 
-6. **Review the query plan**
+5. **Review the query plan**
 
    Agentic retrieval use additional billed tokens behind the scenes for the planning process.
    To see the token usage, select the lightbulb icon on a chat answer. This will open the "Thought process" tab, which shows the amount of tokens used by and the queries produced by the planning process

infra/main.bicep

@@ -41,7 +41,6 @@ param storageSkuName string // Set in main.parameters.json
 
 param defaultReasoningEffort string // Set in main.parameters.json
 param useAgenticRetrieval bool // Set in main.parameters.json
-param enableAgenticRetrievalSourceData bool // Set in main.parameters.json
 
 param userStorageAccountName string = ''
 param userStorageContainerName string = 'user-content'
@@ -424,7 +423,6 @@ var appEnvVariables = {
   USE_SPEECH_OUTPUT_BROWSER: useSpeechOutputBrowser
   USE_SPEECH_OUTPUT_AZURE: useSpeechOutputAzure
   USE_AGENTIC_RETRIEVAL: useAgenticRetrieval
-  ENABLE_AGENTIC_RETRIEVAL_SOURCE_DATA: enableAgenticRetrievalSourceData
   // Chat history settings
   USE_CHAT_HISTORY_BROWSER: useChatHistoryBrowser
   USE_CHAT_HISTORY_COSMOS: useChatHistoryCosmos

infra/main.parameters.json

@@ -338,9 +338,6 @@
     "useAgenticRetrieval": {
       "value": "${USE_AGENTIC_RETRIEVAL=false}"
     },
-    "enableAgenticRetrievalSourceData": {
-      "value": "${ENABLE_AGENTIC_RETRIEVAL_SOURCE_DATA=false}"
-    },
     "ragSearchTextEmbeddings": {
       "value": "${RAG_SEARCH_TEXT_EMBEDDINGS=true}"
     },

tests/conftest.py

@@ -49,8 +49,6 @@
     MockResponse,
     MockTransport,
     mock_retrieval_response,
-    mock_retrieval_response_with_duplicates,
-    mock_retrieval_response_with_missing_doc_key,
     mock_retrieval_response_with_sorting,
     mock_retrieval_response_with_top_limit,
     mock_speak_text_cancelled,
@@ -90,31 +88,23 @@ async def mock_search(self, *args, **kwargs):
 def create_mock_retrieve(response_type="default"):
     """Create a mock_retrieve function that returns different response types.
 
-    Args:
-        response_type: Type of response to return. Options:
-            - "default": mock_retrieval_response()
-            - "sorting": mock_retrieval_response_with_sorting()
-            - "duplicates": mock_retrieval_response_with_duplicates()
-            - "missing_doc_key": mock_retrieval_response_with_missing_doc_key()
-            - "top_limit": mock_retrieval_response_with_top_limit()
+    Supported response_type values:
+      - "default": single reference response
+      - "sorting": multiple refs to test ordering / interleaving
+      - "top_limit": many refs to test early breaking via top limit
     """
 
     async def mock_retrieve_parameterized(self, *args, **kwargs):
         retrieval_request = kwargs.get("retrieval_request")
         assert retrieval_request is not None
-        # New API uses knowledge_sources instead of target_index_params
         assert retrieval_request.knowledge_source_params is not None
         assert len(retrieval_request.knowledge_source_params) == 1
-        params = retrieval_request.knowledge_source_params
-        # parameters may have filter_add_on attribute in new SDK
+        params_list = retrieval_request.knowledge_source_params
+        params = params_list[0]
         self.filter = getattr(params, "filter_add_on", None)
 
         if response_type == "sorting":
             return mock_retrieval_response_with_sorting()
-        elif response_type == "duplicates":
-            return mock_retrieval_response_with_duplicates()
-        elif response_type == "missing_doc_key":
-            return mock_retrieval_response_with_missing_doc_key()
         elif response_type == "top_limit":
             return mock_retrieval_response_with_top_limit()
         else:  # default
@@ -465,7 +455,6 @@ async def mock_exists(*args, **kwargs):
         "AZURE_OPENAI_SEARCHAGENT_MODEL": "gpt-4.1-mini",
         "AZURE_OPENAI_SEARCHAGENT_DEPLOYMENT": "gpt-4.1-mini",
         "USE_AGENTIC_RETRIEVAL": "true",
-        "ENABLE_AGENTIC_RETRIEVAL_SOURCE_DATA": "true",
     }
 ]
 
@@ -479,7 +468,6 @@ async def mock_exists(*args, **kwargs):
         "AZURE_OPENAI_SEARCHAGENT_MODEL": "gpt-4.1-mini",
         "AZURE_OPENAI_SEARCHAGENT_DEPLOYMENT": "gpt-4.1-mini",
         "USE_AGENTIC_RETRIEVAL": "true",
-        "ENABLE_AGENTIC_RETRIEVAL_SOURCE_DATA": "true",
         "AZURE_USE_AUTHENTICATION": "true",
         "AZURE_SERVER_APP_ID": "SERVER_APP",
         "AZURE_SERVER_APP_SECRET": "SECRET",
@@ -1178,37 +1166,3 @@ def chat_approach():
             credential=MockAzureCredential(),
         ),
     )
-
-
-@pytest.fixture
-def chat_approach_with_hydration():
-    return ChatReadRetrieveReadApproach(
-        search_client=SearchClient(endpoint="", index_name="", credential=AzureKeyCredential("")),
-        search_index_name=None,
-        agent_model=None,
-        agent_deployment=None,
-        agent_client=None,
-        auth_helper=None,
-        openai_client=None,
-        chatgpt_model="gpt-4.1-mini",
-        chatgpt_deployment="chat",
-        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",
-        query_speller="lexicon",
-        prompt_manager=PromptyManager(),
-        user_blob_manager=AdlsBlobManager(
-            endpoint="https://test-userstorage-account.dfs.core.windows.net",
-            container="test-userstorage-container",
-            credential=MockAzureCredential(),
-        ),
-        global_blob_manager=BlobManager(  # on normal Azure storage
-            endpoint="https://test-globalstorage-account.blob.core.windows.net",
-            container="test-globalstorage-container",
-            credential=MockAzureCredential(),
-        ),
-    )