Base api version on env var

Author: pamelafox

app/backend/app.py

@@ -573,7 +573,8 @@ async def setup_clients():
         current_app.config[CONFIG_CREDENTIAL] = azure_credential
 
     if OPENAI_HOST.startswith("azure"):
-        api_version = os.getenv("AZURE_OPENAI_API_VERSION") or "2024-03-01-preview"
+        # https://learn.microsoft.com/azure/ai-services/openai/api-version-deprecation#latest-ga-api-release
+        api_version = os.getenv("AZURE_OPENAI_API_VERSION") or "2024-06-01"
         if OPENAI_HOST == "azure_custom":
             current_app.logger.info("OPENAI_HOST is azure_custom, setting up Azure OpenAI custom client")
             if not AZURE_OPENAI_CUSTOM_URL:

app/backend/prepdocs.py

@@ -115,6 +115,7 @@ def setup_embeddings_service(
     openai_custom_url: Union[str, None],
     openai_deployment: Union[str, None],
     openai_dimensions: int,
+    openai_api_version: str,
     openai_key: Union[str, None],
     openai_org: Union[str, None],
     disable_vectors: bool = False,
@@ -134,6 +135,7 @@ def setup_embeddings_service(
             open_ai_deployment=openai_deployment,
             open_ai_model_name=openai_model_name,
             open_ai_dimensions=openai_dimensions,
+            open_ai_api_version=openai_api_version,
             credential=azure_open_ai_credential,
             disable_batch=disable_batch_vectors,
         )
@@ -366,6 +368,8 @@ async def main(strategy: Strategy, setup_index: bool = True):
         openai_service=os.getenv("AZURE_OPENAI_SERVICE"),
         openai_custom_url=os.getenv("AZURE_OPENAI_CUSTOM_URL"),
         openai_deployment=os.getenv("AZURE_OPENAI_EMB_DEPLOYMENT"),
+        # https://learn.microsoft.com/azure/ai-services/openai/api-version-deprecation#latest-ga-api-release
+        openai_api_version=os.getenv("AZURE_OPENAI_API_VERSION") or "2024-06-01",
         openai_dimensions=openai_dimensions,
         openai_key=clean_key_if_exists(openai_key),
         openai_org=os.getenv("OPENAI_ORGANIZATION"),

app/backend/prepdocslib/embeddings.py

@@ -163,6 +163,7 @@ def __init__(
         open_ai_deployment: Union[str, None],
         open_ai_model_name: str,
         open_ai_dimensions: int,
+        open_ai_api_version: str,
         credential: Union[AsyncTokenCredential, AzureKeyCredential],
         open_ai_custom_url: Union[str, None] = None,
         disable_batch: bool = False,
@@ -176,6 +177,7 @@ def __init__(
         else:
             raise ValueError("Either open_ai_service or open_ai_custom_url must be provided")
         self.open_ai_deployment = open_ai_deployment
+        self.open_ai_api_version = open_ai_api_version
         self.credential = credential
 
     async def create_client(self) -> AsyncOpenAI:
@@ -196,7 +198,7 @@ class AuthArgs(TypedDict, total=False):
         return AsyncAzureOpenAI(
             azure_endpoint=self.open_ai_endpoint,
             azure_deployment=self.open_ai_deployment,
-            api_version="2023-05-15",
+            api_version=self.open_ai_api_version,
             **auth_args,
         )
 

tests/test_prepdocs.py

@@ -62,6 +62,7 @@ async def mock_create_client(*args, **kwargs):
         open_ai_deployment="x",
         open_ai_model_name=MOCK_EMBEDDING_MODEL_NAME,
         open_ai_dimensions=MOCK_EMBEDDING_DIMENSIONS,
+        open_ai_api_version="test-api-version",
         credential=MockAzureCredential(),
         disable_batch=False,
     )
@@ -79,6 +80,7 @@ async def mock_create_client(*args, **kwargs):
         open_ai_deployment="x",
         open_ai_model_name=MOCK_EMBEDDING_MODEL_NAME,
         open_ai_dimensions=MOCK_EMBEDDING_DIMENSIONS,
+        open_ai_api_version="test-api-version",
         credential=MockAzureCredential(),
         disable_batch=True,
     )
@@ -149,6 +151,7 @@ async def test_compute_embedding_ratelimiterror_batch(monkeypatch, caplog):
                 open_ai_deployment="x",
                 open_ai_model_name=MOCK_EMBEDDING_MODEL_NAME,
                 open_ai_dimensions=MOCK_EMBEDDING_DIMENSIONS,
+                open_ai_api_version="test-api-version",
                 credential=MockAzureCredential(),
                 disable_batch=False,
             )
@@ -167,6 +170,7 @@ async def test_compute_embedding_ratelimiterror_single(monkeypatch, caplog):
                 open_ai_deployment="x",
                 open_ai_model_name=MOCK_EMBEDDING_MODEL_NAME,
                 open_ai_dimensions=MOCK_EMBEDDING_DIMENSIONS,
+                open_ai_api_version="test-api-version",
                 credential=MockAzureCredential(),
                 disable_batch=True,
             )
@@ -193,6 +197,7 @@ async def test_compute_embedding_autherror(monkeypatch, capsys):
             open_ai_deployment="x",
             open_ai_model_name=MOCK_EMBEDDING_MODEL_NAME,
             open_ai_dimensions=MOCK_EMBEDDING_DIMENSIONS,
+            open_ai_api_version="test-api-version",
             credential=MockAzureCredential(),
             disable_batch=False,
         )
@@ -205,6 +210,7 @@ async def test_compute_embedding_autherror(monkeypatch, capsys):
             open_ai_deployment="x",
             open_ai_model_name=MOCK_EMBEDDING_MODEL_NAME,
             open_ai_dimensions=MOCK_EMBEDDING_DIMENSIONS,
+            open_ai_api_version="test-api-version",
             credential=MockAzureCredential(),
             disable_batch=True,
         )

tests/test_searchmanager.py

@@ -275,6 +275,7 @@ async def mock_upload_documents(self, documents):
         open_ai_deployment="x",
         open_ai_model_name=MOCK_EMBEDDING_MODEL_NAME,
         open_ai_dimensions=MOCK_EMBEDDING_DIMENSIONS,
+        open_ai_api_version="test-api-version",
         credential=AzureKeyCredential("test"),
         disable_batch=True,
     )