fix(azure): gate live tests, add offline coverage, fix api_config and retry errors

.github/workflows/test.yml

@@ -41,6 +41,12 @@ jobs:
           COHERE_API_KEY: ${{ secrets.COHERE_API_KEY }}
           MISTRAL_API_KEY: ${{ secrets.MISTRAL_API_KEY }}
           VOYAGE_API_KEY: ${{ secrets.VOYAGE_API_KEY }}
+          # The live Azure OpenAI tests gate themselves on AZURE_OPENAI_ENDPOINT,
+          # AZURE_OPENAI_API_KEY and OPENAI_API_VERSION: if any of the three is empty
+          # or unset they skip, so an absent secret disables them and re-adding it
+          # re-enables them with no code change. These lines stay wired up for that
+          # reason. (test_non_supported_dtypes is deliberately left ungated -- it
+          # validates dtype before any client is built and needs no network.)
           AZURE_OPENAI_API_KEY: ${{ secrets.AZURE_OPENAI_API_KEY }}
           AZURE_OPENAI_ENDPOINT: ${{ secrets.AZURE_OPENAI_ENDPOINT }}
           AZURE_OPENAI_DEPLOYMENT_NAME: ${{ secrets.AZURE_OPENAI_DEPLOYMENT_NAME }}
@@ -192,6 +198,12 @@ jobs:
           COHERE_API_KEY: ${{ secrets.COHERE_API_KEY }}
           MISTRAL_API_KEY: ${{ secrets.MISTRAL_API_KEY }}
           VOYAGE_API_KEY: ${{ secrets.VOYAGE_API_KEY }}
+          # The Azure cells in docs/user_guide/04_vectorizers.ipynb carry an
+          # unconditional `# NBVAL_SKIP`, so notebook validation skips them whether or
+          # not these secrets are set, and never blocks on getpass for missing
+          # credentials. Restoring the secrets is therefore not enough on its own --
+          # those markers have to be removed too. These lines stay wired up so that
+          # re-enabling needs no workflow change.
           AZURE_OPENAI_API_KEY: ${{ secrets.AZURE_OPENAI_API_KEY }}
           AZURE_OPENAI_ENDPOINT: ${{ secrets.AZURE_OPENAI_ENDPOINT }}
           AZURE_OPENAI_DEPLOYMENT_NAME: ${{ secrets.AZURE_OPENAI_DEPLOYMENT_NAME }}

docs/user_guide/04_vectorizers.ipynb

@@ -192,13 +192,14 @@
       "metadata": {},
       "outputs": [],
       "source": [
+        "# NBVAL_SKIP\n",
         "# additionally to the API Key, setup the API endpoint and version\n",
         "api_key = os.environ.get(\"AZURE_OPENAI_API_KEY\") or getpass.getpass(\"Enter your AzureOpenAI API key: \")\n",
         "api_version = os.environ.get(\"OPENAI_API_VERSION\") or getpass.getpass(\"Enter your AzureOpenAI API version: \")\n",
         "azure_endpoint = os.environ.get(\"AZURE_OPENAI_ENDPOINT\") or getpass.getpass(\"Enter your AzureOpenAI API endpoint: \")\n",
         "deployment_name = os.environ.get(\"AZURE_OPENAI_DEPLOYMENT_NAME\", \"text-embedding-ada-002\")\n",
         "\n",
-        "# Skip Azure examples when required values are missing (e.g. CI or Run All without Azure).\n",
+        "# Skip Azure examples when required values are missing (e.g. Run All without Azure configured).\n",
         "_azure_configured = bool(azure_endpoint and api_key and api_version)\n"
       ]
     },
@@ -208,6 +209,8 @@
       "metadata": {},
       "outputs": [],
       "source": [
+        "# NBVAL_SKIP\n",
+        "# Depends on the Azure OpenAI cell above, which is not executed in CI.\n",
         "from redisvl.utils.vectorize import AzureOpenAITextVectorizer\n",
         "\n",
         "if not _azure_configured:\n",
@@ -234,6 +237,8 @@
       "metadata": {},
       "outputs": [],
       "source": [
+        "# NBVAL_SKIP\n",
+        "# Depends on the Azure OpenAI cell above, which is not executed in CI.\n",
         "# Just like OpenAI, AzureOpenAI supports batching embeddings and asynchronous requests.\n",
         "sentences = [\n",
         "    \"That is a happy dog\",\n",

redisvl/utils/vectorize/text/azureopenai.py

@@ -129,8 +129,8 @@ def _initialize_clients(self, api_config: dict[str, Any] | None, **kwargs):
             ImportError: If the openai library is not installed
             ValueError: If required parameters are not provided
         """
-        if api_config is None:
-            api_config = {}
+        # Copy so we never mutate the caller's dict.
+        api_config = dict(api_config or {})
 
         # Dynamic import of the openai module
         try:
@@ -142,10 +142,8 @@ def _initialize_clients(self, api_config: dict[str, Any] | None, **kwargs):
             )
 
         # Fetch the API key, version and endpoint from api_config or environment variable
-        azure_endpoint = (
-            api_config.pop("azure_endpoint")
-            if api_config
-            else os.getenv("AZURE_OPENAI_ENDPOINT")
+        azure_endpoint = api_config.pop("azure_endpoint", None) or os.getenv(
+            "AZURE_OPENAI_ENDPOINT"
         )
 
         if not azure_endpoint:
@@ -154,10 +152,8 @@ def _initialize_clients(self, api_config: dict[str, Any] | None, **kwargs):
                 "Provide it in api_config or set the AZURE_OPENAI_ENDPOINT environment variable."
             )
 
-        api_version = (
-            api_config.pop("api_version")
-            if api_config
-            else os.getenv("OPENAI_API_VERSION")
+        api_version = api_config.pop("api_version", None) or os.getenv(
+            "OPENAI_API_VERSION"
         )
 
         if not api_version:
@@ -166,11 +162,7 @@ def _initialize_clients(self, api_config: dict[str, Any] | None, **kwargs):
                 "Provide it in api_config or set the OPENAI_API_VERSION environment variable."
             )
 
-        api_key = (
-            api_config.pop("api_key")
-            if api_config
-            else os.getenv("AZURE_OPENAI_API_KEY")
-        )
+        api_key = api_config.pop("api_key", None) or os.getenv("AZURE_OPENAI_API_KEY")
 
         if not api_key:
             raise ValueError(
@@ -219,6 +211,7 @@ def _set_model_dims(self) -> int:
         wait=wait_random_exponential(min=1, max=60),
         stop=stop_after_attempt(6),
         retry=retry_if_not_exception_type(TypeError),
+        reraise=True,
     )
     def _embed(self, content: str = "", text: str = "", **kwargs) -> list[float]:
         """
@@ -253,6 +246,7 @@ def _embed(self, content: str = "", text: str = "", **kwargs) -> list[float]:
         wait=wait_random_exponential(min=1, max=60),
         stop=stop_after_attempt(6),
         retry=retry_if_not_exception_type(TypeError),
+        reraise=True,
     )
     def _embed_many(
         self,
@@ -299,6 +293,7 @@ def _embed_many(
         wait=wait_random_exponential(min=1, max=60),
         stop=stop_after_attempt(6),
         retry=retry_if_not_exception_type(TypeError),
+        reraise=True,
     )
     async def _aembed(self, content: str = "", text: str = "", **kwargs) -> list[float]:
         """
@@ -333,6 +328,7 @@ async def _aembed(self, content: str = "", text: str = "", **kwargs) -> list[flo
         wait=wait_random_exponential(min=1, max=60),
         stop=stop_after_attempt(6),
         retry=retry_if_not_exception_type(TypeError),
+        reraise=True,
     )
     async def _aembed_many(
         self,

tests/integration/test_vectorizers.py

@@ -39,13 +39,33 @@ def embeddings_cache(client):
     cache.clear()
 
 
+# Azure OpenAI live tests need a reachable deployment. _initialize_clients()
+# requires all three of these before it will construct a client, so gate on all
+# three rather than guessing from one. AZURE_OPENAI_DEPLOYMENT_NAME is
+# deliberately excluded -- it has a real default at every call site below.
+_AZURE_CONFIGURED = all(
+    os.getenv(var)
+    for var in ("AZURE_OPENAI_ENDPOINT", "AZURE_OPENAI_API_KEY", "OPENAI_API_VERSION")
+)
+skip_without_azure = pytest.mark.skipif(
+    not _AZURE_CONFIGURED,
+    reason=(
+        "Azure OpenAI is not configured. Set AZURE_OPENAI_ENDPOINT, "
+        "AZURE_OPENAI_API_KEY and OPENAI_API_VERSION to run these, plus "
+        "AZURE_OPENAI_DEPLOYMENT_NAME if your deployment is not named "
+        "text-embedding-ada-002. Offline coverage lives in "
+        "tests/unit/test_azure_openai_vectorizer.py."
+    ),
+)
+
+
 _vectorizer_params = [
     pytest.param(HFTextVectorizer, marks=pytest.mark.requires_hf),
     OpenAITextVectorizer,
     VertexAIVectorizer,
     GoogleGenAIVectorizer,
     CohereTextVectorizer,
-    AzureOpenAITextVectorizer,
+    pytest.param(AzureOpenAITextVectorizer, marks=skip_without_azure),
     BedrockVectorizer,
     MistralAITextVectorizer,
     CustomVectorizer,
@@ -416,7 +436,7 @@ def bad_return_type(text: str) -> str:
 
 
 _dtype_params = [
-    AzureOpenAITextVectorizer,
+    pytest.param(AzureOpenAITextVectorizer, marks=skip_without_azure),
     BedrockVectorizer,
     CohereTextVectorizer,
     CustomVectorizer,