Update available Azure API versions

Author: ccreutzi

+llms/+azure/apiVersions.m

@@ -3,11 +3,17 @@
 
 %   Copyright 2024 The MathWorks, Inc.
     versions = [...
+        "2025-02-01-preview", ...
+        "2025-01-01-preview", ...
+        "2024-12-01-preview", ...
+        "2024-10-01-preview", ...
+        "2024-09-01-preview", ...
         "2024-08-01-preview", ...
         "2024-07-01-preview", ...
         "2024-05-01-preview", ...
         "2024-04-01-preview", ...
         "2024-03-01-preview", ...
+        "2024-10-21", ...
         "2024-06-01", ...
         "2024-02-01", ...
         "2023-05-15", ...

azureChat.m

@@ -107,7 +107,7 @@
                 nvp.DeploymentID             (1,1) string {llms.utils.mustBeNonzeroLengthTextScalar}
                 nvp.APIKey                         {llms.utils.mustBeNonzeroLengthTextScalar}
                 nvp.Tools                    (1,:) {mustBeA(nvp.Tools, "openAIFunction")} = openAIFunction.empty
-                nvp.APIVersion               (1,1) string {mustBeAPIVersion} = "2024-06-01"
+                nvp.APIVersion               (1,1) string {mustBeAPIVersion} = "2024-10-21"
                 nvp.Temperature                    {llms.utils.mustBeValidTemperature} = 1
                 nvp.TopP                           {llms.utils.mustBeValidProbability} = 1
                 nvp.StopSequences                  {llms.utils.mustBeValidStop} = {}

doc/Azure.md

@@ -250,10 +250,10 @@ Create a sample output structure.
 sampleOutput = struct("Animal","Penguin","CanFly",false);
 ```
 
-Connect to Azure OpenAI Service. Use an API version that supports structured outputs.
+Connect to Azure OpenAI Service.
 
 ```matlab
-model = azureChat(APIVersion="2024-08-01-preview");
+model = azureChat;
 ```
 
 Generate structured output.

doc/functions/azureChat.md

@@ -117,7 +117,7 @@ Instead of using the `DeploymentID` name\-value argument, you can also set the e
 
 ### `APIVersion` — API Version
 
-`"2024-06-01"` (default) | `"2024-02-01"` | `"2023-05-15"` | `"2024-05-01-preview"` | `"2024-04-01-preview"` | ...
+`"2024-10-21"` (default) | `"2024-06-01"` | `"2024-02-01"` | `"2025-02-01-preview"` | `"2025-01-01-preview"` | ...
 
 
 After construction, this property is read\-only.

tests/tazureChat.m

@@ -15,8 +15,7 @@
         constructor = @azureChat;
         defaultModel = azureChat;
         visionModel = azureChat(Deployment="gpt-4o");
-        structuredModel = azureChat("APIVersion","2024-08-01-preview",...
-            "Deployment","gpt-4o-2024-08-06");
+        structuredModel = azureChat("Deployment","gpt-4o-2024-08-06");
     end
 
     methods(Test)
@@ -63,7 +62,7 @@ function jsonFormatWithSystemPrompt(testCase)
         end
 
         function responseFormatRequiresNewAPI(testCase)
-            chat = azureChat;
+            chat = azureChat(APIVersion="2024-02-01");
             testCase.verifyError(@() generate(chat, ...
                 "What is the smallest prime?", ...
                 ResponseFormat=struct("number",1)), ...
@@ -127,7 +126,7 @@ function specialErrorForUnsupportedResponseFormat(testCase)
             testCase.assumeFail("Disabled until `llms.internal.callAzureChat` is updated to use `max_completion_tokens` instead of the deprecated `max_tokens` in the OpenAI API.")
 
             testCase.verifyError(@() generate(...
-                azureChat(APIVersion="2024-08-01-preview",DeploymentID="o1-mini"), ...
+                azureChat(DeploymentID="o1-mini"), ...
                 "What is the smallest prime?", ...
                 ResponseFormat=struct("number",1)), ...
                 "llms:noStructuredOutputForAzureDeployment");