Tests for MultipartFormDataHelper and OpenAISyncClientTest #44811
Tests for MultipartFormDataHelper and OpenAISyncClientTest #44811gregocmsft wants to merge 1 commit intoAzure:mainfrom
Conversation
+Tests for OpenAISyncClientTest - ChatCompletionsStreamOptions
github-actions
bot
added
Community Contribution
|
Thank you for your contribution @gregocmsft! We will review the pull request and get back to you soon. |
|
API change check API changes are not detected in this pull request. |
jpalvarezl
left a comment
jpalvarezl
left a comment
There was a problem hiding this comment.
This is great! Thank you. I think you just need to add the analogous tests I mentioned in the PR comment and you are good to go!
| @ParameterizedTest(name = DISPLAY_NAME_WITH_ARGUMENTS) | ||
| @MethodSource("com.azure.ai.openai.TestUtils#getTestParameters") | ||
| public void testGetCompletionsStreamChatCompletionStreamOptions(HttpClient httpClient, | ||
| OpenAIServiceVersion serviceVersion) { | ||
| client = getOpenAIClient(httpClient, serviceVersion); | ||
| getCompletionsRunner((deploymentId, prompt) -> { | ||
| IterableStream<Completions> resultCompletions = client.getCompletionsStream(deploymentId, | ||
| new CompletionsOptions(prompt), new ChatCompletionStreamOptions().setIncludeUsage(true)); | ||
|
|
||
| Object[] result = resultCompletions.stream().toArray(); | ||
| Completions[] completionsArray = Arrays.copyOf(result, result.length, Completions[].class); | ||
|
|
||
| assertTrue(completionsArray.length > 1); | ||
| // First element returns the prompt filter results (no output tokens are present) | ||
| assertFalse(CoreUtils.isNullOrEmpty(completionsArray[0].getPromptFilterResults())); | ||
| // Choices (output tokens) are present in all the elements in between | ||
| for (int i = 1; i < completionsArray.length - 2; i++) { | ||
| assertCompletionsStream(completionsArray[i]); | ||
| } | ||
|
|
||
| // Last element returns the completion tokens (no output tokens are present) | ||
| assertNotNull(completionsArray[completionsArray.length - 1].getUsage()); | ||
| }); | ||
| } | ||
|
|
||
| @ParameterizedTest(name = DISPLAY_NAME_WITH_ARGUMENTS) | ||
| @MethodSource("com.azure.ai.openai.TestUtils#getTestParameters") | ||
| public void testGetChatCompletionsStreamChatCompletionStreamOptionsUsageTrue(HttpClient httpClient, | ||
| OpenAIServiceVersion serviceVersion) { | ||
| client = getOpenAIClient(httpClient, serviceVersion); | ||
| getChatCompletionsStreamUsageRunner((deploymentId, chatCompletionsOptions) -> { | ||
| IterableStream<ChatCompletions> resultChatCompletions = client.getChatCompletionsStream(deploymentId, | ||
| chatCompletionsOptions, new ChatCompletionStreamOptions().setIncludeUsage(true)); | ||
| assertChatCompletionStreamUsage(resultChatCompletions.stream().collect(Collectors.toList())); | ||
| }); | ||
| } | ||
|
|
||
| @ParameterizedTest(name = DISPLAY_NAME_WITH_ARGUMENTS) | ||
| @MethodSource("com.azure.ai.openai.TestUtils#getTestParameters") | ||
| public void testGetChatCompletionsStreamChatCompletionStreamOptionsUsageFalse(HttpClient httpClient, | ||
| OpenAIServiceVersion serviceVersion) { | ||
| client = getOpenAIClient(httpClient, serviceVersion); | ||
| getChatCompletionsRunner((deploymentId, chatMessages) -> { | ||
| IterableStream<ChatCompletions> resultChatCompletions = client.getChatCompletionsStream(deploymentId, | ||
| new ChatCompletionsOptions(chatMessages), new ChatCompletionStreamOptions().setIncludeUsage(false)); | ||
| assertTrue(resultChatCompletions.stream().toArray().length > 1); | ||
| resultChatCompletions.forEach(OpenAIClientTestBase::assertChatCompletionsStream); | ||
| }); | ||
| } |
There was a problem hiding this comment.
Good stuff! I think that all these tests are good. The only thing left to do would be to write the version for the NonAzureOpenAISyncClientTest (which is just copy and pasting this, adjusting the client initialization. After that, adding the async version for both Azure and Non-Azure and you are good to go!
There was a problem hiding this comment.
Hi! I'm not sure how to generate the versions since I have not used any client initialization for these unit tests. I only worked with the tools of MultipartFormDataHelper but the tests do not interact with the API so no client initialization was needed.
Edit: I see what you're saying but I'm not sure why these tests in particular are here since I did not develop them. I will verify with the team what could happen.
|
The CI failures I see are mostly related to the test recording being missing. Once you've added the other tests, give me a heads up and I will add the recording to this PR so we can move it forward. I think you might also have style check issues, although if you are able to run |
3 participants
+Tests for MultipartFormDataHelper
+Tests for OpenAISyncClientTest - ChatCompletionsStreamOptions