.Net: MEVD Built in embedding generation throws when used with AzureOpenAI

[westey-m]

Describe the bug
MEVD Built in embedding generation throws when used with AzureOpenAI with This model does not support specifying dimensions.

To Reproduce
Find RAGWithInMemoryVectorStoreAndPluginAsync sample in samples

Update text embedding generator to azure open ai:

        var textEmbeddingGenerator = new AzureOpenAIClient(new Uri(TestConfiguration.AzureOpenAIEmbeddings.Endpoint), new AzureCliCredential())
            .GetEmbeddingClient(TestConfiguration.AzureOpenAIEmbeddings.DeploymentName)
            .AsIEmbeddingGenerator();

Sample fails:

    System.ClientModel.ClientResultException : HTTP 400 (invalid_request_error: )
    
    This model does not support specifying dimensions.

  Stack Trace: 
    ClientPipelineExtensions.ProcessMessageAsync(ClientPipeline pipeline, PipelineMessage message, RequestOptions options)
    EmbeddingClient.GenerateEmbeddingsAsync(BinaryContent content, RequestOptions options)
    EmbeddingClient.GenerateEmbeddingsAsync(IEnumerable`1 inputs, EmbeddingGenerationOptions options, CancellationToken cancellationToken)
    OpenAIEmbeddingGenerator.GenerateAsync(IEnumerable`1 values, EmbeddingGenerationOptions options, CancellationToken cancellationToken)
    InMemoryVectorStoreRecordCollection`2.UpsertAsync(IEnumerable`1 records, CancellationToken cancellationToken) line 280
    WithPlugins.RAGWithInMemoryVectorStoreAndPluginAsync() line 62```

**Expected behavior**
Sample can generate embeddings

[DavidMarquezF]

Any workarounds for now?

[roji]

@DavidMarquezF we're addressing this for the release next so, so hang tight and things should just work very soon. Otherwise, to get unblocked you can simply generate embeddings yourself outside of Microsoft.Extensions.VectorData and pass those in.