Azure/azure-sdk-for-net Azure.AI.OpenAI_1.0.0-beta.14

on GitHub

1.0.0-beta.14 (2024-03-04)

Features Added

• Text-to-speech using OpenAI TTS models is now supported. See OpenAI's API reference or the Azure OpenAI quickstart for detailed overview and background information.
  • The new method GenerateSpeechFromText exposes this capability on OpenAIClient.
  • Text-to-speech converts text into lifelike spoken audio in a chosen voice, together with other optional configurations.
  • This method works for both Azure OpenAI and non-Azure api.openai.com client configurations

Breaking Changes

"On Your Data" changes:

• Introduced a new type AzureChatExtensionDataSourceResponseCitation for a more structured representation of citation data.
• Correspondingly, updated AzureChatExtensionsMessageContext:
  • Replaced Messages with Citations of type AzureChatExtensionDataSourceResponseCitation.
  • Added Intent as a string type.
• Renamed "AzureCognitiveSearch" to "AzureSearch":
  • AzureCognitiveSearchChatExtensionConfiguration is now AzureSearchChatExtensionConfiguration.
  • AzureCognitiveSearchIndexFieldMappingOptions is now AzureSearchIndexFieldMappingOptions.
• Check the project README for updated code snippets.

Other Changes

• New properties in ChatCompletionsOptions:
  • EnableLogProbabilities: Allows retrieval of log probabilities (REST: logprobs)
  • LogProbabilitiesPerToken: The number of most likely tokens to return per token (REST: top_logprobs)
• Introduced a new property in CompletionsOptions:
  • Suffix: Defines the suffix that follows the completion of inserted text (REST: suffix)
• Image generation response now includes content filtering details (specific to Azure OpenAI endpoint):
  • ImageGenerationData.ContentFilterResults: Information about the content filtering results. (REST: content_filter_results)
  • ImageGenerationData.PromptFilterResults: Information about the content filtering category (REST: prompt_filter_results)