Azure/azure-sdk-for-net Azure.AI.VoiceLive_1.1.0-beta.2

on GitHub

1.1.0-beta.2 (2026-02-19)

Features Added

• MCP (Model Context Protocol) Support: Added comprehensive support for MCP server integration

  • Added VoiceLiveMcpServerDefinition for configuring external MCP servers as tools
  • Added MCPApprovalType enum for controlling tool execution approval workflows ("always", "never")
  • Added VoiceLiveMcpTool class for representing MCP tool definitions with JSON schemas
  • Added MCP-specific session update events:
    • SessionUpdateMcpListToolsInProgress - Tool discovery started
    • SessionUpdateMcpListToolsCompleted - Tool discovery completed
    • SessionUpdateMcpListToolsFailed - Tool discovery failed
    • SessionUpdateResponseMcpCallArgumentsDelta - Tool call arguments streaming
    • SessionUpdateResponseMcpCallArgumentsDone - Tool call arguments complete
    • SessionUpdateResponseMcpCallInProgress - Tool execution started
    • SessionUpdateResponseMcpCallCompleted - Tool execution completed
    • SessionUpdateResponseMcpCallFailed - Tool execution failed
  • Added MCP-specific response items:
    • SessionResponseMcpListToolItem - Contains list of available tools from MCP server
    • SessionResponseMcpCallItem - Represents an MCP tool call with optional approval request
    • SessionResponseMcpApprovalRequestItem - Human-in-the-loop approval request
    • SessionResponseMcpApprovalResponseItem - Approval decision response
  • MCP configuration properties on VoiceLiveMcpServerDefinition:
    • ServerLabel - Unique identifier for the MCP server
    • ServerUrl - HTTP endpoint for the MCP server
    • Authorization - Optional authorization header value
    • Headers - Custom HTTP headers for MCP server requests
    • AllowedTools - Optional list of tool names to enable (whitelist)
    • RequireApproval - Approval policy for tool execution

• Enhanced Avatar Configuration: Extended avatar capabilities with new configuration options

  • Added AvatarConfigTypes enum for avatar type selection ("photo_avatar", "video_avatar")
  • Added PhotoAvatarBaseModes enum for photo avatar model selection ("vasa1")
  • Added AvatarOutputProtocol enum for output protocol selection ("webrtc", "websocket")
  • New properties on AvatarConfiguration:
    • Type - Avatar type (photo or video)
    • Model - Base model for photo avatars
    • OutputProtocol - Protocol for avatar data streaming

• Personal Voice Enhancements: Added prosody and localization controls for Azure Personal Voice

  • New properties on AzurePersonalVoice:
    • Locale - Primary locale for speech synthesis
    • PreferLocales - List of preferred fallback locales
    • Rate - Speech rate adjustment (e.g., "+10%", "-15%")
    • Pitch - Pitch adjustment (e.g., "+2st", "-1st")
    • Volume - Volume adjustment (e.g., "+6dB", "-3dB")
    • Style - Speaking style selection
    • CustomLexiconUrl - URL to custom pronunciation lexicon

• Image Input Support: Added multimodal image input capabilities

  • Added RequestImageContentPart for including images in user messages
  • Added RequestImageContentPartDetail enum for image quality control ("auto", "low", "high")
  • Properties on RequestImageContentPart:
    • Url - Image URL (supports http://, https://, and data: URIs)
    • Detail - Image processing detail level

• Enhanced Token Tracking: Added image token usage metrics

  • Added ImageTokens property to InputTokenDetails for tracking image processing costs
  • Added ImageTokens property to CachedTokenDetails for cached image token tracking

• New Voice Options: Added new OpenAI voice presets

  • OAIVoice.Cedar - Additional voice option
  • OAIVoice.Marin - Additional voice option

• Foundry Agent Support: Added support for agent-centric sessions using Azure AI Foundry agents

  • Added AgentSessionConfig class for configuring Foundry agent sessions:
    • AgentName - The name of the Foundry agent to use
    • ProjectName - The name of the Azure AI project which the agent belongs to
    • AgentVersion - Optional version of the agent to use
    • ConversationId - Optional conversation ID to continue
    • AuthenticationIdentityClientId - Optional client ID for user-assigned managed identity
    • FoundryResourceOverride - Optional Foundry resource name for cross-resource agent mode
  • Added SessionTarget class for specifying session targets (model or agent):
    • SessionTarget.FromModel(string model) - Creates a model-centric session target
    • SessionTarget.FromAgent(AgentSessionConfig agentConfig) - Creates an agent-centric session target
    • Implicit conversions from string and AgentSessionConfig

• Interim Response Configuration: Added support for interim responses during latency or tool execution delays

  • Added InterimResponseConfigBase abstract base class with Triggers and LatencyThresholdMs properties
  • Added InterimResponseTrigger extensible enum (Latency, Tool)
  • Added LlmInterimResponseConfig for AI-generated interim responses with Model, Instructions, and MaxCompletionTokens properties
  • Added StaticInterimResponseConfig for predefined static text interim responses with Texts collection

• New VoiceLiveClient Session Methods: Added new overloads for creating and starting sessions

  • CreateSession(string model) - Creates an unconnected session for a model
  • CreateSession(SessionTarget target) - Creates an unconnected session from a model or agent target
  • CreateSession(VoiceLiveSessionOptions sessionConfig) - Creates an unconnected session from session configuration
  • StartSessionAsync(AgentSessionConfig agentConfig, ...) - Starts a connected session with a Foundry agent
  • StartSessionAsync(SessionTarget target, ...) - Starts a connected session from a model or agent target
  • StartSessionAsync(SessionTarget target, VoiceLiveSessionOptions sessionConfig, ...) - Starts a connected session from a target with additional configuration

• New Service Version: Added ServiceVersion.V2026_01_01_PREVIEW to VoiceLiveClientOptions