Azure.AI.AgentServer.Responses_1.0.0-beta.1

1.0.0-beta.1 (2026-04-14)

Features Added

• Initial release of the Azure.AI.AgentServer.Responses library.
• ASP.NET Core server library implementing the Azure AI Responses API.
• ResponseHandler abstract class for custom response handling with CreateAsync returning IAsyncEnumerable<ResponseStreamEvent>.
• TextResponse convenience class for simple text-only responses with a single delegate.
• ResponseEventStream builder for full control over SSE event generation with automatic sequenceNumber, outputIndex, contentIndex, and itemId tracking.
• ResponseEventStream convenience generators for emitting complete output items in one call:
  • OutputItemMessage(string text) — emits a full text message output item (handles all inner SSE events automatically)
  • OutputItemMessage(string text, IEnumerable<Annotation> annotations) — emits a text message with file annotations
  • OutputItemMessage(IAsyncEnumerable<string> tokens, CancellationToken) — streams tokens as response.output_text.delta events
  • OutputItemFunctionCall(name, callId, arguments) — emits a complete function call output item
  • OutputItemFunctionCallOutput(callId, output) — emits a function call output (no deltas)
  • OutputItemReasoningItem(...) — emits a reasoning output item
  • OutputItemImageGenCall(resultBase64) — emits an image generation result with status transitions
  • OutputItemStructuredOutputs(output) — emits an arbitrary structured JSON output item
  • One-liner convenience generators for all remaining simple output item types: computer calls, local shell calls, function shell calls, apply-patch calls, custom tool call outputs, MCP approval requests/responses, and compaction
• ResponseContext with GetInputItemsAsync() (returns IReadOnlyList<Item>) and GetInputTextAsync() convenience for accessing resolved input items and text content.
• IEnumerable<Item>.GetInputText() extension method for extracting text from any sequence of input items.
• CreateResponse.GetInputExpanded() extension for advanced access to expanded input items as OutputItem types.
• Built-in in-memory response provider and execution tracking.
• Support for default, streaming, background, and streaming+background response modes.
• AgentHostBuilder convenience methods for zero-config server startup via ResponsesServer.Run<T>().
• Protocol identity registration with ServerUserAgentRegistry during route mapping.
• x-agent-response-id header validation matching the Responses API specification.
• Conversation ID round-trip support in both synchronous and SSE streaming modes.
• OpenTelemetry distributed tracing via Azure.AI.AgentServer.Responses activity source.