Agentic.Workflow.Agents

Microsoft Agent Framework integration for Agentic.Workflow. Provides abstractions for LLM-powered workflow steps with conversation continuity and streaming responses.

Installation

dotnet add package Agentic.Workflow.Agents

Features

Agent Steps

Create workflow steps powered by LLM agents:

public class AnalyzeDocumentStep : IAgentStep<DocumentState> { public string GetSystemPrompt() => """ You are a document analyst. Analyze the provided document and extract key insights. Focus on: main topics, sentiment, key entities, and actionable recommendations. """; public Type? GetOutputSchemaType() => typeof(DocumentAnalysis); public async Task<StepResult<DocumentState>> ExecuteAsync( DocumentState state, StepContext context, CancellationToken ct) { // Agent execution handled by generated worker return StepResult<DocumentState>.FromState(state); } }

Conversation Continuity

Enable per-agent conversation threads for context retention:

public record MyState : IWorkflowState, IConversationalState { public Guid WorkflowId { get; init; } public ImmutableDictionary<string, string> SerializedThreads { get; init; } = ImmutableDictionary<string, string>.Empty; public IConversationalState WithSerializedThread(string agentType, string thread) => this with { SerializedThreads = SerializedThreads.SetItem(agentType, thread) }; }

Streaming Responses

Handle real-time token streaming:

public class WebSocketStreamingCallback : IStreamingCallback { public async Task OnTokenReceivedAsync( string token, Guid workflowId, string stepName, CancellationToken ct) { await _hubContext.Clients.Group(workflowId.ToString()) .SendAsync("TokenReceived", token, ct); } public async Task OnResponseCompletedAsync( string fullResponse, Guid workflowId, string stepName, CancellationToken ct) { await _hubContext.Clients.Group(workflowId.ToString()) .SendAsync("ResponseCompleted", fullResponse, ct); } }

Configuration