Examples & Learning Paths

Examples & Learning Paths

Learn to build agentic workflows through structured tutorials that teach you to think about workflow design, not just copy code.

---

How to Use This Documentation

Each example in this section teaches concepts before showing code:

Section	What You’ll Find
Problem Narrative	Real-world scenario motivating the pattern
Learning Objectives	What you’ll understand after reading
Conceptual Foundation	Design decisions, trade-offs, anti-patterns
Progressive Code Reveal	Shape first, then implementation details
”Aha Moment”	Core insight crystallized
Extension Exercises	Guided practice to deepen understanding

---

Learning Paths

Choose a path based on your goals and available time.

Path 1: First Workflow (30 minutes)

Goal: Understand the foundational patterns that all workflows build on.

Step	Read	Learn
1	Basic Workflow	Sequential steps, immutable state, saga pattern
2	Run the ContentPipeline sample	See a workflow in action
3	Exercise: Add a ReserveInventory step to the order workflow

After this path: You can build simple sequential workflows with proper state management.

---

Path 2: AI Agent Patterns (2 hours)

Goal: Learn patterns for building intelligent, adaptive AI systems.

Step	Read	Learn
1	Thompson Sampling	Multi-armed bandit, exploration vs exploitation
2	Run the MultiModelRouter sample	See adaptive model selection
3	Iterative Refinement	Quality loops, [Append] attribute, maxIterations
4	Run the AgenticCoder sample	See test-driven refinement
5	Exercise: Add a custom task category to MultiModelRouter

After this path: You can build AI systems that learn and improve over time.

---

Path 3: Production Patterns (3 hours)

Goal: Master patterns for production-ready workflows with human oversight.

Step	Read	Learn
1	Approval Flow	Human checkpoints, timeout handling, escalation
2	Branching	Conditional routing, transition tables
3	Fork/Join	Parallel execution, state merging
4	Run the ContentPipeline sample	See approval gates and compensation
5	Exercise: Add multi-approver workflow to ContentPipeline

After this path: You can build workflows with human-AI collaboration, parallel processing, and complex routing.

---

Path 4: Complete Mastery (Full Day)

Goal: Understand all patterns and how they compose together.

1. Complete Path 1 (Foundation)
2. Complete Path 2 (AI Patterns)
3. Complete Path 3 (Production Patterns)
4. Capstone Project: Design a workflow that combines:
   • Thompson Sampling for agent selection
   • Iterative refinement for quality
   • Human approval before final action
   • Fork/Join for parallel analysis

---

Pattern Quick Reference

Pattern	When to Use	Key Concept
Basic Workflow	Sequential operations with dependencies	Saga pattern, immutable state
Branching	Different logic for different inputs	Declarative routing, transition tables
Fork/Join	Independent operations that can parallelize	State merging, fail-fast vs continue
Iterative Refinement	Quality improvement through feedback	[Append], maxIterations circuit breaker
Approval Flow	Human decisions in automated workflows	Timeout handling, escalation, audit
Thompson Sampling	Adaptive selection from multiple options	Beta distributions, exploration/exploitation

---

Sample Applications

Runnable projects demonstrating complete implementations.

Sample	Run Command	What It Demonstrates
ContentPipeline	dotnet run --project samples/ContentPipeline	Human approval gates, compensation, audit trails
MultiModelRouter	dotnet run --project samples/MultiModelRouter	Thompson Sampling, intelligent model selection
AgenticCoder	dotnet run --project samples/AgenticCoder	Iterative refinement loops, human checkpoints

Each sample README includes:

• Problem narrative and solution approach
• Conceptual explanation of the patterns used
• Step-by-step walkthrough of the implementation
• Extension exercises for practice

---

Prerequisites

Before running examples, ensure you have:

1. .NET 9.0 or later installed
2. PostgreSQL running (for Marten event store)
3. Strategos packages installed:

dotnet add package LevelUp.Strategos
dotnet add package LevelUp.Strategos.Generators

---

Quick Start Template

Program.cs

var builder = WebApplication.CreateBuilder(args);

// Add Wolverine for message handling
builder.Host.UseWolverine(opts =>
{
    opts.Durability.Mode = DurabilityMode.Solo;
});

// Add Marten for persistence
builder.Services.AddMarten(opts =>
{
    opts.Connection(builder.Configuration.GetConnectionString("Marten")!);
})
.IntegrateWithWolverine();

// Add workflow services
builder.Services.AddStrategos()
    .AddWorkflow<YourWorkflow>();

// Register step dependencies
builder.Services.AddScoped<IYourService, YourServiceImpl>();

var app = builder.Build();
app.MapControllers();
app.Run();

---

Key Principles Across All Patterns

Principle	Why It Matters
State is immutable	Enables replay, debugging, concurrency safety
Steps are resolved via DI	Testability, loose coupling
Failures are explicit	StepResult.Fail() not exceptions
Workflows survive restarts	Durability via Wolverine saga persistence
Everything is audited	Events capture all state transitions

---

What’s Next?

After completing these examples:

1. Read the Learn section for deeper conceptual understanding
2. Explore the API Reference for complete interface documentation
3. Check the samples directory for production-quality implementations
4. Join the community to share your workflows and learn from others