Session 1: Foundation + Local Chat

Duration: 75-90 minutes | Level: Intermediate .NET | Series: OpenClawNet — Microsoft Reactor

Goal#

Build a working AI chatbot with Aspire, local LLMs (Ollama, Azure OpenAI, or GitHub Copilot SDK), and HTTP NDJSON streaming — then understand every layer of the architecture that makes it work.

Prerequisites#

Tool	Version	Install
.NET SDK	10.0+	dot.net/download
Ollama or Foundry Local	Latest	ollama.com / Foundry Local
VS Code	Latest	code.visualstudio.com
GitHub Copilot	Active subscription	VS Code extension

Pre-session setup:

# Option A: Ollama
ollama pull llama3.2
# Option B: Foundry Local
foundrylocal model download --assigned phi-4
 
dotnet workload install aspire

What You'll Learn#

Architecture & Core Abstractions — How 27 projects are organized into clean vertical slices, and why IAgentProvider (Microsoft Agent Framework) is the contract that makes everything pluggable
Local LLM Provider + Data Layer — How HTTP NDJSON streaming works with IAsyncEnumerable, and how EF Core entities store conversations
Gateway + SSE + Blazor — How Minimal APIs, HTTP streaming with NDJSON, and Aspire orchestration come together to deliver a complete chat experience

Session Materials#

Resource	Description
speaker-script.md	Minute-by-minute dual-presenter timeline
demo-scripts.md	Three step-by-step live demos with recovery plans
bonus-demos.md	Additional demos for extra time or Q&A
guide.md	Detailed walkthrough guide (English)
guide-es.md	Detailed walkthrough guide (Spanish)
demo-agents/	Sample agent persona files for Demo 3 (pirate, chef, robot)
code/	Demo-ready code: console app (demo1), bug injection scripts (demo2), agent personalities (demo3)

Projects Covered#

Project	LOC	Description
`OpenClawNet.Models.Abstractions`	93	`IAgentProvider` interface, `ChatRequest`, `ChatResponse`, `ChatMessage` records
`OpenClawNet.Models.Ollama`	181	Ollama provider (`OllamaAgentProvider`) with streaming via OllamaSharp
`OpenClawNet.Models.AzureOpenAI`	185	Azure OpenAI provider via Azure.AI.OpenAI SDK (3 auth modes: API Key, Integrated, Federated)
`OpenClawNet.Models.FoundryLocal`	165	Foundry Local bridge provider via Microsoft.AI.Foundry.Local
`OpenClawNet.Models.GitHubCopilot`	142	GitHub Copilot SDK provider via `GitHub.Copilot.SDK` (v0.2.2) with environment-based auth
`OpenClawNet.Storage`	275	EF Core DbContext, entities (ChatSession, ChatMessageEntity, AgentProfile, etc.)
`OpenClawNet.ServiceDefaults`	105	Aspire service defaults — telemetry, health checks, OpenAPI
`OpenClawNet.AppHost`	18	Aspire orchestration — wires Gateway + Web with service discovery
`OpenClawNet.Gateway`	625	Minimal APIs, HTTP NDJSON streaming endpoint (`POST /api/chat/stream`), Model Providers, Agent Profiles
`OpenClawNet.Web`	28	Blazor web app — chat UI with real-time HTTP NDJSON streaming + inline agent selector badge (provider/model)

Quick Start#

# Clone and build
git clone https://github.com/elbruno/openclawnet.git
cd openclawnet
dotnet build
 
# Verify local LLM is running
ollama list   # Should show llama3.2 (if using Ollama)
 
# Launch the full stack (from repo root)
aspire run

Expected endpoints:

🌐 Web UI: http://localhost:5001
🔌 Gateway API: http://localhost:5000
📊 Aspire Dashboard: https://localhost:15100

Architecture at a Glance#

┌──────────────────────────────────────────────────────┐
│                    Blazor Web UI                      │
│                 (OpenClawNet.Web)                     │
├──────────────────────────────────────────────────────┤
│            HTTP NDJSON + REST API                          │
│             (OpenClawNet.Gateway)                     │
├──────────────────────────────────────────────────────┤
│              RuntimeAgentProvider                     │
│          (routes to active provider)                 │
├─────────┬──────────┬─────────┬──────────┬────────────┤
│ Ollama  │Azure     │Microsoft│Foundry   │GitHub      │
│         │OpenAI    │Foundry  │Local     │Copilot SDK │
├─────────┴──────────┴─────────┴──────────┴────────────┤
│      Storage (EF Core)    │    ServiceDefaults        │
│       + AgentProfile      │       (Aspire)            │
└───────────────────────────┴──────────────────────────┘

Next Session#

Session 2: Tools & Agent Workflows — Give the chatbot superpowers with file system access, web fetching, shell execution, and the agent tool-call loop.