🚀 Session 4: Automation + Cloud

Overview#

Your agent works locally. Users love it. Now three production challenges stand between you and a real platform: cloud providers (GPT-4o, SLAs, team-wide access), automated scheduling (background jobs without a user present), and testing (24 tests proving the architecture works).

This is the series finale. By the end, every piece connects: chat, tools, skills, memory, scheduling, health checks, and cloud providers — a complete AI agent platform built with .NET.

Prerequisites#

✅ Session 3 complete and working
✅ .NET 10 SDK, VS Code or Visual Studio
✅ Local LLM running (Ollama with llama3.2 or Foundry Local)
✅ Understanding of: background services, HTTP clients, unit testing
☁️ Optional: Azure account with Azure OpenAI or Foundry access

What You'll Learn#

☁️ Stage 1: Cloud Providers (12 min)#

IModelClient polymorphism — one interface, three implementations (Local LLM, Azure OpenAI, Foundry)
AzureOpenAIModelClient — Azure.AI.OpenAI SDK integration
FoundryModelClient — custom HTTP client with manual JSON serialization
Options pattern for provider configuration
Switching providers with a single DI line change

⏰ Stage 2: Scheduling + Health (12 min)#

BackgroundService pattern for JobSchedulerService
Cron-based job scheduling with EF Core persistence and audit trail
SchedulerTool — the agent schedules itself
ServiceDefaults — health checks (/health, /alive) + OpenTelemetry
Aspire Dashboard integration

✅ Stage 3: Testing + Production (12 min)#

Test pyramid: 23 unit tests + 1 integration test
Mocking patterns: IModelClient, IDbContextFactory, ITool, ISkillLoader
Test suites: PromptComposerTests, ToolExecutorTests, SkillParserTests, ConversationStoreTests, ToolRegistryTests
Production readiness checklist

🎉 Series Finale Closing (14 min)#

Full platform demo: chat → tools → skills → scheduling → health → Aspire
Four-session series recap with architecture diagram
Where to go from here

Projects Covered#

Project	Key Responsibility
OpenClawNet.Models.AzureOpenAI	Azure OpenAI client (137 LOC)
OpenClawNet.Models.Foundry	Foundry client (195 LOC)
OpenClawNet.Models.Abstractions	IModelClient interface
OpenClawNet.Tools.Scheduler	SchedulerTool with cron support
OpenClawNet.Gateway	JobSchedulerService (BackgroundService)
OpenClawNet.ServiceDefaults	Health checks + OpenTelemetry (105 LOC)
OpenClawNet.UnitTests	23 unit tests
OpenClawNet.IntegrationTests	1 integration test

Session Materials#

Resource	Link
📖 Presenter Guide	session-4-guide.md
📖 Guía (Español)	session-4-guide-es.md
🎤 Speaker Script	speaker-script.md
🤖 Copilot Prompts	copilot-prompts.md
🖥️ Slides	slides.md

Git Checkpoints#

Starting tag: session-4-start (alias: session-3-complete)
Ending tag: session-4-complete

The Full Architecture#

┌─────────────────────────────────────────────────────────┐
│                    OpenClawNet Platform                  │
├──────────┬──────────┬──────────┬───────────────────────────┤
│  Web UI  │ REST API │  Aspire  │     Health Checks       │
├──────────┴──────────┴──────────┴───────────────────────────┤
│                   Agent Orchestrator                      │
│         ┌──────────────────────────────┐                  │
│         │     Prompt Composer          │                  │
│         │  (System + Skills + Memory)  │                  │
│         └──────────────────────────────┘                  │
├──────────┬──────────┬──────────┬───────────────────────────┤
│   Tools  │  Skills  │  Memory  │     Scheduler           │
│ Registry │  Loader  │  Store   │  BackgroundService      │
├──────────┴──────────┴──────────┴───────────────────────────┤
│                   Model Abstraction                       │
│         ┌────────┬──────────┬──────────┐                  │
│         │ Ollama │ Azure AI │ Foundry  │                  │
│         └────────┴──────────┴──────────┘                  │
├────────────────────────────────────────────────────────────┤
│              Storage (EF Core + SQLite)                   │
└────────────────────────────────────────────────────────────┘

Series Recap#

Session	Topic	What We Built
1	Scaffolding + Local Chat	Aspire host, local LLM integration, Gateway, Chat UI
2	Tools + Agent Workflows	ITool, Registry, Executor, approval policies, tool loop
3	Skills + Memory	Markdown skills, YAML parsing, summarization, semantic search
4	Automation + Cloud	Cloud providers, job scheduling, health checks, 24 tests