# Tutorials

End-to-end walkthroughs. Each tutorial takes a real scenario and builds a working agent from scratch.

## Tutorials

### Complete, runnable walkthroughs

- [Reusable Capabilities](capability-pattern.md) — Compose Summarizer, Translator, and SentimentClassifier onto an editorial role with zero method bodies.
- [REST API actions with @WebService](web-service-action.md) — Bind any REST endpoint to an LLM-callable action through annotations alone. GET + query, path templating, POST + body + custom headers, retry.
- [Declarative Guardrails](declarative-guardrails.md) — Validate, sanitise, and bound action parameters and return values through `@InputGuardrail` / `@OutputGuardrail` annotations, no per-action plumbing.
- [Declarative RAG with @KnowledgeSource and @Retrieval](declarative-rag.md) — Wire knowledge sources to a Role and retrieve from them on every action call through annotations alone, no manual vector-store plumbing.
- [Declarative Resilience](declarative-resilience.md) — Wrap action dispatch with structured tracing, latency + counter metrics, and automatic recovery via `@Traced` / `@Metered` / `@Fallback`.
- [Research Agent](research-agent.md) — Agent that searches the web, reads PDFs, and produces cited summaries.

### Architecture sketches + building blocks

Each of these pages describes the shape of a non-trivial system and links to the framework pieces you'd compose to build it. The full code walkthrough is still being written — start the implementation today and open a discussion if anything's missing.

- [Customer Support Bot](customer-support-bot.md) — RAG over a knowledge base, with sentiment-aware escalation handoff.
- [Multi-Agent Research Team](multi-agent-research-team.md) — Coordinator + specialists + judge.
- [Agent with Tool Approval](agent-with-tool-approval.md) — Destructive tools gated through a human approval UI.