Examples Playbook

Real-world examples covering every feature of confused-ai. Pick your level.

---

Full framework map (start here)

For a single narrative that walks every capability area—NorthPeak StoreOps Copilot, import cheat-sheets, architecture diagram, and links back to these tutorials—see 17 · Full framework showcase.

Runnable counterparts in the repo:

• bun run example:showcase — LLM, sessions, workflows, pipeline, health, metrics, optional HTTP.
• bun run example:potential — chunking, circuit breaker, rate limiter, artifacts, profiles, eval metrics, config (no LLM calls).
• bun run example:meridian — full role intelligence platform: 6 personas, handoff, router, supervisor, consensus, RAG, HTTP server.
• bun run example:reasoning — chain-of-thought incident triage with ReasoningManager.
• bun run example:scheduled — nightly market digest cron with ScheduleManager.
• bun run example:code-review — three-agent PR review pipeline with bare(), compose(), pipe().
• bun run example:eval — CI eval regression guard with runEvalSuite + InMemoryEvalStore.

---

Skill Levels

Label	Who it's for
🟢 Beginner	Never used an AI framework. First time here.
🟡 Intermediate	Built basic agents, want real-world patterns.
🔴 Advanced	Production systems, orchestration, resilience.

---

All Examples

#	Example	Level	What you learn
01	Hello World	🟢	Create and run your first agent
02	First Custom Tool	🟢	Build a tool that calls a real API
03	Tool with Approval	🟢	Ask human before executing risky actions
04	Extend & Wrap Tools	🟡	Add middleware, caching, auth to any tool
05	RAG Knowledge Base	🟡	Answer questions from your own documents
06	Persistent Memory	🟡	Remember users and conversations
07	Storage Patterns	🟡	Cache, state, and config in agents
08	Multi-Agent Team	🔴	Specialized agents working together
09	Supervisor Workflow	🔴	One agent delegates to many
10	Database Analyst	🔴	Query a SQL database with natural language
11	Customer Support Bot	🔴	Full bot: sessions + guardrails + handoff
12	Observability & Hooks	🟡	Log, trace, and measure every agent step
13	Production Resilience	🔴	Circuit breakers, retries, fallbacks
14	MCP Filesystem Agent	🔴	Use Model Context Protocol tools
15	Full-Stack App	🔴	Complete app: HTTP API + agent + RAG + storage
16	Intelligent LLM Router	🟡	Auto-route requests to the right model by task, cost & speed
17	Full framework showcase	🔴	Coverage map: one real-world story + every module / import path + links to 01–16
18	Meridian — Role Intelligence Platform	🔴	Full platform demo: 6 role personas, triage handoff, router, supervisor, consensus, compose/pipe, workflows, RAG, guardrails, resilience, health, HTTP runtime
19	Incident Triage Bot	🔴	Chain-of-thought reasoning, ReasoningManager, event streaming, NextAction loop
20	Scheduled Agent Jobs	🔴	Cron scheduling, ScheduleManager, handler registry, run history, enable/disable
21	Code Review Pipeline	🔴	bare(), compose(), pipe(), conditional when hand-off, lifecycle hooks
22	Eval Regression Guard	🟡	runEvalSuite, EvalStore, custom scorer, baseline saving, CI exit code

---

Quick Start (30 seconds)

npm install confused-ai

import { createAgent } from 'confused-ai';

const agent = createAgent({
  name: 'my-first-agent',
  model: 'gpt-4o-mini',
  instructions: 'You are a helpful assistant.',
});

const result = await agent.run('What is 2 + 2?');
console.log(result.text); // "4"

---

Environment Variables

Most examples need these in a .env file:

OPENAI_API_KEY=sk-...        # required for OpenAI models
ANTHROPIC_API_KEY=sk-...     # optional, for Claude models
GOOGLE_AI_API_KEY=...        # optional, for Gemini models