A comprehensive, hands-on AI programming tutorial for IT teams — covering LLM fundamentals, prompt engineering, AI agents, RAG, and production-grade AI application development.
An open-source AI programming knowledge base designed for IT professionals who want to learn AI development from scratch. Built with VitePress + Jupyter Notebooks, it covers everything from your first API call to deploying production AI agents.
Key Features:
- Learn by Doing — Every concept comes with runnable code (Jupyter Notebooks on Google Colab)
- Problem-Driven — Each section starts with a real-world problem you'll face at work
- Multiple Learning Paths — Quick-start (3 hours), role-based, or full curriculum
- Updated for 2026 — Covers GPT-5, Claude 4.6, DeepSeek R1, OpenClaw, MCP protocol, and more
| Part | Chapters | Topics |
|---|---|---|
| Getting Started | Ch1–Ch4 | First AI conversation, AI landscape & model providers, LLM fundamentals (tokens, parameters, reasoning models), Prompt engineering (CoT, structured output) |
| Tools | Ch5–Ch6 | AI coding tools (Cursor, Copilot, Claude Code, Windsurf, Codex CLI), Context engineering (AGENTS.md, Rules, Vibe Coding) |
| Capabilities | Ch7–Ch10 | Function calling & tool use, Multimodal AI (vision, image generation, speech, Realtime API), AI agents (ReAct, OpenAI Agents SDK, LangGraph), Multi-agent systems (handoff, supervisor-worker) |
| Ecosystem | Ch11–Ch12 | AI protocols (MCP, A2A, ANP), RAG & memory (vector search, hybrid retrieval, reranking) |
| Production | Ch13–Ch14 | Guardrails, evaluation, observability, cost optimization, IT practice projects (knowledge base Q&A, code review bot, ops assistant) |
| Extensions | Ch15–Ch16 | AI workflow platforms (Dify, Coze, n8n), AI for non-developers |
Scale: 16 chapters, 60+ sections, 55 Markdown docs, 39 Jupyter Notebooks
| Layer | Choice | Why |
|---|---|---|
| Documentation | VitePress v1.6 | Markdown-driven, fast, great i18n support |
| Code Demos | Python 3.10+ / Jupyter | Standard AI/ML ecosystem |
| Diagrams | Mermaid | Architecture & flow charts inline in Markdown |
| Package Managers | pnpm (site) + uv (Python) | Modern & efficient |
git clone https://github.com/forhow134/ai-coding-guide.git
cd ai-coding-guide
# Install dependencies
pnpm install
# Start dev server
pnpm docs:dev
# Visit http://localhost:5173Or use the management script:
./start.sh dev # Start dev server (hot reload)
./start.sh build # Build static site
./start.sh preview # Preview build output
./start.sh stop # Stop serverOption 1: Google Colab (Recommended)
Click the "Open in Colab" badge in any chapter — runs in browser, no local setup needed.
Option 2: Local Jupyter
pip install openai anthropic google-genai tiktoken jupyter
jupyter notebook demos/./start.sh docker-build # Build image (auto-detects podman/docker)
./start.sh docker-run # Run container at localhost:8080
./start.sh docker-stop # Stop container./start.sh docker-build
./start.sh k8s-deploy # Auto-generates manifests
K8S_NAMESPACE=prod ./start.sh k8s-deploy # Target specific namespace1.1 First AI Chat → 4.1 Prompt Basics → 5.1 IDE Tools
→ 7.1 Function Calling → 9.2 ReAct Agent → 12.1 RAG Basics
→ 14.1 Knowledge Base Project
| Role | Recommended Path |
|---|---|
| Backend Developer | Ch1 → Ch3 → Ch4 → Ch7 → Ch9 → Ch12 → Ch13 → Ch14.1 |
| Frontend Developer | Ch1 → Ch4 → Ch5 → Ch6 → Ch8 → Ch14.4 |
| DevOps / SRE | Ch1 → Ch4 → Ch5 → Ch9 → Ch11 → Ch13 → Ch14.3 |
| Tech Lead / Manager | Ch1 → Ch5 → Ch6 → Ch11 → Ch13 → Ch14.4 |
All 16 chapters in order — from beginner to production-ready.
| Area | Details |
|---|---|
| LLM Fundamentals | Tokens, parameters, temperature, reasoning models (o3, DeepSeek R1), model selection |
| Model Providers | OpenAI, Anthropic Claude, Google Gemini, DeepSeek, Qwen, Mistral, xAI Grok |
| API Gateways | OpenRouter, ZenMux, Azure OpenAI, AWS Bedrock |
| Local Deployment | Ollama, LM Studio, open-source models |
| Prompt Engineering | Zero-shot, few-shot, chain-of-thought (CoT), structured output, advanced techniques |
| AI Coding Tools | Cursor, GitHub Copilot, Windsurf, Claude Code, OpenAI Codex CLI, Gemini CLI |
| Context Engineering | AGENTS.md, project rules, cursor rules, vibe coding |
| Function Calling | OpenAI function calling, Anthropic tool use, streaming, tool orchestration |
| Multimodal AI | GPT-4 Vision, DALL-E 3, Stable Diffusion, text-to-speech, Realtime API |
| AI Agents | ReAct pattern, OpenAI Agents SDK, LangGraph, computer use |
| Multi-Agent | Sequential, parallel, hierarchical architectures, Swarm, handoff patterns |
| AI Protocols | MCP (Model Context Protocol), A2A (Agent-to-Agent), ANP (Agent Network Protocol), AG-UI |
| RAG | Vector databases (ChromaDB), embedding, chunking, reranking, hybrid search, agentic RAG |
| Memory | Short-term / long-term memory, conversation history, persistence |
| Production | Guardrails, LLM evaluation, observability, prompt caching, cost optimization |
| Practice Projects | Knowledge base Q&A, AI code review, ops assistant, team toolchain |
| AI Workflow | Dify, Coze, n8n, Flowise — no-code/low-code AI app builders |
| AI History | From Turing Test (1950) to Transformer (2017) to ChatGPT (2022) to GPT-5 (2025) |
- Create a new directory and Markdown files under
docs/ - Follow the four-part structure: Why → What → How → Reflection
- Add corresponding Jupyter Notebooks under
demos/ - Update sidebar config in
docs/.vitepress/config/ - Mark API cost estimates and last-updated dates
Note: AI evolves fast. This tutorial reflects the state of AI technology as of February 2026. Each section includes a last-updated timestamp, and core demos are pinned to specific package versions.