HandsOnAI
A unified educational toolkit for teaching how modern AI systems actually work — by building and interacting with them directly. Chat, RAG, agents, workflows, loops, and evals: six small, readable modules that make each concept legible.
pip install hands-on-ai
Each module is small, self-contained, and designed so the concept is what students see — not framework plumbing.
Prompt engineering, system prompts, personalities, and memory — the front door to LLMs.
Retrieval-Augmented Generation over your own documents: embeddings, search, grounded answers.
ReAct-style reasoning with tools — multi-step thinking, acting, and observing.
Multi-step tasks as plain folders of stages you can read, review, and rerun.
Do, check, repeat until a goal is met — the core move of agentic systems, made visible.
Score output quality with an LLM judge — the foundation of testing AI systems.
Progressive difficulty, mini-project galleries, and CLI tools for chat, rag, and agent — ready for labs and lectures.
Point it at a local Ollama server, a campus server, or a cloud provider. One config, no code changes.
Once a concept clicks, students graduate to production libraries — LangChain, LlamaIndex, eval harnesses — with real understanding.