Tutorials & Projects
These beginner-friendly tutorials walk you through building small research tools and practicing core development workflows with AI coding assistance and the Research Computing OpenAI-compatible API. Each project stays small on purpose: the goal is to practice a clear workflow, review generated code, and make focused improvements.
Before you start
One guide sets you up for every tutorial on this page: an API key, a coding assistant that can read and edit your projects, and a working sense of how to get good results from the models we host. No prior experience assumed.
That guide covers, in order:
- Getting your API key and a model ID from the Voyager portal, and how to read the model badges so you pick one that can actually edit files.
- Choosing and setting up a tool — OpenCode (recommended), OpenWork, VS Code, or Open WebUI for brainstorming.
- Generating your config in Voyager — presets, agents, and permissions, without hand-editing JSON.
- Why prompting and agents matter more here — plus
AGENTS.md, custom agents, and the RC skills library for HPC and research work.
Each tutorial's Prerequisites section lists anything else it needs (such as Python or Node.js), with beginner install steps built in — so you can jump straight to a project.
Build a Sudoku Solver with a Visual UI installs nothing at all — it is one HTML file you open by double-clicking. A good way to get used to working with an assistant before taking on a project with real data.
Project tutorials
Each project uses the RC OpenAI-compatible API and an AI coding assistant to build something small but real. They make great starting points for a hackathon — pick one, then take it somewhere interesting.
Every tutorial is labeled Beginner or Intermediate. New to coding? Start with the sudoku solver — it needs nothing installed — then move to the Beginner Streamlit apps, which are deliberately tiny and buildable in an afternoon. Work up to an Intermediate project from there.
AI-powered projects
| Tutorial | Level | What you build | Stack | Time |
|---|---|---|---|---|
| Build a Sudoku Solver with a Visual UI | Beginner | A browser sudoku board that generates puzzles and solves them box by box. Nothing to install. | HTML, CSS, JavaScript | 30–40 min |
| Create a Basic Lab Web Portal | Beginner | A small internal homepage for lab updates, projects, and links. | Next.js, TypeScript, Tailwind | 35–45 min |
| Build a Lab Onboarding Assistant CLI | Beginner | A Python CLI that answers questions from local lab notes using the LLM API. | Python, OpenAI SDK | 35–45 min |
| Create a Slurm Queue Snapshot CLI | Beginner | A Python CLI that parses squeue output into a readable summary. | Python, Slurm | 30–40 min |
| Create a CSV Dataset Explorer | Beginner | A Streamlit app that previews, filters, and summarizes CSV files. | Python, Streamlit, pandas | 25–35 min |
| Build a Streamlit AI Chatbot | Beginner | A streaming web chat UI backed by the RC LLM API. | Python, Streamlit, OpenAI SDK | 30–40 min |
| Summarize Research Papers from the Command Line | Intermediate | A CLI that turns a PDF into a structured, section-by-section summary. | Python, OpenAI SDK, pypdf | 35–45 min |
| Generate Slurm Job Scripts with AI | Intermediate | A CLI that drafts sbatch scripts from plain English, grounded in your cluster's rules. | Python, OpenAI SDK, Slurm | 35–45 min |
| Explain and Document Code with AI | Intermediate | A CLI that explains a Python file and suggests docstrings. | Python, OpenAI SDK | 30–40 min |
| Chat with Your CSV | Intermediate | A Streamlit app that answers plain-English questions about a dataset. | Python, Streamlit, pandas, OpenAI SDK | 40–50 min |
| Build a Q&A Assistant over Your Notes | Intermediate | A retrieval (RAG) CLI that answers from a folder of notes with citations. | Python, OpenAI SDK, scikit-learn | 45–55 min |
Foundational skills
| Tutorial | Level | What you build | Stack | Time |
|---|---|---|---|---|
| Track a New Project with Git | Beginner | A version-controlled project with clean commits, .gitignore, and a remote. | Git | 25–35 min |
| Collaborate on a Project with Git | Intermediate | A team workflow with branches, pull requests, and conflict resolution. | Git | 40–50 min |
How these tutorials work
Each tutorial follows the same beginner-safe pattern:
- Set up a small workspace with a known project structure.
- Write a clear first prompt that states the goal, constraints, and definition of done.
- Review generated code like a draft from a teammate before accepting it.
- Ask for one focused adjustment at a time to keep the project from sprawling.
That pattern is not arbitrary — it is what makes locally hosted models reliable. The reasoning behind it, and the prompt skeleton these tutorials use, is in Getting Started with AI.