AgentBoard

A deliberately small full-stack project for testing AI coding agents.

AgentBoard is a local task board with a FastAPI backend, SQLite persistence, a tiny browser UI, automated tests, and CI. The initial version is intentionally simple so an AI agent can safely extend it through small, verifiable tasks.

Why this project is useful for AI development tests

It exercises several common software-engineering skills at once:

• understand an unfamiliar codebase
• modify API behavior
• change a database schema
• preserve backward compatibility
• update a frontend
• write tests before/after changes
• satisfy CI
• document decisions

Current features

• Create tasks
• List tasks
• Filter by status
• Update task title, description, priority, and status
• Delete tasks
• SQLite storage
• Simple single-page browser UI
• API tests with pytest

Stack

• Python 3.12+
• FastAPI
• SQLite (sqlite3, no ORM)
• Vanilla HTML/CSS/JavaScript
• pytest

Run locally

git clone https://github.com/thp32tt/ai-dev-lab.git
cd ai-dev-lab
python -m venv .venv
# Windows: .venv\\Scripts\\activate
# macOS/Linux: source .venv/bin/activate
pip install -e '.[dev]'
uvicorn app.main:app --reload

Open http://127.0.0.1:8000.

Run tests

pytest

AI-agent test workflow

1. Give the coding agent one item from BACKLOG.md.
2. Tell it to inspect AGENTS.md first.
3. Ask it to implement the change, add tests, and explain its choices.
4. Review the diff and CI result instead of giving it implementation hints.
5. Repeat with a harder backlog item.

A good first task is B1: add due dates and an overdue filter.