dev start¶

Scaffold a new AI/ML production project interactively.

dev start

No flags. All configuration is done through interactive prompts.

Prompts¶

1. Project name¶

Project name (or . to scaffold in current folder):

Type a name to create a new subdirectory (e.g. my-agent)
Type . to scaffold directly in the current directory

2. Project structure¶

Choose the folder layout that matches your project type:

Option	Directories created
AI / ML	`app/` `src/inference/` `src/services/` `src/database/` `models/` `tests/` `docs/` `config/` `.github/workflows/`
API	`app/` `src/services/` `src/database/` `tests/` `docs/` `config/`
Minimal	`app/` `tests/` `config/`
None	Empty project — just `pyproject.toml` + `.gitignore`

All structures include:

pyproject.toml (via uv init)
.gitignore (pre-configured for Python + ML artifacts)
.python-version (set to 3.11)
app/main.py
.env + .env.example

3. Optional features¶

Toggle with space, confirm with enter:

Docker

Adds:

Dockerfile
.dockerignore

Claude

Adds:

.claude/
├── CLAUDE.md       # Project context for Claude Code
└── AGENTS.md       # Agent instructions

4. Virtual environment¶

❯ Yes — create .venv now (runs uv venv)
  No  — I'll do it manually

Choosing Yes runs uv venv in the project directory immediately.

What gets created (AI/ML example)¶

my-ai-api/
├── app/
│   └── main.py
├── src/
│   ├── inference/
│   ├── services/
│   └── database/
├── models/
├── tests/
├── docs/
├── config/
│   └── settings.py
├── .github/
│   └── workflows/
├── .claude/            # if Claude selected
│   ├── CLAUDE.md
│   └── AGENTS.md
├── Dockerfile          # if Docker selected
├── .dockerignore       # if Docker selected
├── .env
├── .env.example
├── .gitignore
├── .python-version
└── pyproject.toml

Idempotent¶

Running dev start on an existing directory is safe — it will not overwrite files that already exist.