Template System

How It Works

Templates are layered via Copier + Jinja2:

1. Base template — shared by all projects (pyproject.toml, CI, gitignore)

2. Archetype template — project shape (service adds infra/, Makefile, .env)

3. Fragment templates — per-group scaffolding (API routes, DB models, etc.)

Group-Aware Rendering

All templates receive groups (list of selected group IDs) in Jinja2 context:

{% if "database" in groups %}
from {{ project_slug }}.core.database import get_session
{% endif %}

This means fragments auto-wire based on what’s selected:

• .env.example only includes vars for your groups

• Settings only has fields for selected groups

• Lifespan connects DB + Redis if those groups are present

• Health check pings actual dependencies

• Test fixtures match selected groups

Source Code Fragments

Fragment	What it generates
settings_module	Pydantic Settings + .env.example
api_app	FastAPI factory, routes, middleware, deps, main.py
db_models	SQLAlchemy engine, session, Base model, mixins
redis_client	Async Redis client + connection pool
auth_jwt	JWT create/decode + Bearer auth dependency
agent_app	LangGraph agent with graph, state, tools, prompts