scaffold-fastloom-service

Scaffold a new fastloom service

Generate the three files every fastloom service needs at its project root, plus the pyproject.toml extras hint.

Confirm with the user first

Ask in one round:

1. Service name (used as the Python package and PROJECT_NAME).
2. Capabilities needed — multi-select from: mongo, rabbit, kafka, redis, mcp, celery, httpx, openai. fastapi is always included.
3. IAM mode — OIDC (OIDC_URL) or OAuth2 (authorizationUrl + tokenUrl) or none for local dev.

Generate

Use the user's answers to produce these files. Write to cwd unless the user specifies a subdirectory.

pyproject.toml (snippet to add or merge)

[project]
name = "<service-name>"
requires-python = ">=3.12,<3.14"
dependencies = [
    "fastloom[fastapi,<other-selected-extras>]>=0.4,<0.5",
]

settings.py

Compose Settings from BaseGeneralSettings + LauncherSettings plus exactly the capability mixins the user picked. Always alias Configs as TC. Add a TenantSettings(BaseModel) placeholder with a name: str field. Reference: docs/settings.md in the fastloom repo.

tenants.yaml

A default: block with the SCREAMING_SNAKE_CASE infrastructure fields (MONGO_URI, RABBIT_URI, REDIS_URL — only the ones matching selected extras), ENVIRONMENT: development, PROJECT_NAME, APP_PORT: 8000, DEBUG: true, observability toggles off. Add one example tenant block under default: with just name: and an optional website_url:. Do not put MONGO_DATABASE or any infrastructure URI in the per-tenant overrides — databases are shared across tenants; partition via the tenant field on documents.

app.py

A minimal App(...) declaration: one demo router with a /ping route, models_module=models and signals_module=signals only if the user picked mongo / rabbit respectively. Include cache_healthcheck=True if redis is selected.

Optional starter package

If the user said they want a fully working starter, also create:

• <service-name>/__init__.py
• <service-name>/api/__init__.py + a ping.py router with GET /ping returning {"pong": "ok"}
• <service-name>/models/__init__.py if mongo was picked
• <service-name>/signals/__init__.py if rabbit was picked

Rules to follow

• Pydantic-everywhere, no Any, no untyped dicts (see fastloom's .claude/rules/typing.md if available).
• Settings field names are SCREAMING_SNAKE_CASE for capability fields; the per-tenant business-domain fields are snake_case.
• No MONGO_DATABASE in tenant overrides.
• Use Str[T] for URL/DSN fields in Settings.
• Use Field(default_factory=...) for runtime defaults.

Done

Tell the user what to do next:

poetry install
launch
# Then visit http://localhost:8000/api/<service-name>/docs