_{by SQLStack}

Fensu checks repository architecture: import boundaries, module roles, and function shape for Python, Rust, TypeScript, JavaScript, and Svelte.

• Boundaries: stop one domain from importing another domain’s private helpers.
• Roles: keep entry points, helpers, models, and types in their owning modules.
• Function shape: keep orchestrators small, dataflow explicit, and signatures clear.

Get started

Requires Python 3.12+. Install in your project’s Python environment:

pip install fensu

From an existing Python repository’s pyproject.toml directory:

fensu init --yes
fensu check

fensu init --yes detects your project, creates fensu.toml, runs an initial check, and installs agent guidance. Existing configuration is preserved.

Run fensu check locally and in CI to enforce your policy. For established repositories, adopt rules incrementally using explicit boundaries and exceptions.

See the quickstart for empty repositories, existing configuration, and language-specific setup.

A finding and its fix

An order entry point importing inventory’s private helper crosses a domain boundary:

# src/shop/orders/main/submit.py
from shop.inventory._helpers.reserve import reserve

Excerpt from a real fensu check --no-color run with FFL102 selected:

FFL102  import 'shop.inventory._helpers.reserve' reaches into internal structure of 'shop.inventory'
 --> src/shop/orders/main/submit.py:1:0
  = help: Import from classes, models, types, constants, exceptions, or a thin main/ entry.

Publish the operation through inventory’s thin main/ entry, then import that public surface instead:

from shop.inventory.main.reserve import reserve

That change clears this boundary finding; it does not disable the rule. Inspect a diagnostic’s full contract and settings with fensu rule FFL102.

The default layout

A Python domain owns its behavior and vocabulary:

src/shop/
└── inventory/
    ├── main/           # public entry points and orchestration
    │   └── reserve.py
    ├── _helpers/       # private implementation phases
    ├── classes/        # one class per module
    ├── models.py       # data models
    ├── types.py        # type declarations
    ├── constants.py
    └── exceptions.py

A leaf needs meaningful main/ behavior, not an empty placeholder. Domains can split into named subdomains; tests mirror the code they cover. Tooling has its own scope. See the architecture model for complete role contracts, layouts, and defaults.

Explore the repository

Command	Purpose
fensu rule FFL102	Explain a rule and its effective settings.
fensu map submit --root src --depth 2	Show proven static Python calls, without running the program.
fensu dupes --since origin/main --diff	Review changed duplicated-code clusters; advisory, not a blocking check.
fensu skills --check	Verify installed guidance matches the current policy.

Call mapping needs neither Fensu configuration nor rule adoption. It marks unresolved calls, cycles, and depth limits instead of guessing dynamic targets. See the map reference and duplicated-code guide.

Documentation

fensu.io/docs is the full manual; the README is the starting point.

Start here	Reference
Quickstart	Installation, initialization, and reading diagnostics.
Adoption	Roll out enforcement safely on existing code.
Configuration	Targets, ownership roots, selection, thresholds, exceptions, and ignores.
Architecture	Roles, domain boundaries, tooling, and test layout.
Rust · Web	Analyzer-specific setup and rule packs.
CLI · Agent guidance	Commands, execution, and skill update safety.
Caching	Cache identities, controls, and safe cleanup.
Custom rules · RuleContext API	Author, configure, and test project-specific policies.

Python dead-code enforcement has a separate repository guide covering reachability, explicit dynamic roots, and adoption caveats.

Part of SQLStack

Fensu is part of SQLStack, a family of tools for data engineers.

• SQLBuild — SQL transformations with compile-time checks, end-to-end tests, and safe renames.
• SQLClean — Fast SQL linting and deterministic formatting.
• SQLTide — Build, replay, and safely deploy ClickHouse streaming pipelines with SQL.