tempest-cli

Documentação: Português (BR) · English (US)

One command for the quality gate of any Python project — ruff + mypy + pytest, with a typing-strictness dial that lives in your pyproject.toml instead of in four different Makefile targets.

Framework-agnostic on purpose: Django, Flask, Litestar, FastAPI, a library, a script. The only runtime dependency is typer.

uv add --dev tempest-cli

tempest-cli check          # lint + fmt-check + type + test, in order, stops at the first failure
tempest-cli fix            # every ruff autofix, then format
tempest-cli type -s strict # override the configured strictness for one run

tc check                   # `tc` is the short alias for the same program

Why it exists

The four commands are always the same, and every project rewrites them slightly differently — a Makefile here, a tox.ini there, a CI job that drifts from what runs locally. tempest-cli check is the same gate on your machine and in CI, and --strictness makes "how much typing do we enforce" a value in pyproject.toml rather than a flag someone remembered to pass.

Commands

Every command takes an optional path (tempest-cli lint src/) and returns the underlying tool's exit code, so CI reads it exactly as it would read ruff directly.

Typing strictness

[tool.tempest]
typing_strictness = "strict"   # lenient | standard | strict

The level adds flags on top of your own [tool.ruff] / [tool.mypy] — it never relaxes what you already configured:

ANN401 is never enabled at any level: Any is a legitimate annotation. The levels enforce that things are annotated, never that they avoid Any.

Override per run with --strictness / -s. Absent config means standard.

PR descriptions from the branch itself

tempest-cli pr-prompt | claude -p
tempest-cli pr-prompt develop --lang en --out pr_prompt.txt

The prompt carries the repository's own pull-request template (or a bundled PT-BR / EN-US default), the rules that stop a model from handing back the template with its placeholders intact, and the branch context: commit subjects, changed files, and a bounded excerpt of each patch.

Diffs are read as base...head — the merge-base diff the forge shows — so commits that landed on the base after the branch started are not attributed to it. Whatever the bounds leave out is stated inside the prompt, so a partial diff reads as partial.

Use it as a library

from tempest_cli import load_tempest_config, run_full_check

config = load_tempest_config()
exit_code = run_full_check(".", config=config)

And to expose the same gate from your own CLI, without copying command bodies:

import typer

from tempest_cli.main import register_commands

cli: typer.Typer = typer.Typer(name="mytool")
register_commands(cli)

Installing the tools

tempest-cli shells out to whatever ruff, mypy and pytest it finds — it does not pin them, so your project chooses the versions. The lookup runs in this order:

1. the environments of the run — the interpreter's own directory, then $VIRTUAL_ENV, then the nearest .venv up the tree;
2. PATH, skipping a pyenv/asdf shim that dispatches nowhere (the one that answers pyenv: ruff: command not found);
3. uv run --with <tool> <tool>, when uv is available.

To install the three alongside it:

uv add --dev "tempest-cli[tools]"

Relationship with tempest-fastapi-sdk

This package was extracted from tempest-fastapi-sdk, where the same gate shipped as tempest check. Getting it meant installing FastAPI, SQLAlchemy, Alembic and Pydantic — 38.7 MB of dependencies and roughly 0.5 s of import time per invocation, for four commands that never touch any of it.

tempest check keeps working: the SDK depends on this package and registers the same commands. Both stay in sync because there is now one implementation.

License

MIT — see LICENSE.