UV and Ruff using the same pyproject.toml

[TheXer]

Hello,

I have a somewhat specific use case and a question about it. Is it possible to configure both UV and Ruff to use the same pyproject.toml file without syncing all dependencies? What I want is a single pyproject.toml that defines:

• The project metadata (name, version, Python version, etc.)
• Project dependencies
• Tool configuration (e.g., ruff, mypy, etc.)

For example:

[project]
name = "random_name"
version = "0.1.0"
requires-python = "==3.12.*"
dependencies = [
    "loguru==0.7.2",
    ...
]

[tool.ruff]
line-length = 120
lint.ignore = ["PLR0913", "RET504", "PLR0912"]
lint.select = ["E", "F", "UP", "I", "BLE", "T20", "RET", "SIM"]
exclude = ["*.yaml", "*.yml", "*.pyi"]

[tool.ruff.lint.per-file-ignores]
"Test/*" = ["PLR0911", "PLR0915"]

When I use uv to create a virtual environment based on this file, it installs all dependencies. That’s fine for development.

However, for testing and static analysis (e.g., ruff, mypy), I want to use uv to run these tools in isolation without installing the full dependency list. For example, using uv tool run ruff check . - this creates a lightweight venv, but the tool doesn’t use the settings from pyproject.toml—as expected, per the docs. On the other hand, if I run uv venv first, it syncs all project dependencies, and only then will ruff pick up the config from pyproject.toml.

My question is:
Is there a way to run tools like ruff or mypy with uv tool run in a way that still honors the settings from pyproject.toml, without triggering a full sync of all project dependencies?
UV is fast, so speed isn’t the problem—but with 30+ dependencies, it feels wasteful to download and install them just to run static analysis that doesn't require them.

Thanks!

EDIT:
To clarify why this matters—it's mainly for use in CI pipelines. In that context, I want the setup to be as lightweight as possible, while still respecting the pyproject.toml configuration. However, I want to avoid syncing and installing all project dependencies, since they're not needed just to run tools like ruff or mypy.

[ntBre]

Thanks for opening a discussion!

For example, using uv tool run ruff check . - this creates a lightweight venv, but the tool doesn’t use the settings from pyproject.toml—as expected, per the docs

Is that expected? That seems surprising to me, and ruff seems to respect my pyproject.toml when I tried this locally.

try.py that I had laying around from another issue:

import random

from tenacity import Retrying, retry_if_exception_type, stop_after_attempt


def try_lottery() -> float:
    for attempt in Retrying(
        retry=retry_if_exception_type(RuntimeError), stop=stop_after_attempt(3)
    ):
        with attempt:
            if random.random() < 0.001:  # Some stochastic condition
                return 1e6
            raise RuntimeError("The lottery was not won.")

it should trigger RET503, a non-default rule.

pyproject.toml enabling the rule:

[project]
requires-python = ">=3.13"

[tool.ruff]
lint.select = ["RET"]

uv command and output:

> uv tool run ruff check --no-cache .
try.py:6:1: RET503 Missing explicit `return` at the end of function able to return non-`None` value
   |
 6 | / def try_lottery() -> float:
 7 | |     for attempt in Retrying(
 8 | |         retry=retry_if_exception_type(RuntimeError), stop=stop_after_attempt(3)
 9 | |     ):
10 | |         with attempt:
11 | |             if random.random() < 0.001:  # Some stochastic condition
12 | |                 return 1e6
13 | |             raise RuntimeError("The lottery was not won.")
   | |__________________________________________________________^ RET503
   |
   = help: Add explicit `return` statement

Found 1 error.
No fixes available (1 hidden fix can be enabled with the `--unsafe-fixes` option).

[TheXer]

You're absolutely right - thank you so much for taking the time to test this. I realize now that I was blinded by this paragraph in the documentation:

If you are running a tool in a project and the tool requires that your project is installed, e.g., when using pytest or mypy, you'll want to use uv run instead of uvx. Otherwise, the tool will be run in a virtual environment that is isolated from your project.

The link to the paragraph (under Note).

Based on that, I assumed something wasn’t working correctly on my end, because uvx is regarded as a drop-in replacement for uv tool run. I also misunderstood the term “project” in the documentation - I thought it referred specifically to the pyproject.toml file, rather than the broader concept of a development project, if that makes sense. And also I had newer version installed (because I was using uv tool ruff check .) instead of picking the specific version that I have in my project (uv tool [email protected] check .) so I was confused by the newer syntax.

After some playing around and testing, it is working correctly and my question is then answered :)