A modular collection of Python utilities for Typer, Pydantic, and Debugging. Built for high-performance developer workflows using uv and hatchling.
pytools is designed as a "buffet" library—install only the submodules you need to keep your project dependencies lean.
Add pytools to your project using uv. You can choose specific "extras" to pull in only the dependencies required for that module.
# Install just the Typer utilities (and their deps)
uv add "pytools[typer] @ git+ssh://git@github.com/buchanan-solutions/pytools.git"
# Install everything
uv add "pytools[all] @ git+ssh://git@github.com/buchanan-solutions/pytools.git"
| Extra | Description | Key Dependencies |
|---|---|---|
core |
Foundational logic used by all modules. | None (Std Lib only) |
typer |
CLI helpers, context management, and rich logging. | typer, rich, python-dotenv |
pydantic |
Custom validators and base model configurations. | pydantic |
pydantic-settings |
Environment management and config loading. | pydantic-settings |
debugging |
Enhanced print utilities and object inspection. | icecream |
Phased logging: build a pure LoggingState with bootstrap, adjust it, then call apply_state to activate the stdlib runtime. The same state can be exported for worker processes (for example Uvicorn log_config).
Quick usage:
import logging
from pytools.logging import bootstrap, merge_logger_levels, set_root_level
from pytools.logging.apply import apply_state
state = bootstrap(root_level=logging.INFO, rich=True)
state = merge_logger_levels(state, {"my.package": logging.DEBUG})
state = set_root_level(state, logging.DEBUG)
apply_state(state)Important — handlers are not objects in config: bootstrap(..., handlers=...) and LoggingState.handlers follow the standard library’s logging.config.dictConfig shape. You pass handler definitions (dicts), not live logging.Handler instances. Custom handlers are wired with a "()": "dotted.import.path.to_callable" entry: that string must resolve to a callable that returns a logging.Handler (often a zero-argument function). At apply time, dictConfig imports that callable and invokes it. Closures over module-level state in your package are fine; the constraint is that the config tree itself must stay serializable as plain dicts and strings (so worker processes and export_config stay coherent).
Default console handler when you omit handlers: colored stderr (create_colored_level_handler), or Rich (create_rich_console_handler) when rich=True.
For step-by-step examples (CLI, Uvicorn, custom handler factories), see src/pytools/logging/USAGE.md.
from pytools.typer.utils.get_from_ctx import get_from_ctx
from my_app.models import Config
@app.command()
def run(ctx: typer.Context):
config = get_from_ctx(ctx, "config", Config)from pytools.debugging.debug_print import debug_print
data = {"status": "success", "meta": [1, 2, 3]}
debug_print(data)This project uses the src layout and hatchling backend.
# Clone and sync environment
git clone https://github.com/your-org/pytools.git
cd pytools
uv sync --all-extras
# Run tests
uv run pytest
MIT © 2026 Buchanan Information Systems Inc.