typed-argparse-gen 0.3.0

A small typed parser generator for lightweight Python CLIs.

typed-argparse

typed-argparse is a small annotation-driven CLI helper for people who want typed parser generation without adopting a large framework.

It is aimed at simple command-line tools, scripting utilities, and lightweight LLM-facing CLIs where you want a small API surface, explicit Python functions, and standard argparse behavior instead of a large abstraction layer.

The core package has no required third-party runtime dependencies. The main API is build_parser(), which converts a typed function signature into an argparse-compatible parser. The package also includes Argument, Option, unwrap(), call(), parse_and_call(), and build_subcommand_parser(), plus optional cmd2 integration through with_annotated().

Why use it:

• small API surface
• typed function signatures instead of command classes or decorators everywhere
• easy argparse.Namespace -> kwargs conversion through unwrap()
• a good fit for short-lived tools and simple LLM CLI wrappers
• optional cmd2 integration when you want an interactive shell

Install

The package is published on PyPI as typed-argparse-gen and imported as typed_argparse.

uv add typed-argparse-gen

For cmd2 integration:

uv add "typed-argparse-gen[cmd2]"

Main feature: build_parser()

Use build_parser() when you want annotation-driven parser generation directly.

import argparse
from typing import Annotated

from typed_argparse import Argument, Option, build_parser


def greet(
    name: Annotated[str, Argument(help_text="Who to greet")],
    count: Annotated[int, Option("--count", "-c", help_text="Number of greetings")] = 1,
    loud: bool = False,
) -> None:
    pass


parser = build_parser(greet)
args = parser.parse_args(["Kelvin", "--count", "2", "--loud"])

assert isinstance(parser, argparse.ArgumentParser)
assert args.name == "Kelvin"
assert args.count == 2
assert args.loud is True

build_parser_from_function() is still available as an alias.

By default:

• parameters without defaults become positional arguments
• parameters with defaults become --option flags
• bool = False becomes store_true
• bool = True becomes store_false on --no-name
• Annotated[..., Argument(...)] and Annotated[..., Option(...)] override the defaults

Unwrapping and calling

Use unwrap() to convert a parsed argparse.Namespace into plain values. Use call() and parse_and_call() for the common dispatch path.

from typed_argparse import build_parser, call, parse_and_call, unwrap

parser = build_parser(greet)
namespace = parser.parse_args(["Kelvin", "--count", "2"])

kwargs = unwrap(namespace)
values = unwrap(namespace, as_tuple=True)

assert kwargs == {"name": "Kelvin", "count": 2, "loud": False}
assert values == ("Kelvin", 2, False)

assert call(greet, namespace) == ["Hello Kelvin", "Hello Kelvin"]
assert parse_and_call(greet, ["Kelvin", "--loud"]) == ["HELLO KELVIN"]

Subcommands

Use build_subcommand_parser() or parse_and_call() with a mapping for small multi-command CLIs.

from typed_argparse import build_subcommand_parser, parse_and_call

commands = {"greet": greet, "tag": tag}

parser = build_subcommand_parser(commands)
namespace = parser.parse_args(["tag", "release", "--labels", "stable"])

result = parse_and_call(commands, ["tag", "release", "--labels", "stable"])

Optional cmd2 support

If you are using cmd2, you can use the same metadata model with with_annotated(). Install the optional cmd2 extra first.

For typed code, prefer direct imports:

import cmd2
from typing import Annotated

from typed_argparse import Argument, Option, with_annotated


class App(cmd2.Cmd):
    def sport_choices(self) -> list[str]:
        return ["football", "basketball", "tennis"]

    @with_annotated
    def do_play(
        self,
        sport: Annotated[str, Argument(choices_provider=sport_choices)],
        venue: Annotated[str, Option("--venue", "-v")] = "home",
    ) -> None:
        self.poutput(f"{sport=} {venue=}")

Development

uv sync --group dev
uv run ruff check .
uv run ty check
uv run pytest
uv build

Release flow

This repo is configured for semantic versioning with conventional commits.

• Push conventional commits to main
• GitHub Actions runs tests and lint with uv
• python-semantic-release determines the next version, updates CHANGELOG.md, tags the release, and publishes to PyPI

Expected commit prefixes include feat:, fix:, and perf:. Breaking changes should use ! or a BREAKING CHANGE: footer.