`base-cli`

base-cli is the PyPI distribution; import it in Python as base_cli.

Install it with:

python -m pip install base-cli

Release builds, TestPyPI rehearsals, and protected PyPI publication are documented in docs/releasing.md. The package exposes base_cli.__version__, which matches the distribution version.

The package is distributed under the Apache License 2.0. Base itself remains licensed separately under AGPL-3.0-or-later.

base_cli is a small Python framework for writing command-line tools with a consistent lifecycle. It is designed to be embedded by applications rather than to define an application's project model. Base is one consumer of the library, not part of its generic contract.

It is intentionally thin. Click still owns argument parsing and command execution, while base_cli provides reusable lifecycle behavior:

standard command options such as --debug, --quiet, --environment, --config, --keep-temp, and --log-file
structured logging to stderr and, by default, to a persistent per-run log file
optional project discovery and configuration policies supplied by the consumer
per-run temp directories, persistent cache directories, and cleanup hooks
sensitive argument redaction in debug invocation logs
a command context object shared by command code and helper functions
test helpers built on Click's CliRunner

Design Goals

CLI tools should be easy to write, but not magical. A command should be explicitly registered, receive an explicit Context, and use standard Python functions instead of import-time side effects.

The package follows these rules:

Decorator-driven setup: commands opt in by creating an App and decorating a function.
Logs go to stderr: user-facing program output can stay on stdout, while logs remain redirectable and skippable.
Every run has a context: logs, paths, configuration, environment, and cleanup are available through one object. Project metadata is available when the selected consumer profile supplies it.
No import-time filesystem writes: state directories are created only when a command runs.
Consumer-neutral, Click-compatible: command authors keep using familiar Click concepts such as options and arguments.

Consumer Profiles

App accepts a CliProfile that supplies the policies which vary between applications: project discovery, configuration, runtime placement, and optional history persistence.

Standalone consumers use the generic profile by default. It can also be passed explicitly when making the policy boundary visible:

app = base_cli.App(
    name="hello",
    version="0.1.0",
    profile=base_cli.CliProfile.generic(),
)

The generic profile has no manifest filename convention, no product-owned configuration directory, and no implicit history writer. Applications can provide those policies through callbacks or build their own profile. The consumer-owned adapters should supply any product-specific policies. See docs/consumer-profiles.md for the boundary and migration guidance.

Public API

The supported facade is import base_cli. It exports the command lifecycle (App, Context, run_app, decorators, and logging helpers), command filters, and the structured command protocol helpers. Consumer-owned user configuration is passed through Context.user_config; the library does not impose a schema. The corresponding modules are also available as base_cli.command_filters, base_cli.command_protocol, and base_cli.history.

The command protocol owns only generic framing, field validation, and schema registration. It ships with no application record types and uses COMMAND_PROTOCOL_V1 by default. A consumer can register its own schemas and pass a compatibility protocol_header when it must interoperate with an existing peer protocol.

Command filters use consumer-neutral name normalization by default. Consumers with legacy command names can pass a normalizer callback to normalize_command_filter, normalize_command_filters, and command_matches to define compatibility aliases or prefixes.

Low-level implementation helpers are intentionally not included in the module __all__ surfaces. Downstream code should use the documented facade or the explicitly supported symbols from those modules.

Minimal Command

from __future__ import annotations

import base_cli


app = base_cli.App(
    name="hello",
    version="0.1.0",
    profile=base_cli.CliProfile.generic(),
)


@app.command()
@base_cli.option("--name", required=True)
def main(ctx: base_cli.Context, name: str) -> None:
    ctx.log.info("starting hello")
    print(f"hello {name}")


if __name__ == "__main__":
    raise SystemExit(base_cli.run_app(app))

Running this command directly as a Python package automatically adds the standard options:

hello --name Ada
hello --debug --name Ada
hello --quiet --name Ada
hello --environment prod --name Ada
hello --keep-temp --name Ada
hello --log-file /tmp/hello.log --name Ada

Long options use Click's native syntax, so both --name Ada and --name=Ada are accepted. These options belong to the application-level lifecycle. A consumer may expose them through its own launcher or compose them with a higher-level command wrapper.

Command Registration

Use App when you want a named command:

app = base_cli.App(name="workspace-tools", version="0.1.0")

Register the command function explicitly:

@app.command()
def main(ctx: base_cli.Context) -> None:
    ...

The command function always receives ctx as its first argument. User-defined options and arguments are passed after the standard lifecycle options have been removed from Click's keyword arguments.

For small scripts, the module-level decorators are available:

@base_cli.command()
def main(ctx: base_cli.Context) -> None:
    ...

Prefer an explicit App when command names, versions, or consumer policies should be visible at the top of the module.

Use @app.subcommand() when one CLI needs multiple verbs while keeping the standard context, logging, redaction, and cleanup lifecycle for each invocation:

app = base_cli.App(
    name="workspace-tools",
    version="0.1.0",
    help="Inspect and sync workspace projects.",
)


@app.subcommand()
@base_cli.argument("project")
def status(ctx: base_cli.Context, project: str) -> None:
    ctx.log.info("checking %s", project)


@app.subcommand("sync")
@base_cli.option("--dry-run", is_flag=True)
def sync_project(ctx: base_cli.Context, dry_run: bool) -> None:
    if ctx.dry_run:
        ctx.log.info("previewing sync")

Subcommands use the same base_cli.option() and base_cli.argument() metadata as single commands. App(help=...) appears in the command group's --help output. For subcommand apps, prefer standard options before the subcommand name, for example workspace-tools --debug status demo. The post-subcommand form, such as workspace-tools status --debug demo, remains accepted for compatibility. Use either @app.command() for a single-command CLI or @app.subcommand() for a command group; do not mix the two registration styles on one App.

Options And Arguments

base_cli.option and base_cli.argument mirror Click's decorators:

@app.command()
@base_cli.argument("project")
@base_cli.option("--workspace", type=str)
def main(ctx: base_cli.Context, project: str, workspace: str | None) -> None:
    ...

Use sensitive=True for options whose values should not appear in invocation logs:

@base_cli.option("--token", sensitive=True, required=True)
def main(ctx: base_cli.Context, token: str) -> None:
    ...

Both --token secret and --token=secret are accepted and redacted in debug logs.

Use dry_run=True when a nonstandard option should drive ctx.dry_run and the lifecycle's default durable-write suppression:

@base_cli.option("--preview", is_flag=True, dry_run=True)
def main(ctx: base_cli.Context, preview: bool) -> None:
    if ctx.dry_run:
        ctx.log.info("previewing changes")

The conventional dry_run parameter is recognized automatically, so commands using @base_cli.option("--dry-run", is_flag=True) do not need the marker. Only one option on a command may be marked dry_run=True; duplicate dry-run markers fail during command registration so authors do not accidentally ship an option that is ignored by ctx.dry_run.

Standard Options

Every base_cli.App command gets these options:

--debug: enable DEBUG logging on the user-facing stderr stream.
--quiet, -q: suppress INFO logs on the user-facing stderr stream.
--environment <name>: set ctx.environment for the run.
--config <path>: merge an additional YAML config file.
--keep-temp: preserve the run's temp directory after command completion.
--log-file <path>: write the persistent log to a specific file.
--version: shown when the App was created with a version.

The command receives only its own application-specific options. Standard options are consumed before the command function is called.

Exit Codes

Use base_cli.ExitCode when command code or tests need to name standard command result meanings:

ExitCode.SUCCESS (0): the command completed successfully.
ExitCode.FAILURE (1): the command was valid, but an operational problem prevented successful completion.
ExitCode.USAGE_ERROR (2): the command could not proceed because user input, configuration, or environment setup was invalid or incomplete.
ExitCode.INTERRUPTED (130): the user interrupted the command with Ctrl+C.

Existing commands can keep returning integers. New code should prefer the named constants when it makes intent clearer:

if ctx.project_root is None:
    ctx.log.error("run this command from a project recognized by the consumer")
    return base_cli.ExitCode.USAGE_ERROR

run_app() is the process boundary for production entry points. It preserves Click's messages and exit codes for usage and application errors, reports an explicit abort as 1, and reports Ctrl+C during startup or command execution as 130 without a traceback. After the command outcome has settled, history, metadata, and cleanup are best-effort teardown: even a second interrupt there cannot replace the primary result.

An unexpected exception returns 1 with a stable, detail-free message. The run ID and diagnostic-log path are included when context and file logging are available. The traceback is kept in the persistent log when enabled and is shown on stderr with an effective --debug setting. A failure before option parsing can provide a traceback only when --debug is an unambiguous leading flag; otherwise the message says that diagnostic context was unavailable. Embedding code that needs the original exception can pass the keyword-only reraise_unexpected=True argument to run_app().

Command result or exception	`outcome`	Exit code	Default message
`None` or returned `0`	`success`	0	none
returned `2`	`usage_error`	2	none
another returned nonzero integer	`nonzero_return`	returned value	none
`click.UsageError`	`usage_error`	exception code	Click usage error
another `click.ClickException`	`click_error`	exception code	Click error
`click.Abort`	`aborted`	1	`Aborted!`
`Ctrl`+`C`	`interrupted`	130	`Interrupted.`
`SystemExit`	`system_exit`	normalized payload	string payload, if any
another unexpected exception	`unexpected_error`	1	stable internal-error message

For SystemExit, a missing payload becomes 0, an integer payload is preserved, and any other payload is printed and normalized to 1.

Context

Context is the object command code should pass around instead of rediscovering runtime paths or global settings.

Important fields include:

ctx.cli_name: normalized CLI name used for state paths and logger names.
ctx.run_id: timestamp plus short random suffix for this invocation.
ctx.application_home: optional application home supplied by the profile.
ctx.project_root: project root returned by the profile, when any.
ctx.workspace_root: optional workspace root supplied by user configuration.
ctx.manifest_path: project metadata path returned by the profile, when any.
ctx.history_scope: history scope supplied by the profile or its compatibility adapter.
ctx.history_parent_run_id: optional parent invocation ID supplied by the consumer.
ctx.runtime_owner: consumer-defined runtime owner; the generic profile uses default.
ctx.owner_root: application namespace root under the configured cache root.
ctx.run_root: this invocation's run bundle.
ctx.state_dir: owner root (compatibility alias).
ctx.log_dir: run-bundle log directory.
ctx.cache_dir: persistent component cache directory.
ctx.temp_dir: per-run temp directory inside the bundle.
ctx.log_file: the run's shared logs/primary.log, or None when persistent logging is disabled.
ctx.config: merged configuration dictionary.
ctx.user_config: opaque consumer-owned user configuration returned by the profile, or None for the generic default.
ctx.environment: active environment, defaulting to dev.
ctx.debug: whether debug logging is enabled for the stderr stream.
ctx.quiet: whether INFO logs are suppressed on the stderr stream.
ctx.dry_run: whether the command is running in a no-durable-write mode.
ctx.keep_temp: whether ctx.temp_dir should survive cleanup.
ctx.log: standard Python logger configured by base_cli.

Helpers can retrieve the active context without threading it through every call:

from base_cli import get_current_context


def helper() -> None:
    ctx = get_current_context()
    ctx.log.debug("helper is running")

get_current_context() is valid only while a base_cli.App command is running.

Logging

base_cli configures two handlers:

a user-facing stderr handler at INFO by default, DEBUG with --debug, or WARNING with --quiet / -q
a persistent file handler that records DEBUG logs when persistent logging is enabled

--quiet suppresses INFO output on the user-facing stream but still shows warnings and errors. --debug and --quiet cannot be used together. Persistent log files still receive DEBUG-level detail, including INFO messages suppressed from stderr. User-facing logs use colors automatically on interactive terminals; persistent log files remain plain text. Set NO_COLOR=1 or BASE_CLI_COLOR=0 to disable colors. A consumer wrapper may add its own color option and map it to the environment variable.

Click also provides shell completion. For an app named hello, request a completion script with _HELLO_COMPLETE=bash_source hello, replacing bash with zsh or fish as needed. base_cli leaves installation to the caller so shell startup files remain under user control.

Advanced tests and CI wrappers can call base_cli.configure_logger(..., stream=..., formatter=...) to capture user-facing logs or apply a custom formatter. Leave those arguments as None to keep the default stderr stream and formatter. Log timestamps use the host's local timezone and include its numeric offset by default. A consumer can set LOG_UTC=1 to use UTC and include an explicit UTC marker.

This setting affects log presentation only. Run metadata, history records, and run IDs retain their canonical UTC representation.

Commands that inspect runtime artifacts can use base_cli.App(log_to_file=False) to keep the standard context, --debug, and --quiet behavior without creating default logs/, cache/, or tmp/<run-id>/ directories. An explicit --log-file <path> still enables file logging for that invocation.

Commands running with ctx.dry_run also skip default logs/, cache/, and tmp/<run-id>/ creation. Passing --log-file <path> still writes to that explicit file so tests and diagnostics can inspect dry-run logs when needed.

The generic profile does not write command history. A profile may provide a history writer to persist redacted command metadata, timing, exit status, project context, and a pointer to the raw log file. History writes should be best-effort and should not fail the user command when an index cannot be updated.

High-frequency tools can set base_cli.App(max_log_files=<count>) to keep at most that many default persistent log files across the owner's run bundles. Retention runs during startup after the current run's default log file is resolved, and the current run's log file is never pruned. The policy is skipped for ctx.dry_run, log_to_file=False, and explicit --log-file paths so no-durable-write modes and caller-selected log locations stay under caller control. Use this as a small guardrail for busy local tools; an application can provide broader maintenance commands for caches, logs, and retained temp files.

Logs use a stable, human-readable shape:

2026-05-26 12:34:56 INFO    path/to/file.py:42 message

Use either ctx.log directly:

ctx.log.info("processed %s items", count)

or the convenience functions:

base_cli.log_debug("cache_dir=%s", ctx.cache_dir)
base_cli.log_info("done")
base_cli.log_warning("using fallback")
base_cli.log_error("failed")

Program output should still use stdout when another command might consume it. Logs should stay on stderr so users can redirect or ignore logs without losing the real command output.

Config Precedence

The generic profile has no implicit configuration files. It loads the file passed through --config, when present, and otherwise starts with an empty configuration dictionary. Standard command-line options are applied by the lifecycle after the profile's configuration is loaded; for example, --environment prod overrides environment: dev from an explicit configuration file.

ctx.config exposes the dictionary returned by the profile. ctx.user_config exposes the opaque user-configuration value returned by the profile. Consumers that need user files, project files, environment variables, or a merge precedence must implement those policies in CliProfile.load_config and CliProfile.load_user_config; base_cli does not define the value's fields.

Project Discovery

The generic profile does not discover projects or assume a manifest filename. Its ctx.project_root and ctx.manifest_path fields are None unless the consumer supplies a discover_project policy. A profile can discover projects from a manifest, workspace, repository metadata, or any other application-owned source and return a ProjectInfo value.

Commands that require a project should validate the profile-provided value explicitly and return a clear usage error or actionable message.

Runtime Directories

The generic profile uses the configured cache root and an application namespace to create per-run logs, caches, and temporary directories. Pass cache_root to CliProfile.generic() for deterministic placement in tests or applications; otherwise the platform cache directory is used. Linux and WSL2 follow XDG_CACHE_HOME or ~/.cache, macOS uses ~/Library/Caches, and Windows uses %LOCALAPPDATA% (falling back to ~/AppData/Local). Set BASE_CLI_CACHE_DIR to override the default on any platform. The generic profile does not prescribe a product-wide cache name or cleanup command.

Each lifecycle-owned invocation is a run bundle containing a private run.json, logs/, and tmp/, while persistent component caches live in the owner's cache directory. When persistence succeeds, run.json begins with status: "running" and is finalized with status, outcome, exit_code, ended_at, and duration_ms, including command failures and interruptions. The stable outcome values are success, usage_error, nonzero_return, click_error, aborted, interrupted, system_exit, and unexpected_error. Terminal-write failures are warnings and cannot change the process result; base-cli then removes a matching or corrupt owned record on a best-effort basis so history data cannot masquerade as authoritative core data.

Parsing errors, help, and version requests occur before the command lifecycle owns a bundle and therefore do not create one. Neither do inherited runtimes, log_to_file=False, or dry-run invocations; an explicit log path can still receive diagnostics in the latter two modes. Context startup is transactional: if directory creation, logger setup, or retention fails, base-cli closes partially installed handlers and removes new bundle-local temp/log artifacts and empty directories. Pre-existing content, persistent component caches, and parent-runtime data are preserved.

On POSIX, base-cli enforces owner-only 0600/0700 modes. On Windows, the default user-local cache root relies on inherited user-profile ACLs; consumers using a custom cache root must provide the appropriate ACL themselves.

See Platform support for the supported Linux, WSL2, macOS, and native Windows boundaries. Native Windows support covers the generic base-cli framework; it does not imply native Windows support for Base or basectl.

Use ctx.on_cleanup() for cleanup work that should happen even when helper code does not own the main command wrapper:

def close_connection() -> None:
    connection.close()


ctx.on_cleanup(close_connection)

Cleanup hooks run before temp directory removal. Hook failures are logged as warnings and do not prevent later hooks from running.

Testing

Use base_cli.testing.invoke for unit tests:

from pathlib import Path

from base_cli.testing import invoke


def test_command(tmp_path: Path) -> None:
    project = tmp_path / "project"
    project.mkdir()

    result = invoke(
        app,
        ["--name", "Ada"],
        home=tmp_path,
        cwd=project,
    )

    assert result.exit_code == 0
    assert "hello Ada" in result.stdout

The helper wraps Click's CliRunner but routes the invocation through the same run_app() boundary used by production entry points. Option parsing (including native forms such as --name=Ada), effective and logged argv, exit-code and error normalization, lifecycle behavior, and command-group dispatch therefore follow the production path.

By default, unexpected exceptions receive the production-safe rendering and exit code. Pass the keyword-only reraise_unexpected=True argument when a test needs the original unexpected exception in result.exception:

result = invoke(app, [], home=tmp_path, reraise_unexpected=True)
assert isinstance(result.exception, RuntimeError)

As with direct CliRunner use, a handled nonzero exit normally also gives Result.exception a SystemExit carrying that exit code. That does not by itself indicate an unexpected crash; assert result.exit_code and the rendered stdout or stderr for expected usage or application failures.

invoke() sets HOME plus the relevant USERPROFILE, LOCALAPPDATA, and XDG_CACHE_HOME values when requested, and supplies cwd to the invocation for the duration of the test. Calls that use cwd are serialized and the caller's cwd is restored afterward, but this remains process-global: do not use it concurrently with code that changes cwd outside invoke() or from threads spawned by the invoked command. A generic profile should receive project fixtures through its discover_project callback. The helper does not create or interpret any product-specific manifest fixture.

When home is supplied, invoke() provides an isolated default cache environment for tests. Pass env={"BASE_CLI_CACHE_DIR": str(path)} when a test needs an explicit cache location.

When To Use `base_cli`

Use base_cli for Python commands that need a predictable command lifecycle: standard options, logging, redaction, runtime state, cleanup, and test helpers. Standalone consumers should use CliProfile.generic() or provide an explicit profile with their own project and configuration policies.

It is a good fit for:

project discovery commands
setup and artifact management commands
developer workflow commands
CLIs that need predictable logs, temp directories, and config precedence

It is not meant to replace Click, Typer, argparse, or rich terminal UI frameworks. It is the reusable layer around command lifecycle, context, logging, configuration, and state.

Name		Name	Last commit message	Last commit date
Latest commit History 110 Commits
.github		.github
docs		docs
lib/python/base_cli		lib/python/base_cli
scripts		scripts
tests		tests
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
VERSION		VERSION
base_manifest.yaml		base_manifest.yaml
pyproject.toml		pyproject.toml

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Repository files navigation

`base-cli`

Design Goals

Consumer Profiles

Public API

Minimal Command

Command Registration

Options And Arguments

Standard Options

Exit Codes

Context

Logging

Config Precedence

Project Discovery

Runtime Directories

Testing

When To Use `base_cli`

About

Releases

Packages

Contributors

Languages

Uh oh!

Folders and files

Latest commit

History

Repository files navigation

base-cli

Design Goals

Consumer Profiles

Public API

Minimal Command

Command Registration

Options And Arguments

Standard Options

Exit Codes

Context

Logging

Config Precedence

Project Discovery

Runtime Directories

Testing

When To Use base_cli

About

Resources

Contributing

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages

`base-cli`

When To Use `base_cli`