Logging and Console Output

console is your task’s output and always prints. Logs are diagnostics gated by verbosity, from -v to -vvv, with per-module levels and a JSON format.

bakefile has two output channels, and they don’t affect each other:

• console is your task’s user-facing output (results, status). It always prints, regardless of verbosity.
• Logs are bakefile’s own diagnostics, plus any logging calls you make in tasks. These are gated by verbosity.

Console output

console (imported from bake) is a thin wrapper over Rich. Print task output through it rather than print:

from bake import Bakebook, command, console


class MyBakebook(Bakebook):
    @command()
    def status(self) -> None:
        console.echo("Building...")  # plain output, stdout
        console.success("Build done")  # ✅ SUCCESS   (stderr)
        console.warning("Low disk")  # ⚠️ WARNING   (stderr)
        console.error("Build failed")  # ❌ ERROR     (stderr)

Helpers:

• console.echo(msg) prints to stdout (your task’s normal output).
• console.success, info, warning, error(msg) print a labeled line to stderr. In GitHub Actions, warning and error become ::warning:: / ::error:: annotations.
• console.prefix(msg, label=..., emoji_code=..., label_style=...) prints a custom labeled line when the built-in labels don’t fit. label is required - use info for the default INFO label.
• console.cmd(cmd_str) prints a command as ❯ <cmd>, which is what ctx.run uses to show the command it runs. The arrow accepts arrow_style= (default green, e.g. arrow_style="bold red" for failed commands). The command text always stays plain.
• console.script_block(title, script) pretty-prints a multi-line script, used by run_script.

Every helper accepts rich print kwargs (style, emoji, markup, highlight, …) and they apply to the message only - the label chrome is rendered as a Text object and is never affected. Chrome styling has its own param names (label_style on prefix, arrow_style on cmd) so rich kwargs are never shadowed. So console.success("tests[unit] passed", markup=False) keeps the green label but prints the message literally. Note that with markup on (the default), rich parses the message: [unit]-style tags are consumed.

Stream model: echo is for machine-readable data (stdout, safe to pipe). Everything else is human progress output (stderr). Any helper also accepts no_color=True to emit zero ANSI codes - use it when the output is parsed by another tool:

console.echo(value, no_color=True)  # clean stdout, no ANSI codes

For anything else, the raw Rich consoles are also exposed: console.out / console.err (stdout / stderr, color) and console.plain_out / console.plain_err (no color).

console output, plain print(), and command output are all separate from logs. They always print, regardless of verbosity.

Logging

bakefile logs through loguru, and all logs go to stderr. Standard-library logging is bridged into it, so any logging.getLogger(__name__).info(...) in your tasks honors the same settings:

import logging

logger = logging.getLogger(__name__)


@bakebook.command()
def task(self):
    logger.info("starting")  # visible at -vv and above

Three settings control logs. The first two decide what shows, and the third picks the format.

Verbosity

Verbosity sets the global floor. Anything below it is dropped. The default is 0 (silent).

Flag	Env	Level
(none)	BAKE_LOG_VERBOSITY=0	silent (no logs)
-v	BAKE_LOG_VERBOSITY=1	warning
-vv	BAKE_LOG_VERBOSITY=2	info
-vvv	BAKE_LOG_VERBOSITY=3	debug

bake build              # silent
bake -v build           # warning + error
bake -vv build          # adds info
bake -vvv build         # adds debug (everything)

Per-module levels

--bake-log (env BAKE_LOG) is a comma-separated list of level or module=level entries. It raises or lowers specific modules independent of verbosity. The default is warning,bake=debug,bakelib=debug,bakefile=debug (the tool’s internals and your bakefile.py at debug, everything else at warning):

# Set one level for everything (the root level, always required)
bake --bake-log debug build                        # all modules at debug
bake --bake-log warning build                      # all modules at warning

# Raise one module above the root
bake --bake-log warning,bake=debug build           # bake internals at debug, rest warning

# Target your own bakefile.py (it loads as module "bakefile")
bake --bake-log warning,bakefile=debug build       # your bakefile.py at debug

# Env var form: quiet bakefile.py, keep bake internals (restate them)
BAKE_LOG="warning,bake=debug,bakelib=debug,bakefile=warning" bake build

--bake-log replaces the default instead of merging with it, so restate any module you want to keep. In the last line, bake and bakelib stay at debug while bakefile.py is silenced to warning.

A log line shows only if it clears both the verbosity floor and its module’s level. So with the default BAKE_LOG, -v surfaces bakefile’s warnings and errors, and -vvv surfaces its debug logs too.

Format

--log-pretty / --no-log-pretty (env BAKE_LOG_PRETTY, default pretty) chooses between pretty colored text and JSON (one object per line, for CI and log shipping):

bake --no-log-pretty build     # JSON logs

Advanced

For advanced needs, override setup_logging() (e.g. a custom JSON sink like GCPJsonSink for GCP Cloud Logging) or get_bake_log_thread_local_context() (inject trace IDs into each log line) on your Bakebook.