taggin 0.1.1

Taggin logging with dynamic tags, structured storage, and CLI search.

Taggin

Welcome

Taggin is a tiny layer on top of the Python stdlib logger that treats attribute access as dynamic tags (log.TRAIN.BATCH("...")). Tags can be filtered via glob patterns, assigned custom log levels, rate limited, and every log is mirrored into a structured store that can be searched or exported.

Installation

Install via pip (or pixi/uv) with:

pip install taggin

The CLI depends on arrow and cyclopts (included), colorful console output uses rich, and saving/reading Parquet requires pandas plus a backend such as pyarrow or fastparquet.

Usage

Quick start

from datetime import datetime
from taggin import (
    setup_logger,
    get_log_storage,
    set_visible_tags,
    set_tag_style,
)

log = setup_logger(enable_color=True)  # enables Rich-powered colors where supported
set_visible_tags(["TRAIN.*", "io.net"])
set_tag_style("TRAIN.START", color="green", emoji="🚂")

log.info("Booting")
log.TRAIN.START("epoch=%s", 1)
log.io.net("connected to redis")

storage = get_log_storage()
storage.save_text("logs/run.txt")                # new file
storage.save_text("logs/run.txt", append=True)   # append to existing file
storage.save_parquet("logs/run.parquet")         # requires pandas + pyarrow

recent = storage.search_by_date(start=datetime.utcnow().replace(hour=0, minute=0))
by_tag = storage.search_by_tag("TRAIN.*")
approx = storage.search_fuzzy("redis connection failed", threshold=0.5)

When enable_color=True, tags render in color (and can add emoji via set_tag_style). Disable the flag to fall back to plain text.

All structured entries store the timestamp, level, logger name, tag, and message so they remain queryable even if the original message contains its own time or date. This makes ad-hoc debugging easy whether you prefer grepping the text artifact or using a DataFrame/Parquet workflow.

CLI search utility

A small cyclopts-powered CLI is available for exploring saved logs without writing Python. Assuming you saved either a structured text log or Parquet file:

taggin by-tag logs/run.txt "TRAIN.*"
taggin by-date logs/run.parquet --start "2025-01-01" --end "2025-01-05"
taggin fuzzy logs/run.txt "redis timeout" --threshold 0.4 --limit 5
taggin tags logs/run.txt                    # list all known tags
taggin by-tag logs/run.txt "TRAIN.*" --json-output   # machine-friendly

Each command prints matching entries in the same concise [TAG] message style, or JSON (when --json-output is provided) for downstream scripting.

Tests

Run the small pytest suite (which also exercises the Parquet writer when pandas is available) with:

pytest

Documentation

This repo ships MkDocs docs (Home, Examples, API Reference). Preview locally:

mkdocs serve

or build static files via mkdocs build. The content lives under docs/.