ML Visualizer

A data toolkit for ingesting, transforming, querying, and visualizing machine learning datasets. Manage your entire data pipeline — from raw ingestion through profiling and validation — all from the command line.

Commands

Each data command (ingest, transform, query, etc.) works in two modes:

• - Without arguments — displays the 20 most recent entries of that type
• With arguments — saves the input as a new timestamped entry

Data Storage

All data is stored as plain-text log files in ~/.local/share/ml-visualizer/:

• - Each command type gets its own log file (e.g., ingest.log, transform.log, visualize.log)
• Entries are stored in timestamp|value format for easy parsing
• A unified history.log tracks all activity across command types
• Export to JSON, CSV, or TXT at any time with the export command

Set the ML_VISUALIZER_DIR environment variable to override the default data directory.

Requirements

• - Bash 4.0+ (uses set -euo pipefail)
• Standard Unix utilities: date, wc, du, tail, grep, sed, INLINECODE37
• No external dependencies or API keys required

When to Use

1. 1. Building a data pipeline journal — use ingest, transform, and pipeline to document each step of your ML data preparation workflow
2. Tracking data quality — use validate and profile to log validation checks and profiling runs, ensuring data integrity before model training
3. Logging visualization requests — use visualize to record what charts and plots you've generated for model diagnostics (confusion matrices, ROC curves, feature importance)
4. Managing dataset schemas — use schema to document the structure of your datasets, track schema changes over time, and share definitions with your team
5. Auditing data operations — use search, recent, and stats to review your complete data processing history and find specific operations

Examples

CODEBLOCK0

Output

All commands print results to stdout. Redirect to a file if needed:

CODEBLOCK1