skilium-eda 0.1.0

Lightweight agentic EDA library for data science

skilium-eda

Lightweight agentic EDA library for data science. From raw data to insights, charts, and reports in one command.

Install

pip install skilium-eda

With S3 support:

pip install skilium-eda[s3]

Quick Start

Python API

from skilium_eda import DataEngine, EDAPipeline

# Load and explore
df = DataEngine.load("data.csv")
summary = DataEngine.get_summary(df)
profile = DataEngine.profile(df)

# Full pipeline
pipeline = EDAPipeline("data.csv", output_dir="./reports")
results = pipeline.run()

S3 Support

# Load from S3
df = DataEngine.load("s3://my-bucket/datasets/sales.csv")

# Full pipeline from S3
pipeline = EDAPipeline("s3://my-bucket/datasets/sales.csv", output_dir="./reports")
results = pipeline.run()

S3 authentication uses standard AWS credential chains (env vars, ~/.aws/credentials, IAM roles).

CLI

# Full pipeline
skilium-eda run data.csv --output-dir ./reports

# Profile only
skilium-eda profile data.csv --output profile.json

# S3
skilium-eda run s3://bucket/data.csv --output-dir ./reports

Features

• Load: CSV, Excel, Parquet, JSON/JSONL — local files and S3 URLs
• Clean: Missing values, duplicates, type inference, outlier removal
• Profile: Statistics, correlations, distributions, quality metrics
• Agent: Rule-based insights, chart selection, action items
• Visualize: Distributions, correlations, missing values, boxplots, pairplots
• Report: Self-contained HTML with embedded charts, Markdown export

Project Structure

src/skilium_eda/
  __init__.py    # Clean exports
  core.py        # DataEngine: load, clean, profile (with S3)
  agent.py       # EDAAgent: insights, chart selection
  viz.py         # ChartEngine: matplotlib/seaborn charts
  report.py      # HTML and Markdown report generation
  pipeline.py    # EDAPipeline: orchestrate full workflow
  cli.py         # skilium-eda CLI commands

6 modules, ~2,800 lines, production-ready.