featcat 0.1.0

AI-Powered Feature Catalog for Data Science teams

featcat

AI-Powered Feature Catalog for Data Science Teams

Tiếng Việt

featcat is a lightweight Feature Catalog designed for Data Science teams. It is not a Feature Store (no online serving) — it's a metadata management tool with an AI layer for searching, documenting, and monitoring feature quality.

The Problem

• Features scattered everywhere: Parquet files stored across local disks, S3, and MinIO — nobody knows what features exist
• Missing documentation: Dataset columns have no descriptions; new team members don't know what avg_session_duration means
• Hard to find the right features: Starting a new project (e.g. churn prediction) with no idea which features are already available
• Undetected data drift: Feature distributions change silently until model performance degrades

Key Features

Module	Description	Phase
Catalog	Register data sources, scan Parquet to auto-extract schema + stats	1
AI Discovery	Describe a use case → AI recommends relevant features + suggests new ones	2
Auto-doc	LLM automatically generates documentation for each feature	2
NL Query	Ask in natural language (English or Vietnamese), AI finds relevant features	2
Monitoring	PSI drift detection, null spikes, range violations	3
TUI	Terminal UI with dashboard, feature browser, AI chat	3
S3 Support	Read Parquet directly from S3/MinIO — never copies data locally	1
Caching	Cache LLM responses to speed up doc generation and NL queries	3

Quick Start

# 1. Clone and install
git clone https://github.com/codepawl/featcat.git && cd featcat
uv venv && source .venv/bin/activate
uv pip install -e ".[dev]"

# 2. Initialize catalog
featcat init

# 3. Register and scan a data source
featcat source add device_perf /data/features/device_performance.parquet
featcat source scan device_perf

# 4. Browse features
featcat feature list
featcat feature info device_perf.cpu_usage

# 5. (Optional) Enable AI features — requires Ollama
ollama serve &
ollama pull qwen2.5:7b
featcat discover "churn prediction for telecom customers"
featcat ask "features related to user behavior"

TUI (Terminal UI)

uv pip install -e ".[tui]"
featcat ui

Keybindings: D Dashboard | F Features | M Monitor | C Chat | Q Quit | ? Help

System Health Check

featcat doctor

[x] Python 3.10+
[x] SQLite catalog exists (catalog.db)
[x] Ollama running at localhost:11434
[x] Model qwen2.5:7b available
[x] 14 features registered
[x] 10 features have docs (71.4%)
[ ] 2 features have drift warnings

Tech Stack

• Python 3.10+ | SQLite (metadata only, never copies data)
• Typer + Rich (CLI) | Textual (TUI)
• PyArrow (Parquet schema + stats) | s3fs (S3/MinIO)
• Ollama (local LLM) | Pydantic (models + config)

Project Structure

featcat/
├── catalog/        # Models, DB, scanner, storage backends
├── llm/            # LLM abstraction (Ollama, llama.cpp)
├── plugins/        # Discovery, Autodoc, Monitoring, NL Query
├── utils/          # Prompts, catalog context, statistics, cache
├── tui/            # Textual TUI (screens, widgets)
├── config.py       # Pydantic settings
└── cli.py          # Typer CLI entry point

License

MIT