data-ai

Intelligent file organizer using semantic similarity. Define categories with keywords, and let AI sort your files automatically.

Features

• YAML Configuration — Define categories with keywords and optional example documents
• Local AI — Uses Ollama for embeddings (nomic-embed-text) and vision (llava)
• Multiple Modes — Single sort, preview+apply, or continuous watch
• Smart Matching — Hybrid similarity using keywords and example documents
• Interactive — Prompts for uncertain classifications

Installation

# Install with uv
uv tool install data-ai

# Or with pipx
pipx install data-ai

Requirements

• Python 3.11+
• Ollama running locally
• Tesseract for OCR: brew install tesseract (macOS) or apt install tesseract-ocr (Linux)

Pull required models:

ollama pull nomic-embed-text
ollama pull llava  # For image description

Quick Start

# Create config
data-ai init

# Edit ~/.config/data-ai/config.yaml with your categories

# Sort files
data-ai sort ./inbox

# Or preview first
data-ai scan ./inbox
data-ai apply

# Or watch continuously
data-ai watch ./inbox

Configuration

settings:
  ollama_model: "nomic-embed-text"
  vision_model: "llava"
  similarity_threshold: 0.6
  inbox: "./inbox"

categories:
  Invoices/Outgoing:
    keywords:
      - "Invoice"
      - "Rechnung"
      - "Amount due"
    examples:
      - "./examples/sample-invoice.pdf"

  Contracts:
    keywords:
      - "Contract"
      - "Agreement"
      - "Terms"
    examples: []

Commands

Command	Description
data-ai init	Create default config
data-ai sort [DIR]	Sort files immediately
data-ai scan [DIR]	Preview what would be sorted
data-ai apply	Execute last scan
data-ai watch [DIR]	Watch and sort new files
data-ai test FILE	Test single file classification
data-ai config	Show current settings
data-ai categories	List categories and keywords

License

MIT