Skip to content

README.md

Latest commit

 

History

History
64 lines (47 loc) · 1.78 KB

README.md

File metadata and controls

64 lines (47 loc) · 1.78 KB
Raw

AutoSorter

Automated Background Document Classification System

A headless Windows background service that monitors your Downloads folder, classifies files by academic subject using sentence embeddings, and auto-sorts them into structured Desktop folders.

How It Works

  1. Monitors Downloads/ for new files via watchdog
  2. Extracts text from PDFs, DOCX, PPTX, images (OCR), and code files
  3. Classifies using all-MiniLM-L6-v2 sentence embeddings + cosine similarity
  4. Moves high-confidence files to Desktop/Subjects/<Category>/
  5. Keeps low-confidence files in Downloads untouched

Supported File Types

Type Extensions
Documents .pdf, .docx, .pptx
Images .jpg, .jpeg, .png
Code .py, .ipynb, .c, .lex

Quick Start

# Install dependencies
pip install -r requirements.txt

# Run directly
python -m src.main

# Or install as startup service
install.bat

Configuration

config/categories.json

Edit this file to change subject categories. No retraining needed:

{
  "RL": "Reinforcement Learning, Q-learning, policy gradient, MDP",
  "DL": "Deep Learning, neural networks, CNN, RNN, transformers"
}

config/config.json

Runtime settings:

  • confidence_threshold: Minimum similarity score to move a file (default: 0.50)
  • max_file_size_mb: Skip files larger than this (default: 100)
  • worker_threads: Concurrent processing threads (default: 2)
  • model_name: Sentence transformer model (default: all-MiniLM-L6-v2)

Requirements

  • Python 3.8+
  • Tesseract OCR (for scanned PDFs and images)
  • ~500MB RAM at runtime

Logs

Located at %LOCALAPPDATA%\AutoSorter\logs\autosorter.log

  • Rotating: 5MB × 5 files
  • Every processed file is logged with category, score, and action