Skip to content

CLAUDE.md

Latest commit

 

History

History
55 lines (39 loc) · 1.81 KB

CLAUDE.md

File metadata and controls

55 lines (39 loc) · 1.81 KB
Raw

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

A FastAI-based web application for classifying chess piece images. The model can identify six chess pieces: Bishop, King, Knight, Pawn, Queen, and Rook.

Key Commands

Local Development with Docker

docker build -t fastai-v3 . && docker run --rm -it -p 5000:5000 fastai-v3

Run Server Locally

python app/server.py serve

Architecture

Core Components

  • app/server.py: Main FastAI/Starlette server that:

    • Downloads pre-trained model from Google Drive on startup
    • Serves static files and HTML frontend
    • Provides /analyze endpoint for image classification
    • Runs on port from environment variable or defaults to 50000

  • Frontend: Simple vanilla JavaScript interface in:

    • app/view/index.html: UI for image upload
    • app/static/client.js: Handles file selection and API calls
    • app/static/style.css: Styling

Model Management

  • Model (export.pkl) is downloaded from Google Drive URL at runtime
  • Download URL is hardcoded in server.py: export_file_url
  • Model is loaded using FastAI's load_learner()
  • Classes are predefined: ['Bishop', 'King', 'Knight', 'Pawn', 'Queen', 'Rook']

Deployment Configuration

  • Dockerfile: Python 3.7.9 base image with FastAI dependencies
  • Procfile: Heroku deployment configuration
  • requirements.txt: Pins specific PyTorch CPU versions for deployment
  • Port configured via PORT environment variable

Key Considerations

  • The app uses CPU-only PyTorch versions for deployment efficiency
  • Model file is not stored in repo to keep deployment lightweight
  • CORS is enabled for all origins in the Starlette middleware
  • Image processing uses FastAI's open_image() from BytesIO stream