🐜 Ant v0.4.0

High-ratio lossless compressor + Agent Mode that saves 70–90% tokens when feeding repos to AI coding tools.

GitHub Repository

pip install antcompress[agent]

Commands

• ant init [path] → Generates an .ant_context/ tree replica for efficient IDE exploration + agent rules.
• ant compress <path> → tiny .ant file (95–99% smaller)
• ant decompress <file.ant> → restore exact original
• ant pack-ai <folder> → Agent Mode (JSON for IDEs or MD for phone)
• ant unpack-ai <edited.md> → AI changes → new .ant

Quick Start

ant pack-ai . --stdout --model claude

Phone Workflow (no Termux)

1. ant pack-ai my-repo/
2. Send my-repo.ai.md to phone.
3. Paste into Claude/Grok/ChatGPT.
4. Save AI reply as edited.md.
5. Send back → ant unpack-ai edited.md → ant decompress updated.ant

Agent Mode in IDEs

Cursor / Continue.dev

Add to ~/.continue/config.json:

"context": [{
  "provider": "command",
  "params": {
    "command": "ant pack-ai . --format json --model claude --max-tokens 120000 --stdout",
    "name": "ant-repo"
  }
}]

Type @ant-repo in chat.

Aider

aider --context-cmd "ant pack-ai . --format json --model claude --max-tokens 128000 --stdout"

GitHub Copilot

Run ant pack-ai . --stdout and paste the output into Copilot Chat to give it full project context while staying under the token limit.

Google Antigravity

Since Antigravity can use your terminal, simply ask it to: "Use Ant to pack this repo for context" or "Use Ant to apply changes from this edited.md file."

Made with ❤️ by Okediya Ayobami Oluwaseunfunmi