Antigravity MRL v2.0 (Beast Mode Edition)

The "cURL for Memory" - Lightweight, Serverless Matryoshka Indexing.

Release v2.0 Features:

• 🦁 Beast Mode (--beast): Disables safety brakes for high-bandwidth ingestion (50x batch size).
• 🧠 GAG Protocol: Investigates index using the "Chief Operations Architect" persona (No Fluff).
• 🌲 Evergreen Models: Enforces gemini-flash-latest for maximum reasoning.
• 🔪 MRL Slicing: Compresses knowledge by 6x (768d -> 128d) for ultra-lightweight portability.

Usage

1. Beast Mode (Ingest)

Index a massive codebase in seconds:

python python/indexer.py --path ./my_huge_repo --beast

2. The Inquisition (Ask)

Query your codebase with the GAG v2.0 Persona:

python python/indexer.py --ask "Find the memory leak in the buffer logic"

3. File Size Hack (MRL Slicing)

Reduce index size by 83% for mobile/embedded use:

python python/indexer.py --dim 128

Configuration

MRL follows the Antigravity "Evergreen" Mandate. It defaults to models/gemini-flash-latest unless overridden by GEMINI_MODEL env var.

Environment Variables:

• GOOGLE_API_KEY: Required.
• GEMINI_MODEL: Optional override (e.g. models/gemini-1.5-pro).