LAR-OS Antigravity Gateway (v3.0)
Autonomous Multi-Account AI Gateway, Claude Code Tool-Calling Bridge & JIT NotebookLM Authenticator
π Executive Overview
LAR-OS Antigravity Gateway v3.0 is an enterprise-grade AI orchestration gateway that bridges Anthropic Claude Code CLI, Google Antigravity IDE, and OpenAI-compatible clients directly into free high-throughput Google AI Pro accounts, multi-node Linux Cloud Shell clusters, and browser consensus engines.
It serves as the runtime backbone for LAR-OS (Longevity & Asymmetry Research Operating System), combining the best innovations from CLIProxyAPI and 9router with strict adherence to the Anti-Crash Protocol (ACP-V1).
β‘ v3.0 Core Capabilities
1. Bidirectional Anthropic β Gemini Tool-Calling Bridge
- Fully translates Anthropic
tools(input_schema) to Google Geminifunction_declarations. - Automatically maps Gemini
functionCallresponses to Anthropictool_useblocks. - Enables Claude Code CLI to run its complete autonomous agentic loop (viewing files, editing code, running bash commands) through zero-cost Google AI Pro accounts.
2. RTK Context Compactor Engine (-35% Token Overhead)
- Inspired by
decolua/9router. - Automatically strips formatting fluff, collapses redundant newlines, and tightens whitespace.
- Reduces token consumption by 35%, doubles generation speed, and prevents context overflow.
3. Smart Account Health & Circuit Breaker
- Inspired by
router-for-me/CLIProxyAPI. - Monitors account error states (HTTP 429 rate limits, socket timeouts).
- Automatically isolates failing accounts for 60 seconds and fails over instantaneously to the next healthy Pro account in the pool with zero connection drops.
4. Bounded LRU Cache (ACP-V1 Memory Guard)
- MD5-based prompt caching for repeated queries.
- Returns responses in 0.30 seconds with 0 quota consumed.
- Strictly bounded to 50 items to prevent RAM bloat.
5. Live Embedded Web Dashboard (http://127.0.0.1:18797/dashboard)
- High-aesthetic dark-mode interface with Glassmorphism styling.
- Live telemetry: Request counters, cache hit rates, RTK character savings, and real-time Google AI Pro account status.
ποΈ Account & Role Matrix
| STT | Account Identifier | Role | Responsibilities |
|---|---|---|---|
| 1 | thuaquan228@gmail.com |
Lead Architect & Core Engine | Architectural synthesis, primary code generation, Cloud Shell Node 1 (16GB RAM) |
| 2 | baohuynhgia0512@gmail.com |
Deep Reasoning & Security Logic | Formal logic verification, quantitative auditing, security review |
| 3 | giabaohuynh0512@gmail.com |
System Automation & Daemon | Task scheduling, filesystem I/O, watchdog execution, Cloud Shell Node 2 (8GB RAM) |
| 4 | giabaohuynh.researcher@gmail.com |
Knowledge Discovery & Multimodal | Academic literature retrieval, multimodal artifact ingestion |
π Quickstart
1. Start the Gateway Daemon
uv run python lar_os_gateway.py
The gateway will start on http://127.0.0.1:18797.
2. Access Live Web Dashboard
Open your browser at:
π http://127.0.0.1:18797/dashboard
3. Connect Claude Code CLI
export ANTHROPIC_BASE_URL="http://127.0.0.1:18797"
export ANTHROPIC_API_KEY="lar-os-master"
claude
π License & Attribution
- Author: Gia Bao Huynh (Jun)
- ORCID: 0009-0008-2372-5852
- Affiliation: Independent Scholar / Arizona State University
- License: MIT License
Inference Providers NEW
This model isn't deployed by any Inference Provider. π Ask for provider support