LaRuche

LaRuche - Networked Edge AI System

“Plug in AI. That’s it.”

Plug a LaRuche node into your local network and AI becomes available to connected devices. Zero configuration, zero cloud dependency, and privacy-first by design.

Architecture

Local network

  LaRuche Core (LLM/RAG) <---- LAND protocol (mDNS + HTTP) ----> LaRuche Pro (VLM/Code)
            |                                                           |
            +---------------------- Swarm intelligence ------------------+

Clients
  - VS Code extension
  - Web UI
  - CLI / SDK
  - IoT integrations

Workspace structure

laruche/
|-- land-protocol/        # Core LAND protocol library
|   `-- src/
|       |-- lib.rs
|       |-- capabilities.rs
|       |-- manifest.rs
|       |-- discovery.rs
|       |-- auth.rs
|       |-- qos.rs
|       |-- swarm.rs
|       `-- error.rs
|
|-- laruche-node/         # Node daemon
|   `-- src/main.rs
|
|-- laruche-client/       # Rust client SDK
|   `-- src/lib.rs
|
|-- laruche-cli/          # CLI tool
|   `-- src/main.rs
|
`-- laruche-dashboard/    # Web dashboard
    `-- src/
        |-- main.rs
        `-- templates/dashboard.html

Quick start

Prerequisites

• Rust 1.75+
• Ollama running locally or on your LAN

Windows notes:

• Option 1 (recommended): MSVC toolchain
  • Install Visual Studio Build Tools with “Desktop development with C++”
  • rustup default stable-x86_64-pc-windows-msvc
• Option 2: GNU/MSYS2
  • Install MSYS2 and MinGW binutils
  • Add C:\msys64\mingw64\bin to PATH
  • rustup default stable-x86_64-pc-windows-gnu

1. Pull a model

ollama pull mistral

2. Build

cargo check --workspace

3. Run a node

cargo run -p laruche-node

4. Open the dashboard

http://localhost:8419/dashboard

Current implementation notes

• LAND mDNS re-announce is periodic (2s) to keep nodes visible.
• LAND listener stale timeout is 45s (from land-protocol).
• /swarm includes node port and merged capabilities (HTTP + mDNS fallback).
• /health returns plain text OK.
• Dashboard keeps transiently missing nodes for 10 polls before removal.
• VS Code extension uses:
  • mDNS node-loss grace: 12s
  • swarm stale grace: 6 polls

Node configuration

The node loads configuration in this order:

1. laruche.toml (or LARUCHE_CONFIG=<path>)
2. Environment variables (override file values)

Environment variables

• LARUCHE_NAME (default random laruche-xxxxxx)
• LARUCHE_TIER (nano|core|pro|max)
• LARUCHE_PORT (default 8419)
• LARUCHE_DASH_PORT (default 8420)
• OLLAMA_URL (default http://127.0.0.1:11434)
• LARUCHE_MODEL (default model name)
• LARUCHE_CAP (primary capability, example: llm)
• LARUCHE_CAP2 + LARUCHE_MODEL2 (optional second capability/model)

Example laruche.toml

node_name = "laruche-salon"
tier = "core"
ollama_url = "http://127.0.0.1:11434"
default_model = "mistral"
api_port = 8419
dashboard_port = 8420

[[capabilities]]
capability = "llm"
model_name = "mistral"
model_size = "7B"
quantization = "Q4_K_M"

[[capabilities]]
capability = "code"
model_name = "deepseek-coder"

API

Core endpoints

• GET / - Node status (CPU/RAM, queue, capabilities)
• GET /health - Health check (OK)
• GET /nodes - Discovered peers (mDNS view)
• GET /swarm - Collective view (self + peers, with port)
• GET /models - Local Ollama models
• GET /swarm/models - Models across swarm
• POST /infer - Inference request
• GET /activity - Recent node activity log
• POST /auth/request - Auth request (POC)
• POST /auth/approve - Approve auth request (POC)
• GET /dashboard - Embedded dashboard

Example inference request

{
  "prompt": "Explain ownership in Rust",
  "capability": "code",
  "model": "deepseek-coder",
  "qos": "normal",
  "max_tokens": 1024,
  "temperature": 0.7
}

VS Code extension

cd laruche-vscode
npm install
npm run compile

Then launch the Extension Development Host with F5 from VS Code.

Rust client SDK

use laruche_client::{Cap, LaRuche};

#[tokio::main]
async fn main() {
    let client = LaRuche::discover().await.unwrap();
    let answer = client.ask_with("Write a Rust iterator", Cap::Code).await.unwrap();
    println!("{}", answer.text);
}

Notes

• This workspace currently depends on land-protocol from GitHub (workspace.dependencies).
• If you iterate locally on protocol changes, update dependency resolution before release.

This site is open source. Improve this page.