360router

Smart AI model router — local first, cloud when needed.

What is 360router?

360router intelligently routes AI requests between local models (Ollama, LM Studio, vLLM) and cloud providers (Anthropic, OpenAI, Groq, Gemini, Grok). It automatically selects the best provider based on:

• Request complexity
• Provider availability
• Circuit breaker state
• Your configured preferences

Key features:

• Local-first routing (privacy by default)
• Automatic fallback to cloud when local unavailable
• Circuit breaker prevents repeated calls to failing providers
• Basic PII detection and privacy counter
• Opt-in telemetry for performance insights
• Zero configuration for common local setups

Installation

npm install -g 360router

Quick Start

1. Setup

360router init

This will:

• Scan for local providers (Ollama, LM Studio, vLLM, Jan)
• Optionally add remote local providers
• Optionally add cloud providers (Anthropic, OpenAI, etc.)

2. Check status

360router status

Shows provider health, latency, model counts, and privacy stats.

3. Route a message

360router route "What is 2+2?"

Usage as Library

import { route } from '360router';

const result = await route([
  { role: 'user', content: 'Hello!' }
]);

console.log(result.content);
console.log(`Provider: ${result.provider}, Model: ${result.model}`);

Configuration

Configuration is stored in ~/.config/360router/config.json (cross-platform).

Provider Configuration

interface ProviderConfig {
  name: string;
  kind: 'local' | 'cloud';
  enabled: boolean;
  baseUrl?: string;      // For local providers
  apiKey?: string;       // For cloud providers
  label?: string;        // Display name
}

Routing Logic

360router uses a 5-step routing process:

1. Complexity — Classify request (simple/medium/complex/expert)
2. Sensitivity — Check for PII (cloud-bound only)
3. Tier — Select appropriate model tier
4. Queue — Try local providers first, then cloud
5. Fallback — Circuit breaker prevents repeated failures

Supported Providers

Local (OpenAI-compatible)

• Ollama (port 11434)
• LM Studio (port 1234)
• vLLM (port 8000)
• Jan (port 1337)
• Custom endpoints

Cloud

• Anthropic (Claude)
• OpenAI (GPT)
• Groq
• Google Gemini
• xAI Grok

Privacy

• Local requests never leave your machine
• Cloud requests are scanned for basic PII patterns (SSN, email, phone, credit card)
• Privacy counter tracks cloud-bound requests with potential sensitive data
• Telemetry is opt-in and never collects prompts, responses, or API keys

Telemetry

If enabled, 360router collects:

• Provider latency
• Success/failure rates
• Error codes
• OS and version

Never collected:

• Prompt content
• Response content
• API keys
• IP addresses
• Personal identifiers

Managed API Keys

Don't want to manage API keys? Get managed cloud access at 360ops.ai from $3.99/mo.

License

MIT

Support

• Documentation: https://360ops.ai/router
• Issues: https://github.com/360ops/360router/issues
• Email: [email protected]