# sdk

VAOS™ SDK

First-class Python. REST and WebSocket APIs from any language. Plugin spec for extending the runtime.

REST API ref → GitHub ↗ Quickstart →
// PYTHON

vaos (Python)

The reference SDK. Async, typed, idiomatic. pip install vaos

// REST

HTTP / REST

Stable JSON API for any language. OpenAPI spec available.

// WS / MQTT

Streaming

WebSocket and MQTT topics for live scene streams and events.

// JS / TS

vaos-js (beta)

TypeScript SDK for browser dashboards and Node.js integrations.

// GO

vaos-go (alpha)

Go bindings for high-throughput action emitters and integrators.

// PLUGIN

Plugin spec

Extend perception, reasoning, and action layers — typed plugin contracts.

Common patterns

Subscribe to scenes

subscribe.pypython
from vaos import Runtime

runtime = Runtime(camera=0)

async for scene in runtime.scenes():
    print(scene.summary, scene.confidence)

Emit a validated action

action.pypython
await runtime.action(
  "slow_down",
  target="amr-fleet-nav",
  validation=True   # gate on scene confidence (default)
)

Query scene memory

memory.pypython
recent = await runtime.memory.query(
  type="restricted_zone_proximity",
  window="1h"
)
print(f"{len(recent)} similar events in last hour")

Subscribe via WebSocket (any language)

websocketjs
const ws = new WebSocket("ws://localhost:8800/scenes");
ws.addEventListener("message", (e) => {
  const scene = JSON.parse(e.data);
  console.log(scene.summary, scene.metrics.latency_ms);
});
Patent pending. The VAOS™ runtime, sensor-aware perception, and validation-gated action are protected by patents pending. Open-source under MIT for the SDK and runtime; commercial licensing available for derivative hardware/silicon.