# vaos toolkits

Modular vision runtime ecosystem

VAOS is not a monolithic SDK — it's a layered set of toolkits. Pick what you need: build, deploy, accelerate, scale. Each toolkit ships with its own CLI, docs, and runtime surfaces.

                    Applications
                          ↑
                      App Toolkit
                          ↑
                      OEM Toolkit
                          ↑
                    SceneLM Toolkit
                          ↑
                      Edge Toolkit
                          ↑
                    Developer Toolkit
                          ↑
              VAOS Core Toolkit

Each toolkit composes upward. Core is the runtime kernel. Apps sit on top.

# the six toolkits

Pick by capability, not by vendor

Runtime Kernel

VAOS Core Toolkit

The foundational runtime layer of VAOS.

  • Scene Contracts
  • Validation runtime
  • MQTT transport
  • Memory interface
  • Runtime APIs
  • Decision engine
Build & Debug

VAOS Developer Toolkit

Build and test vision-native agents locally.

  • Dashboard
  • CLI
  • Simulators
  • Trace viewer
  • Plugin SDK
  • Validators
Hardware Runtime

VAOS Edge Toolkit

Deploy VAOS on edge AI hardware and robotics.

  • RK3588 backend
  • Jetson backend
  • Raspberry Pi
  • NPU runtimes
  • Optimised inference
Structured Cognition

SceneLM Toolkit

Cognition infrastructure for structured scenes.

  • Scene reasoning
  • Memory reasoning
  • Multi-camera cognition
  • Predictive cognition
  • Evaluation suite
Commercial Deployment

VAOS OEM Toolkit

Integrate VAOS into production devices and fleets.

  • Provisioning
  • Fleet management
  • OTA updates
  • Deployment policies
  • Security configs
Ecosystem Layer

VAOS App Toolkit

Build applications on top of the VAOS runtime.

  • Runtime APIs
  • App templates
  • Runtime bindings
  • Automation SDKs
# dependencies

Toolkit relationships

ToolkitDepends on
Core—
DeveloperCore
EdgeCore
SceneLMCore + Edge
OEMEdge + SceneLM
AppsSDK + Runtime
VAOS is a modular vision-native runtime ecosystem built around structured perception, cognition orchestration, validation-aware execution, persistent memory, hardware portability, and event-driven runtime APIs — not a single AI application.
# install

Installable, modular, operational

$ vaos install core
$ vaos install developer
$ vaos install edge-rk3588
$ vaos install scenelm-lite
$ vaos install oem
$ vaos install apps
# roadmap

Ecosystem evolution

VersionToolkit milestone
v1.6Validation runtime stable
v1.7Memory layer stable
v2.0Developer ecosystem GA
v2.1Edge acceleration GA
v2.3SceneLM cognition GA
v2.5OEM deployment GA
v3.0App ecosystem GA

A programmable runtime platform.

Not another AI camera demo. A real runtime ecosystem.

Try the Developer Toolkit→ Browse on GitHub