VEE Audio User Guide

VEE Audio is an application framework for building voice-enabled and audio-rich applications on embedded devices. It hosts multiple audio applications on a single device, each isolated as a sandboxed MicroEJ Feature, under an Audio Kernel that manages hardware resources, audio focus arbitration, and connectivity.

VEE Audio targets audio consumer products: headsets, earbuds, smart glasses, smart speakers, sound bars. It runs on low-power audio SoCs such as the Qualcomm Snapdragon S7 platform.

VEE Audio is built on the MicroEJ Multi-Sandbox platform, enabling independent audio applications to coexist, cooperate, and be updated at runtime, without reflashing the device.

VEE Audio simulator window

Feature Overview

VEE Audio sits between the Qualcomm S7 platform and a set of sandboxed applications. On one side, a phone streams music over Bluetooth (A2DP) or takes voice calls (HFP), and cloud AI backends power the Voice Assistants over Wi-Fi. On the other side, the applications (Voice Assistants, web radio, utilities) run as isolated MicroEJ Features. In between, VEE Audio arbitrates access to the audio output, pausing and resuming Bluetooth playback transparently for the applications. The diagram below shows what Qualcomm provides (the S7 platform) and what VEE Audio adds on top: a managed multi-sandbox runtime and the audio framework. Features marked as planned are on the roadmap and are not part of the current package.

Phone

Media playbackmusic, podcasts, A2DP Callsvoice calls, HFP

Cloud & Backend Services

AI voice backendsOpenAI · Azure · Alibaba Web servicesradio streams · web search Developer PCremote monitor · audio capture

Bluetooth: A2DP / HFP / AVRCP

Wi‑Fi: WebSocket / HTTPS / TLS

Device running VEE Audio

Sandboxed Applications Voice assistants, streaming & utility apps, developed by OEMs or 3rd parties

runs on

VEE Audio provided by MicroEJ

Audio Arbitration

Who plays on the one audio output

▾

Audio Focus

Transparent arbitration of the one audio output

Audio Player

One call plays a sound, focus handled

Volume Control

System volume of the one output

Application Framework

Voice assistants & audio apps, orchestrated

▾

Voice Assistants

Pluggable AI backends: OpenAI, Azure, Alibaba

Voice Functions

Apps expose capabilities to the assistant

AI Query

Apps query the assistant's AI directly

Application Orchestration

One app live: assistant & apps replace each other

User Interface

Voice, buttons & LEDs become intents

▾

Voice Commands

The primary channel: ask for anything

Button Intents

Double press: assistant · single: play/pause

LED Feedback

Device state signaled through a LED service

Wear Detectionplanned

Auto-pause on removal, resume on wear

Configuration & Storage

Settings & keys that survive reboots

▾

Persistent Store

Configuration & keys that survive reboots

GAIA Configurationplanned

Setup from a phone app (Qualcomm GAIA)

Core Runtime

Managed Java & C, native bridge, events

▾

Multi-Language VEE

Managed Java & Managed C on one engine

SNI

Java-to-C bridge to the platform

Event Queue

Async native events into managed Java

Application Lifecycle

Sandboxed apps: lifecycle, isolation, interfaces

▾

Multi-Sandbox Framework

Isolated apps, install at build or runtime

Sandbox Isolation

Third-party code cannot harm the device

Shared Interfaces

Apps reach the kernel via pure interfaces

App Connectplanned

Install & manage apps on a deployed device

Audio I/O

Audio streams, formats & decoders

▾

Audio Streams

Play & record through simple streams

Formats & Decoders

MP3, PCM, A-law; pluggable decoders

Protected Playbackplanned

Widevine DRM for premium streaming

Networking

Wi‑Fi, TLS, HTTP & WebSocket

▾

Wi-Fi

IP transport to the cloud backends

TLS / SSL

Secure channels for AI & streaming

HTTP & WebSocket

REST calls & realtime voice streams

Discovery & Network State

mDNS publication & Wi‑Fi state for apps

integrates with: MicroEJ audio source · SNI natives & events

Qualcomm ADK provided by Qualcomm

Hardware Snapdragon S7 Gen 1 sound SoC

VEE Audio Framework

The VEE Audio Framework describes the architecture and core concepts of VEE Audio:

• the VEE Audio Kernel: lifecycle management, audio focus arbitration, and dynamic Feature deployment,

• the VEE Audio Service Library: ApplicationService, AudioService, ConnectivityService, PersistentStoreService, LedService, EventService, MdnsService, and the Audio Focus concept,

• Audio Applications: the application model, with regular audio applications and the Voice Assistant (AI-backend agnostic design, Function Calling, AI Query API, and secure WebSocket communication with AI backends).

Reference Applications

The Reference Applications chapter presents the ready-to-use applications shipped with VEE Audio:

• three reference Voice Assistants: ChatGPT, Voice Live, and Qwen,

• five audio applications: Web Radio, Voice Note, Web Search, Volume, and Alarm. The all-in-one composite deployment for evaluation and prototyping.

Note

For more information about VEE Audio, please contact your MicroEJ sales representative.