The Voxide Engine.

A precise, highly technical look at the capabilities and underlying architecture of the Voxide client-side SDK.

Middleware Pipelines

Intercept, validate, or block AI executions before they run. Perfect for verifying user authentication states or rate-limiting heavy API calls.

Idempotent Manifest Sync

On every initialization, the SDK ships a snapshot of your capabilities. Versioning is handled automatically, making deployments zero-risk.

Adaptive Rate Control

Adjust audio streaming sample rates dynamically based on network latency to ensure VAD turns stay snappy even on weak mobile connections.

Double-Guard Security

Domain whitelist CORS protection on the server, paired with explicit user-facing UI confirmation prompts for any mutating database action.

Data flow architecture
1. Client Audio

User speaks. Mic samples at 16kHz. Float32 array is converted to PCM-16 and chunked over active WebSocket.

➔
2. State Injection

Browser automatically appends latest bindState() JSON frame to request context before matching tools.

>>
3. Action Executed

Gemini matches tools, triggers client-side handler in browser context, and returns status output to stream.