VoiceInput 0.1.0

VoiceInput 0.1.0 is the first public release. A normal npm install gets 0.1.0 across all six packages: provider, core, react, openai, elevenlabs, and deepgram.

npm install @voiceinput/react @voiceinput/openai

This release brings the package API and documentation together for the 0.1 launch, with streaming dictation, cursor-aware editing, accessible React controls, and OpenAI, ElevenLabs, and Deepgram adapters. See the documentation.

Package APIs

• Shared server handler types are VoiceTokenAuthorization, VoiceTokenRateLimitResult, and VoiceTokenHandlerContext from @voiceinput/provider.
• Server handlers accept providerTokenUrl to override the provider endpoint.
• React's getTriggerProps() supplies accessible trigger bindings.
• Token-issued callback metadata uses expiresAt in epoch milliseconds for OpenAI and Deepgram.

Validation and artifacts

Published from 5a5ba22, using the exact tarballs from the successful main CI run. Checks cover packed React 18/19 consumers, Node 22/24/26, 92 end-to-end browser checks, automated accessibility, and secret scanning. Live provider checks passed for all three providers, including two Chromium recordings each.

The attached manifest contains all six SHA-512 hashes. The candidate ZIP is the original CI artifact. The release record retains the complete generated notes, registry verification, and observed failures with their successful rechecks. Physical mobile, Safari microphone, and manual screen-reader coverage remain unverified; see the support policy.