OpenScribe v0.1.1

Highlights\n- Fixed false "Anthropic key required" prompt by using server/env truth for mixed mode auth status.\n- Hardened desktop microphone readiness and permission recovery flow.\n- Added preferred mic selection + fallback to default device when stale device IDs are saved.\n- Added macOS audio-input entitlement for packaged app/helper bundles to enable correct mic permission behavior.\n\n## Artifacts\n- macOS arm64 DMG and ZIP\n- Release manifest and checksums

v0.1.0 - Developer Preview

OpenScribe v0.1.0 is a developer preview of a local-first, MIT-licensed AI medical scribe.

What this release includes

• Web app + desktop (Electron) surfaces
• Audio capture and transcription via OpenAI Whisper API
• Structured draft note generation via Anthropic / OpenAI
• Editable note sections (CC, HPI, ROS, PE, Assessment, Plan)
• AES-GCM encrypted local storage
• Baseline test suite for core packages (llm, pipeline)

Quick start

• Node.js 18+ and pnpm
• pnpm install
• pnpm setup
• Add OPENAI_API_KEY and/or ANTHROPIC_API_KEY to apps/web/.env.local
• pnpm dev (web) or pnpm dev:desktop (desktop)

Data handling

• Local-first: encounters, transcripts, notes stored locally (encrypted)
• Audio processed in-memory; not persisted by default
• No analytics or telemetry

Known limitations (expected in v0.1.0)

• Not production-ready; no compliance claims
• Limited audit logging and admin controls
• Browser storage limits apply (~5-10MB typical)
• Error handling and recovery still improving
• No EHR integration

Contributing

• Issues and PRs welcome
• Priority: error handling, tests, documentation, accessibility, provider adapters

Repo
https://github.com/sammargolis/OpenScribe