v1.0.0

Beta release. OpenStream runs from source; this DMG is an unsigned convenience build.

---

Local-first voice dictation for macOS. Hold a key, talk, let go, and the words land at your cursor. Nothing you say leaves the machine.

Before it types a line break, OpenStream checks which app you are in. A spoken "new paragraph" becomes a real newline in a document, but in a terminal or a chat box, where a stray newline would submit something half-finished, the words just carry on.

• Transcription runs Parakeet (NVIDIA) on the Apple Neural Engine, via FluidAudio.
• A deterministic rules engine tidies the text. A small local model places paragraph breaks and never rewrites your words.
• No account, no cloud, no telemetry.

Requirements: Apple Silicon, macOS 14 or newer.

Install: the supported path is git clone + npm install + npm start — see the README. The attached DMG is unsigned, so Gatekeeper will warn on first open (right-click the app → Open, or allow it in System Settings → Privacy & Security).

Known rough edges (tracked on the v1.1 milestone): transcription can mishear proper nouns on short clips; spoken numbers come through spelled out; the first dictation right after launch is slower while the accessibility API warms up.