Skip to content

Releases: glebis/cenno

cenno v0.3.1

Choose a tag to compare

@glebis glebis released this 23 Jun 07:44

Key changes:

  • 🔊 Sound-out / voice-out — prompts can be read aloud, with an on-device neural voice (Supertonic)
  • ⚙️ In-app settings window — configure voice, behavior, and defaults from the tray
  • 📱 Cross-device routing — an iOS / watchOS companion surfaces prompts on your phone and Watch
  • 🧩 dismiss_pending MCP tool — for agent-driven voice loops
  • 🛟 Launch fix — 0.3.0 wouldn't open on some Macs; 0.3.1 fixes it and hardens the release pipeline so it can't recur

Fixed (0.3.1)

  • 0.3.0 would not launch ("The application "cenno" can't be opened"). The
    0.3.0 build added restricted iCloud/CloudKit entitlements for the relay, but a
    Developer-ID (non-App-Store) bundle ships no embedded.provisionprofile to
    authorize them, so macOS AMFI killed the app at spawn (exit 137). The iCloud
    entitlements are removed; the relay is already gated to run without them, so no
    functionality is lost on the Mac build. The release pipeline now launch-tests
    the built app and rejects restricted entitlements lacking a provisioning
    profile, so this class of brick cannot ship again.

What landed in 0.3.0

Added

  • Sound-out (voice-out): cenno can read a prompt aloud when it appears.
    Opt-in, urgency-gated, with an optional short say summary. Two engines —
    the fast macOS system voice and Supertonic, a fully on-device neural
    voice (10 styles, F1–F5 / M1–M5) that downloads and runs locally.
  • In-app settings window (tray → "cenno settings…"): a tabbed
    Settings / Integration / About window styled after the website. Configure
    voice (engine, voice, urgency, audio output device, model download /
    delete), behavior (launch at login, hide from Dock), and defaults — written
    to ~/.cenno and applied without a restart.
  • dismiss_pending MCP tool: unparks pending prompts and hides the panel
    immediately, for agent-driven voice loops that capture the answer via an
    external speech-to-text.
  • Cross-device prompt routing (second screens) and an iOS / watchOS
    companion app that surfaces prompts on the phone and Watch.

Fixed

  • CloudKit relay no longer crashes the app on builds without the iCloud
    entitlement: the relay is gated on the entitlement being present instead of
    trapping inside CloudKit (uncatchable from Swift).

cenno v0.2.0

Choose a tag to compare

@glebis glebis released this 11 Jun 08:14

Added

  • Slider catalog control (standard SliderApi + minLabel/maxLabel/ selectAction): continuous "how much" answers, committing on thumb release or Enter.
  • DateTimeInput catalog control (standard API + submitAction): native date/time pickers for "when?" answers. - Image catalog control (standard ImageApi): display-only visual context in prompts.
  • In-app updates from GitHub releases via the Tauri updater, with Check for updates… in the tray menu (signature-verified; installs and restarts only after confirmation).
  • docs/design/CONTROLS.md: the catalog inventory and the recipe for adding controls.
  • Repository meta: LICENSE (Apache-2.0), NOTICE, AUTHORSHIP, SECURITY, CONTRIBUTING, CODE_OF_CONDUCT, CHANGELOG.

cenno 0.1.0 — un cenno

Choose a tag to compare

@glebis glebis released this 10 Jun 19:32

First release. cenno is a macOS runtime that lets AI agents ask you questions through minimal floating surfaces — agents fanno un cenno, you answer with one.

Signed & notarized (Developer ID, ticket stapled) — opens cleanly on any Apple Silicon Mac.

What's in 0.1.0

  • ask_user via MCP or CLI — one line in .mcp.json gives any agent a voice on your screen; answers return as tool results ({answer, via, elapsed_s})
  • Non-activating floating panel — never steals keyboard focus; draggable, remembers position, sizes to content; single-click answers
  • Reporter-style design system — full-bleed flow colors (mood / question / EMA / reminder / ambient), W3C DTCG tokens, A2UI rendering with a custom catalog; rich a2ui payloads pass a validated boundary
  • Inputs: text, choice chips, 1–7 scale with dot pagination, confirm, multi-step progress — plus 30 rotating deadpan confirmations
  • SQLite history + cenno export — every prompt outcome recorded locally (0600), JSON/CSV dump with --since
  • Tray controls — pause (15 min → until tomorrow), fullscreen quiet mode (per-screen, notch-aware), launch at login (default on), all persisted
  • --mcp-stdio bridge — auto-launches the app in tray mode; stale-socket recovery

Install

Download the DMG, drag cenno to Applications. Or build from source: see README.

Coming next

EMA scheduling engine (recurring check-ins), voice input via whisper.cpp + BYOK, tray inbox for missed prompts.

🤖 Generated with Claude Code