Callsign

Callsign is an open-source, Windows-first desktop voice assistant for visible, identity-gated computer control.

The project starts with a simple interaction contract:

Callsign -> identity verification -> command -> visible action

The Free core is the public product: inspectable, local-first, and designed so users can always see when Callsign is listening, what it heard, and how to stop it.

Current alpha

The current repo is focused on the Alpha v1 line. v1.0 is the first public MVP:

• create a local profile and callsign
• record and review voice enrollment samples
• run a background service for wake/session state
• detect the Callsign wake cue through the audio path
• show callsign.gif with a live readout
• verify the user's callsign/voice before command capture
• launch installed apps through a visible Start menu flow
• expose stop, cancel, timeout, lockout, and runtime status states

The codebase also contains early surfaces for the broader Alpha v1 direction, including browser launch helpers, system control helpers, file search, visible-control routing, richer command parsing, and extension packs. Those are part of the v1.x path toward practical Windows Voice Access parity, not a reason to weaken the v1.0 release gate.

The update server now lives in the separate local repo at update-server-repo/, which owns the Dockerized server, manifests, deployment docs, and the WinForms Update Manager used to publish and deploy releases.

Open-source promise

Callsign is Free-first.

• The Free tier is MIT-licensed and useful on its own.
• The public repo contains the setup app, service runtime, visible overlay, profile/enrollment flow, command-routing contracts, docs, and tests for the open core.
• Stable Windows 11 Voice Access parity belongs in the Free core and is tracked in docs/reference/VOICE_ACCESS_PARITY_MATRIX.md.
• Alpha v1 features remain free until at least beta.
• The Free core must not depend on private code or paid entitlement.

This project should be understandable from source. If Callsign hears you, verifies you, or acts for you, the public core should make that behavior inspectable.

Closed-source future

Callsign is also being designed for a future commercial layer.

Future Pro and Advanced features may ship as closed-source extension libraries: deeper Windows, WSL, Linux, browser, workflow, diagnostics, and specialized command catalogs that can evolve faster than the open core.

That boundary is intentional:

• Free stays the public trust layer.
• Pro and Advanced can expand the ceiling.
• Proprietary tier material belongs only in /closed-source/, which is ignored by git.
• Paid extensions must still pass the same identity, visibility, policy, and audit expectations.

Alpha v1 release ladder

The repository tracks two release numbering lanes:

• Major milestones: 0.0.3a, 0.0.4a, 0.0.5a, 0.1.0a, and 1.0.0a.
• Micro patches: 0.0.01a, 0.0.02a, etc., for focused stabilization between majors.

Release	Scope
v1.0 alpha	Background service wake detection, callsign identity verification, callsign.gif wake overlay with live text readout, and visible Start menu app launch.
v1.1 alpha	Dictation with visible review before insertion, copy, paste, or other text actions.
v1.2 alpha	Browser control for visible open, search, navigation, and safe bounded browser tasks.
v1.3 alpha	System control for Windows, WSL, and Linux, including file search results shown or opened through Explorer.
v1.4 alpha	Parity hardening: close the Voice Access matrix, polish overlays, expand tests, replay imported-pack and update splashes, and produce a release-candidate installer/site update.
Beta or later	Revisit Pro and Advanced packaging, entitlement, signed extension libraries, and continuously updated command catalogs.

Product structure

• src/Callsign.UI is setup, onboarding, monitoring, profile management, voice enrollment, overlay/readout UI, and user-visible controls.
• src/Callsign.Service is the background runtime for wake, identity, session orchestration, runtime status, and visible launch flow.
• tests/Callsign.AlphaSmoke contains the current alpha smoke coverage.
• docs/reference contains the canonical markdown used to generate the public website.
• docs/reference/VOICE_ACCESS_PARITY_MATRIX.md tracks the 100% Windows Voice Access parity target.
• docs/index.html is the generated public landing page.
• CANON.md is the root product canon, mirrored in docs/reference/CANON.md.

Safety model

Callsign is not a hidden desktop automation framework.

• Wake audio alone cannot launch commands.
• Transcript text alone must not wake or authorize a session.
• Identity must pass before command capture.
• v1.0 actions stay visible and local.
• Arbitrary shell execution is out of scope.
• Passwords, 2FA codes, payments, account deletion, security settings, and silent external submissions are blocked.
• Screenshots, clipboard contents, file contents, and UI trees are sensitive and must not be sent to cloud models unless the user explicitly opts in.

Build and verify

From a fresh checkout:

.\buildcallsign.ps1
dotnet run --project tests/Callsign.AlphaSmoke/Callsign.AlphaSmoke.csproj
python scripts/build_site.py
.\scripts\prepare-release-packet.ps1
.\scripts\voice_access_parity_evidence.ps1 -WriteManualEvidenceTemplate
.\scripts\verify-release-readiness.ps1 -WebsiteDownloadUrl "http://your-host/downloads/Callsign-Setup.exe" -RequireWebsiteVerification

Preview the public site locally:

python -m http.server 8000 --directory docs

Then open http://localhost:8000.