WhisloAI

Desktop AI assistant for fast translation and dictation from any app.

Download

Download the latest prebuilt app for macOS:

• https://github.com/Canopix/WhisloAI/releases

Overview

WhisloAI helps you translate and dictate text from anywhere in your desktop workflow.

Why I Built This

I built WhisloAI because I want to keep improving my English every day.

I used to do the same loop all the time: write something, translate it, tweak the tone, paste it back, and repeat. After a while, I thought: why not automate this?

So this app started as a personal tool for my daily workflow. I am sharing it publicly because I want honest feedback, I want to see if it is useful to others, and I want to keep learning by building in public with the community.

Current MVP focuses on:

• Translate: source language to target language (configurable in Settings)
• Dictate: record voice, transcribe, translate, and insert back
• Bring-your-own provider credentials (OpenAI and OpenAI-compatible endpoints)
• Quick overlay actions near focused text inputs

Key Features

• Lightweight desktop app built with Tauri
• macOS floating anchor + quick popover workflow
• One-click selected text translation (Translate selection)
• Provider management (add/update/select active provider)
• Connection test against /models
• Voice dictation path (Record -> transcribe -> translate -> insert)
• Auto-insert at cursor with manual paste fallback
• First-run onboarding for required permissions
• Local API key persistence for configured providers

Tech Stack

• Tauri (Rust backend)
• HTML/CSS/JavaScript frontend
• Local settings + provider configuration persisted on disk

Requirements

• Node.js 20+
• Rust toolchain (stable)
• Tauri system dependencies for your OS
• cmake (required for local Whisper transcription)
  • macOS: brew install cmake

Quick Start

npm install
npm run dev

Build

npm run build
npm run tauri build

Local macOS Install and Signing (Contributors)

For day-to-day local installs on macOS, use:

bash scripts/build-install-local-macos.sh

This script:

1. Builds the frontend and Tauri app bundle
2. Installs the app to /Applications/WhisloAI.app
3. Re-signs the installed app
4. Opens the installed app

If you see:

A public key has been found, but no private key. Make sure to set TAURI_SIGNING_PRIVATE_KEY

that error is about updater artifact signing, not local app execution. The script can still continue with local install.

To avoid macOS Accessibility/Automation permission desync across local rebuilds, do not rely on ad-hoc signing.

1. Install a valid Developer ID Application certificate plus private key in your login keychain
2. Confirm identity is available:

security find-identity -v -p codesigning

3. Export your identity so local scripts use stable signing:

export WHISLOAI_CODESIGN_IDENTITY="Developer ID Application: Your Name (TEAMID)"

4. Verify installed app signature is not ad-hoc:

codesign -dvv /Applications/WhisloAI.app 2>&1 | rg "Signature=|Authority=|TeamIdentifier="

Expected: Authority=Developer ID Application... and TeamIdentifier=<your team id>.

If permissions still look enabled in System Settings but checks fail inside the app, use a full local reset first:

npm run reset:local:blank -- --yes

This command:

1. Closes WhisloAI
2. Removes /Applications/WhisloAI.app
3. Cleans local app data in ~/Library for com.whisloai.desktop
4. Resets TCC entries for Accessibility, Automation (AppleEvents), and Microphone

If you also want to remove keychain credentials:

WHISLOAI_RESET_KEYCHAIN=1 npm run reset:local:blank -- --yes

Manual fallback (TCC only):

tccutil reset Accessibility com.whisloai.desktop
tccutil reset AppleEvents com.whisloai.desktop

Then re-enable WhisloAI in:

• Privacy & Security > Accessibility
• Privacy & Security > Automation (System Events)

Publish a Release (Maintainers)

1. Update version values:
   • package.json
   • src-tauri/Cargo.toml
   • src-tauri/tauri.conf.json
2. Commit and push your changes.
3. Create and push a release tag:

git tag v0.1.1
git push origin v0.1.1

4. Configure updater signing secrets in GitHub repository settings:
   • TAURI_SIGNING_PRIVATE_KEY (content of your updater private key)
   • TAURI_SIGNING_PRIVATE_KEY_PASSWORD (optional, only if your key has a password)
5. GitHub Actions workflow .github/workflows/release.yml builds artifacts for macOS (Apple Silicon + Intel), signs updater assets, and attaches them to a GitHub Release.
6. Note: updater signing (TAURI_SIGNING_PRIVATE_KEY) is different from macOS app codesigning (APPLE_CERTIFICATE, APPLE_SIGNING_IDENTITY, APPLE_TEAM_ID).

Usage

Core flow

1. Open a text field in any app
2. Open the WhisloAI widget
3. Choose Translate selection (or use dictation)
4. Insert output back into your active text field

Widget workflow

• Keep the widget open while writing
• Use quick actions to translate selected text or dictate
• Configure provider, languages, and writing modes from Settings

Privacy and Security

• API keys are currently persisted locally in app configuration (base64-encoded, not encrypted) to keep setup simple during MVP
• You control provider selection and credentials
• Recommended before public deployment: add explicit telemetry and data-handling documentation

Project Status

WhisloAI is an active MVP and moving toward a broader open source release.

License

This project is licensed under the PolyForm Noncommercial 1.0.0 license.

• Noncommercial use is allowed under the terms in LICENSE.
• Commercial use requires a separate commercial license from the maintainer.

Contributing

Contributions are welcome.

Suggested flow:

1. Open an issue with context and expected behavior
2. Submit a focused PR with clear testing notes
3. Keep changes scoped and avoid mixing refactors with feature work

Maintainer

• Built by @emanuel_build