rtk-ai · pszymkowiak · Feb 10, 2026 · Feb 10, 2026 · Feb 10, 2026
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
@@ -5,6 +5,7 @@ on:
     branches: [main]
   pull_request:
     branches: [main]
+  workflow_dispatch:
 
 env:
   CARGO_TERM_COLOR: always

diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
@@ -123,3 +123,97 @@ jobs:
           files: release/*
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
+  homebrew:
+    name: Update Homebrew formula
+    needs: [release]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Get version
+        id: version
+        run: |
+          TAG="${{ inputs.tag }}"
+          if [ -z "$TAG" ]; then
+            TAG="${{ github.event.release.tag_name }}"
+          fi
+          VERSION="${TAG#v}"
+          echo "tag=$TAG" >> $GITHUB_OUTPUT
+          echo "version=$VERSION" >> $GITHUB_OUTPUT
+
+      - name: Download checksums
+        run: |
+          gh release download "${{ steps.version.outputs.tag }}" \
+            --repo rtk-ai/vox \
+            --pattern checksums.txt
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Parse checksums
+        id: sha
+        run: |
+          echo "arm=$(grep aarch64-apple-darwin checksums.txt | awk '{print $1}')" >> $GITHUB_OUTPUT
+          echo "intel=$(grep x86_64-apple-darwin checksums.txt | awk '{print $1}')" >> $GITHUB_OUTPUT
+          echo "linux=$(grep x86_64-unknown-linux-gnu checksums.txt | awk '{print $1}')" >> $GITHUB_OUTPUT
+
+      - name: Update formula
+        run: |
+          VERSION="${{ steps.version.outputs.version }}"
+          TAG="${{ steps.version.outputs.tag }}"
+          cat > vox.rb << 'FORMULA'
+          class Vox < Formula
+            desc "Cross-platform TTS CLI — local voice synthesis with three backends"
+            homepage "https://github.com/rtk-ai/vox"
+            version "VERSION_PLACEHOLDER"
+            license "MIT"
+
+            on_macos do
+              on_intel do
+                url "https://github.com/rtk-ai/vox/releases/download/TAG_PLACEHOLDER/vox-x86_64-apple-darwin.tar.gz"
+                sha256 "SHA_INTEL_PLACEHOLDER"
+              end
+              on_arm do
+                url "https://github.com/rtk-ai/vox/releases/download/TAG_PLACEHOLDER/vox-aarch64-apple-darwin.tar.gz"
+                sha256 "SHA_ARM_PLACEHOLDER"
+              end
+            end
+
+            on_linux do
+              on_intel do
+                url "https://github.com/rtk-ai/vox/releases/download/TAG_PLACEHOLDER/vox-x86_64-unknown-linux-gnu.tar.gz"
+                sha256 "SHA_LINUX_PLACEHOLDER"
+              end
+            end
+
+            def install
+              bin.install "vox"
+            end
+
+            test do
+              assert_match "vox #{version}", shell_output("#{bin}/vox --version")
+            end
+          end
+          FORMULA
+          sed -i "s/VERSION_PLACEHOLDER/$VERSION/g" vox.rb
+          sed -i "s/TAG_PLACEHOLDER/$TAG/g" vox.rb
+          sed -i "s/SHA_INTEL_PLACEHOLDER/${{ steps.sha.outputs.intel }}/g" vox.rb
+          sed -i "s/SHA_ARM_PLACEHOLDER/${{ steps.sha.outputs.arm }}/g" vox.rb
+          sed -i "s/SHA_LINUX_PLACEHOLDER/${{ steps.sha.outputs.linux }}/g" vox.rb
+          # Remove leading spaces from heredoc
+          sed -i 's/^          //' vox.rb
+
+      - name: Push to homebrew-tap
+        run: |
+          CONTENT=$(base64 -w 0 vox.rb)
+          SHA=$(gh api repos/rtk-ai/homebrew-tap/contents/Formula/vox.rb --jq '.sha' 2>/dev/null || echo "")
+          if [ -n "$SHA" ]; then
+            gh api -X PUT repos/rtk-ai/homebrew-tap/contents/Formula/vox.rb \
+              -f message="vox ${{ steps.version.outputs.version }}" \
+              -f content="$CONTENT" \
+              -f sha="$SHA"
+          else
+            gh api -X PUT repos/rtk-ai/homebrew-tap/contents/Formula/vox.rb \
+              -f message="vox ${{ steps.version.outputs.version }}" \
+              -f content="$CONTENT"
+          fi
+        env:
+          GH_TOKEN: ${{ secrets.HOMEBREW_TAP_TOKEN }}
diff --git a/Formula/vox.rb b/Formula/vox.rb
diff --git a/README.md b/README.md
@@ -1,94 +1,85 @@
 # vox
 
-TTS CLI for macOS — local voice synthesis with Qwen and system `say`.
+Cross-platform TTS CLI — local voice synthesis with three backends.
 
-## Features
-
-- **Two backends**: macOS native `say` and [Qwen TTS](https://github.com/ml-explore/mlx-audio) (local, Apple Silicon)
-- **Voice cloning**: clone a voice from an audio sample, use it for all speech
-- **Voice chat**: have a spoken conversation with Claude (STT + LLM + TTS)
-- **Pipeline playback**: multi-sentence text plays without gaps between chunks
-- **Preferences**: persist backend, voice, language, rate, style settings
+```
+                         vox
+                          |
+            +-------------+-------------+
+            |             |             |
+          say          qwen        qwen-native
+       (macOS)     (MLX/Python)    (pure Rust)
+        native      Apple Silicon   cross-platform
+                                   CPU/Metal/CUDA
+                          |
+                        rodio
+                    (audio playback)
+```
 
 ## Install
 
 ```bash
-# Quick install (macOS)
-curl -fsSL https://raw.githubusercontent.com/rtk-ai/vox/main/install.sh | sh
-
 # From source
 cargo install --path .
 
-# Or via Homebrew (coming soon)
-brew tap rtk-ai/tap && brew install vox
+# Quick install (macOS / Linux / WSL)
+curl -fsSL https://raw.githubusercontent.com/rtk-ai/vox/main/install.sh | sh
 ```
 
-### Requirements
+| Platform | Default backend | GPU |
+|----------|----------------|-----|
+| macOS | `say` | `--features metal` |
+| Linux / WSL | `qwen-native` | `--features cuda` |
 
-- macOS (uses `say` and `afplay`)
+Linux requires `sudo apt install libasound2-dev`.
 
-For the Qwen backend (local neural TTS, Apple Silicon only):
+## Usage with Claude Code
 
 ```bash
-brew install python3
-pip install mlx-audio
+vox init                # all integrations (default)
+vox init -m mcp         # MCP server only
+vox init -m cli         # CLI hook only
+vox init -m skill       # slash command only
 ```
 
-This pulls in [mlx-audio](https://github.com/ml-explore/mlx-audio) which provides both TTS (`mlx_audio.tts`) and STT (`mlx_audio.stt`). The model `mlx-community/Qwen3-TTS-12Hz-0.6B-Base-bf16` is downloaded automatically on first use (~1.2 GB).
+Each mode sets up a different integration:
 
-For voice chat:
+| Mode | What it does |
+|------|-------------|
+| `mcp` | Registers `vox serve` as an MCP server in `~/.claude.json` (Claude Code) and Claude Desktop config. Exposes 8 tools: `vox_speak`, `vox_list_voices`, `vox_clone_*`, `vox_config_*`, `vox_stats`. |
+| `cli` | Creates a `CLAUDE.md` in your project with instructions for Claude to call `vox` after significant tasks. Adds a `Stop` hook in `.claude/settings.json` that says "Terminé" after each response. |
+| `skill` | Creates a `/speak` slash command in `~/.claude/commands/speak.md`. |
+| `all` | Runs all three modes (default). |
 
-```bash
-brew install sox                  # audio recording (rec command)
-export ANTHROPIC_API_KEY=sk-ant-...
+```
+  Claude Code
+      |
+   MCP stdio
+      |
+  vox serve ──> vox_speak, vox_list_voices, ...
 ```
 
-## Usage
+Running `vox init` again is safe — it skips files that are already configured.
+
+## Standalone CLI
 
 ```bash
-# Speak text (default: say backend)
 vox "Hello, world."
-
-# Use Qwen backend
-vox -b qwen "Bonjour le monde."
-
-# Pipe from stdin
+vox -b qwen-native "Cross-platform TTS."
 echo "Hello" | vox
-
-# List voices
 vox --list-voices
-vox -b qwen --list-voices
-
-# Set voice and language
-vox -b qwen -v Chelsie -l en "Good morning."
 ```
 
 ### Voice cloning
 
 ```bash
-# Add a clone from an audio file
-vox clone add patrick --audio ~/voice.wav --text "Transcription of the audio"
-
-# Record a clone from microphone
+vox clone add patrick --audio ~/voice.wav --text "Transcription"
 vox clone record myvoice --duration 10
-
-# Use a cloned voice
 vox -v patrick "This speaks with your voice."
-
-# List / remove clones
 vox clone list
 vox clone remove patrick
 ```
 
-### Voice chat
-
-```bash
-# Start a voice conversation with Claude
-export ANTHROPIC_API_KEY=sk-ant-...
-vox chat
-vox chat -v patrick -l fr
-```
-
 ### Preferences
 
 ```bash
@@ -99,34 +90,30 @@ vox config set voice Chelsie
 vox config reset
 ```
 
-### Stats
-
-```bash
-vox stats
-```
-
-## AI Integration
+### Optional: Qwen backend (macOS)
 
-Set up your project so that Claude Code provides spoken summaries after completing tasks:
+Neural TTS via Python/MLX on Apple Silicon:
 
 ```bash
-cd your-project
-vox init
+uv pip install mlx-audio
 ```
 
-This creates:
-- **CLAUDE.md** — instructions for Claude to call `vox` after significant tasks
-- **.claude/settings.json** — a `Stop` hook that says "Terminé" after each response
+Model downloaded automatically on first use (~1.2 GB).
 
-Running `vox init` again is safe — it skips files that are already configured.
+## Data
 
-## Configuration
+All state is stored locally in `~/.config/vox/`:
+
+```
+~/.config/vox/
+├── vox.db          # SQLite: preferences, voice clones, usage logs
+└── clones/         # audio files for voice clones
+```
 
 | Env var | Description |
 |---------|-------------|
-| `VOX_CONFIG_DIR` | Override config directory (default: `~/.config/vox/`) |
-| `VOX_DB_PATH` | Override database path (default: `~/.config/vox/vox.db`) |
-| `ANTHROPIC_API_KEY` | Required for `vox chat` |
+| `VOX_CONFIG_DIR` | Override config directory |
+| `VOX_DB_PATH` | Override database path |
 
 ## License