agentvibes 4.6.8 → 5.1.0

package/.claude/hooks-windows/session-start-tts.ps1

@@ -1,114 +1,114 @@
-#
-# File: .claude/hooks-windows/session-start-tts.ps1
-#
-# AgentVibes SessionStart Hook for Windows
-# Outputs JSON with hookSpecificOutput.additionalContext for reliable context injection.
-# Mirrors session-start-tts.sh — keep both in sync.
-#
-
-$ErrorActionPreference = "Stop"
-
-# Get script directory
-$ScriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
-
-# Check if AgentVibes is installed
-if (-not (Test-Path (Join-Path $ScriptDir "play-tts.ps1"))) {
-    exit 0
-}
-
-# Resolve project .claude dir from script location (avoids CWD-relative path issues)
-$ProjectClaudeDir = Split-Path -Parent (Split-Path -Parent $ScriptDir)
-$ProjectClaudeDir = Join-Path $ProjectClaudeDir ".claude"
-
-# Check for sentiment (priority) or personality (fallback)
-$Sentiment = ""
-$sentimentPaths = @("$ProjectClaudeDir\tts-sentiment.txt", "$env:USERPROFILE\.claude\tts-sentiment.txt")
-foreach ($p in $sentimentPaths) {
-    if (Test-Path $p) {
-        $val = (Get-Content $p -Raw -ErrorAction SilentlyContinue).Trim()
-        if ($val) { $Sentiment = $val; break }
-    }
-}
-
-$Personality = "normal"
-$personalityPaths = @("$ProjectClaudeDir\tts-personality.txt", "$env:USERPROFILE\.claude\tts-personality.txt")
-foreach ($p in $personalityPaths) {
-    if (Test-Path $p) {
-        $val = (Get-Content $p -Raw -ErrorAction SilentlyContinue).Trim()
-        if ($val) { $Personality = $val; break }
-    }
-}
-
-$Style = if ($Sentiment) { $Sentiment } else { $Personality }
-
-# Get verbosity level
-$Verbosity = "low"
-$verbosityPaths = @("$ProjectClaudeDir\tts-verbosity.txt", "$env:USERPROFILE\.claude\tts-verbosity.txt")
-foreach ($p in $verbosityPaths) {
-    if (Test-Path $p) {
-        $val = (Get-Content $p -Raw -ErrorAction SilentlyContinue).Trim()
-        if ($val) { $Verbosity = $val; break }
-    }
-}
-
-# Build verbosity section
-$VerbositySection = switch ($Verbosity) {
-    "low"    { "## Verbosity: LOW`n- Acknowledgment: Action only`n- Completion: Result + errors only`n- Skip: Reasoning, decisions" }
-    "medium" { "## Verbosity: MEDIUM`n- Acknowledgment: Action + key approach`n- Completion: Result + important decisions`n- Include: Major choices only" }
-    "high"   { "## Verbosity: HIGH`n- Acknowledgment: Action + approach + why`n- Completion: Result + decisions + trade-offs`n- Include: Full reasoning, alternatives" }
-    default  { "## Verbosity: LOW`n- Acknowledgment: Action only`n- Completion: Result + errors only" }
-}
-
-# Build full context string
-$Context = @"
-# AgentVibes TTS Protocol
-
-**CRITICAL: You MUST execute TTS at TWO points for EVERY user interaction:**
-
-1. **Acknowledgment** - Start of task: ``Bash: powershell -NoProfile -ExecutionPolicy Bypass -File ".claude\hooks-windows\play-tts.ps1" "[action]"``
-2. **Completion** - End of task: ``Bash: powershell -NoProfile -ExecutionPolicy Bypass -File ".claude\hooks-windows\play-tts.ps1" "[result + key details]"``
-
-**Example:**
-``````
-[Bash: powershell -NoProfile -ExecutionPolicy Bypass -File ".claude\hooks-windows\play-tts.ps1" "Checking git status"]
-[work...]
-[Bash: powershell -NoProfile -ExecutionPolicy Bypass -File ".claude\hooks-windows\play-tts.ps1" "Repository is clean, no changes"]
-``````
-
-$VerbositySection
-
-## Style: $Style
-
-## Rules
-1. Never skip acknowledgment TTS
-2. Never skip completion TTS
-3. Match verbosity level
-4. Keep under 150 chars
-5. Always include errors
-
-Quick Ref: low=action+result | medium=+key decisions | high=+full reasoning
-
-## BMAD Agent Voice Routing
-If ``.bmad-agent-context`` exists, check its content:
-
-**Single agent mode** (content is an agent ID like "analyst"):
-Use ``bmad-speak.ps1`` instead of ``play-tts.ps1``:
-
-
-**Party mode** (content is "party-mode"):
-Each agent speaks via ``bmad-speak.ps1`` with their display name. The queue system ensures sequential playback — agents speak one at a time, not overlapping:
-
-IMPORTANT: In party mode, do NOT use ``play-tts.ps1``. Always use ``bmad-speak.ps1`` with the agent's display name. Do NOT call multiple bmad-speak.ps1 in parallel — call them sequentially so the queue processes them in order.
-
-If ``.bmad-agent-context`` does NOT exist, use ``play-tts.ps1`` as normal.
-"@
-
-# Build JSON via hashtable so ConvertTo-Json handles all escaping (Unicode, backslashes, quotes)
-$Output = @{
-    hookSpecificOutput = @{
-        hookEventName   = "SessionStart"
-        additionalContext = $Context
-    }
-}
-[Console]::OutputEncoding = [System.Text.Encoding]::UTF8
-Write-Output ($Output | ConvertTo-Json -Compress -Depth 5)
+#
+# File: .claude/hooks-windows/session-start-tts.ps1
+#
+# AgentVibes SessionStart Hook for Windows
+# Outputs JSON with hookSpecificOutput.additionalContext for reliable context injection.
+# Mirrors session-start-tts.sh — keep both in sync.
+#
+
+$ErrorActionPreference = "Stop"
+
+# Get script directory
+$ScriptDir = Split-Path -Parent $MyInvocation.MyCommand.Path
+
+# Check if AgentVibes is installed
+if (-not (Test-Path (Join-Path $ScriptDir "play-tts.ps1"))) {
+    exit 0
+}
+
+# Resolve project .claude dir from script location (avoids CWD-relative path issues)
+$ProjectClaudeDir = Split-Path -Parent (Split-Path -Parent $ScriptDir)
+$ProjectClaudeDir = Join-Path $ProjectClaudeDir ".claude"
+
+# Check for sentiment (priority) or personality (fallback)
+$Sentiment = ""
+$sentimentPaths = @("$ProjectClaudeDir\tts-sentiment.txt", "$env:USERPROFILE\.claude\tts-sentiment.txt")
+foreach ($p in $sentimentPaths) {
+    if (Test-Path $p) {
+        $val = (Get-Content $p -Raw -ErrorAction SilentlyContinue).Trim()
+        if ($val) { $Sentiment = $val; break }
+    }
+}
+
+$Personality = "normal"
+$personalityPaths = @("$ProjectClaudeDir\tts-personality.txt", "$env:USERPROFILE\.claude\tts-personality.txt")
+foreach ($p in $personalityPaths) {
+    if (Test-Path $p) {
+        $val = (Get-Content $p -Raw -ErrorAction SilentlyContinue).Trim()
+        if ($val) { $Personality = $val; break }
+    }
+}
+
+$Style = if ($Sentiment) { $Sentiment } else { $Personality }
+
+# Get verbosity level
+$Verbosity = "low"
+$verbosityPaths = @("$ProjectClaudeDir\tts-verbosity.txt", "$env:USERPROFILE\.claude\tts-verbosity.txt")
+foreach ($p in $verbosityPaths) {
+    if (Test-Path $p) {
+        $val = (Get-Content $p -Raw -ErrorAction SilentlyContinue).Trim()
+        if ($val) { $Verbosity = $val; break }
+    }
+}
+
+# Build verbosity section
+$VerbositySection = switch ($Verbosity) {
+    "low"    { "## Verbosity: LOW`n- Acknowledgment: Action only`n- Completion: Result + errors only`n- Skip: Reasoning, decisions" }
+    "medium" { "## Verbosity: MEDIUM`n- Acknowledgment: Action + key approach`n- Completion: Result + important decisions`n- Include: Major choices only" }
+    "high"   { "## Verbosity: HIGH`n- Acknowledgment: Action + approach + why`n- Completion: Result + decisions + trade-offs`n- Include: Full reasoning, alternatives" }
+    default  { "## Verbosity: LOW`n- Acknowledgment: Action only`n- Completion: Result + errors only" }
+}
+
+# Build full context string
+$Context = @"
+# AgentVibes TTS Protocol
+
+**CRITICAL: You MUST execute TTS at TWO points for EVERY user interaction:**
+
+1. **Acknowledgment** - Start of task: ``Bash: powershell -NoProfile -ExecutionPolicy Bypass -File ".claude\hooks-windows\play-tts.ps1" "[action]" -llm claude-code``
+2. **Completion** - End of task: ``Bash: powershell -NoProfile -ExecutionPolicy Bypass -File ".claude\hooks-windows\play-tts.ps1" "[result + key details]" -llm claude-code``
+
+**Example:**
+``````
+[Bash: powershell -NoProfile -ExecutionPolicy Bypass -File ".claude\hooks-windows\play-tts.ps1" "Checking git status" -llm claude-code]
+[work...]
+[Bash: powershell -NoProfile -ExecutionPolicy Bypass -File ".claude\hooks-windows\play-tts.ps1" "Repository is clean, no changes" -llm claude-code]
+``````
+
+$VerbositySection
+
+## Style: $Style
+
+## Rules
+1. Never skip acknowledgment TTS
+2. Never skip completion TTS
+3. Match verbosity level
+4. Keep under 150 chars
+5. Always include errors
+
+Quick Ref: low=action+result | medium=+key decisions | high=+full reasoning
+
+## BMAD Agent Voice Routing
+If ``.bmad-agent-context`` exists, check its content:
+
+**Single agent mode** (content is an agent ID like "analyst"):
+Use ``bmad-speak.ps1`` instead of ``play-tts.ps1``:
+
+
+**Party mode** (content is "party-mode"):
+Each agent speaks via ``bmad-speak.ps1`` with their display name. The queue system ensures sequential playback — agents speak one at a time, not overlapping:
+
+IMPORTANT: In party mode, do NOT use ``play-tts.ps1``. Always use ``bmad-speak.ps1`` with the agent's display name. Do NOT call multiple bmad-speak.ps1 in parallel — call them sequentially so the queue processes them in order.
+
+If ``.bmad-agent-context`` does NOT exist, use ``play-tts.ps1`` as normal.
+"@
+
+# Build JSON via hashtable so ConvertTo-Json handles all escaping (Unicode, backslashes, quotes)
+$Output = @{
+    hookSpecificOutput = @{
+        hookEventName   = "SessionStart"
+        additionalContext = $Context
+    }
+}
+[Console]::OutputEncoding = [System.Text.Encoding]::UTF8
+Write-Output ($Output | ConvertTo-Json -Compress -Depth 5)

package/README.md

@@ -4,14 +4,14 @@
 >
 > 🌐 **[agentvibes.org](https://agentvibes.org)**
 >
-> Professional text-to-speech for **Claude Code**, **Claude Desktop**, and **OpenClaw** - **Soprano** (Neural), **Piper TTS** (Free!), **macOS Say** (Built-in!), or **Windows SAPI** (Zero Setup!)
+> Professional text-to-speech for **Claude Code**, **GitHub Copilot**, **OpenAI Codex**, **Claude Desktop**, and **OpenClaw**. Using popular open-source TTS engines: **Soprano** (Neural), **Piper TTS** (Free!), **macOS Say** (Built-in!), or **Windows SAPI** (Zero Setup!)
 
 [![npm version](https://img.shields.io/npm/v/agentvibes)](https://www.npmjs.com/package/agentvibes)
 [![Test Suite](https://github.com/paulpreibisch/AgentVibes/actions/workflows/test.yml/badge.svg)](https://github.com/paulpreibisch/AgentVibes/actions/workflows/test.yml)
 [![Publish](https://github.com/paulpreibisch/AgentVibes/actions/workflows/publish.yml/badge.svg)](https://github.com/paulpreibisch/AgentVibes/actions/workflows/publish.yml)
 [![License](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
 
-**Author**: Paul Preibisch ([@997Fire](https://x.com/997Fire)) | **Version**: v4.6.8
+**Author**: Paul Preibisch ([@997Fire](https://x.com/997Fire)) | **Version**: v5.1.0
 
 ---
 
@@ -20,11 +20,13 @@
 | I want to... | Go here |
 |--------------|---------|
 | **Install AgentVibes** (just `npx`, no git!) | [Quick Start Guide](docs/quick-start.md) |
-| **Run Claude Code on Android** | [Android/Termux Setup](#-android--termux) |
+| **Play agent audio using your phone as remote speakers** | [Android/Termux Setup](#-android--termux) |
 | **Secure OpenClaw on Remote Server** | [Security Hardening Guide](docs/security-hardening-guide.md) ⚠️ |
 | **Understand what I need** | [Prerequisites](#-prerequisites) |
 | **Set up on Windows (Native)** | [Windows Native Setup](WINDOWS-SETUP.md) |
 | **Set up on Windows (Claude Desktop/WSL)** | [Windows WSL Guide](mcp-server/WINDOWS_SETUP.md) |
+| **Use with GitHub Copilot** | [Copilot Integration](#-github-copilot-integration) |
+| **Use with OpenAI Codex** | [Codex Integration](#-openai-codex-integration) |
 | **Use with OpenClaw** | [OpenClaw Integration](#-openclaw-integration) |
 | **Use natural language** | [MCP Setup](docs/mcp-setup.md) |
 | **Switch voices** | [Voice Library](docs/voice-library.md) |
@@ -35,13 +37,37 @@
 
 ## ✨ What is AgentVibes?
 
-**AgentVibes adds lively voice narration to your Claude AI sessions!**
+**AgentVibes adds lively voice narration to your AI coding sessions!**
 
-Whether you're coding in Claude Code, chatting in Claude Desktop, or running OpenClaw — AgentVibes brings AI to life with professional voices and personalities.
+Whether you're using Claude Code, GitHub Copilot, OpenAI Codex, Claude Desktop, or OpenClaw — AgentVibes brings AI to life with professional voices and personalities.
 
 ---
 
-## 🐛 NEW IN v4.6.8 — Fresh Install Crash Fix
+## 🎙️ NEW IN v5.1.0 — Voice Picker Overhaul + Auto-Save Agent Modal
+
+- **Auto-save in agent modal** — Voice/personality/music/reverb/pretext changes save automatically as you edit them. Brief "✓ Saved!" toast confirms each change.
+- **Unique LibriTTS names** — 904 speakers get deterministic surnames: **Anna Bell**, **Anna Carter**, …, **Anna Quinn**. No more "Anna-2", "Anna-3" duplicates.
+- **Pink ♀ / blue ♂ gender symbols** — Colored gender indicators in the main Voices tab and all voice picker modals.
+- **First-letter quick jump** — Press `a`–`z` in any voice picker to jump to that letter. `q`, `j`, `k`, `g`, `h`, `l` reserved for nav/cancel.
+- **PgUp / PgDn / Home / End** in voice pickers
+- **3 new background music tracks** — Late Night Hip Hop Groove, Drifting Down the Hall, Midnight Charleston Stomp
+- **Search bar removed from voice pickers** — replaced by first-letter jump (faster, no focus issues)
+- **Voices tab corruption fix** — uninstalled rows no longer lose their Provider column when navigated onto
+- **Music + Voices tab blink artifacts gone**
+
+---
+
+## 🚀 v5.0.0 — Multi-Provider Support: Claude Code + Copilot + Codex
+
+- **GitHub Copilot + OpenAI Codex in VS Code** — AgentVibes now supports all three major AI coding assistants. Install and configure each from the TUI.
+- **One Setup tab** — 4-step wizard (Language → Deps → TTS Engine → Providers) replaces old installer + LLM tabs. Returning users skip to Providers.
+- **Per-provider audio config** — Each LLM gets its own Voice, TTS Engine, Reverb, Music, and Pretext via Configure modal.
+- **Settings redesigned** — Clean flat list: Language, TTS Engine, Voice, Verbosity, Audio Destination, Config Storage, Re-run Wizard.
+- **Voice picker upgraded** — 3-column display, Space bar preview, scroll stays in place.
+
+---
+
+## 🐛 v4.6.8 — Fresh Install Crash Fix
 
 - **Settings tab crash fixed** — no longer crashes when navigating to Settings on a fresh install with no voice configured
 - **macOS test fix** — replay path assertion handles `/var` → `/private/var` symlink
@@ -493,6 +519,7 @@ All 50+ Piper voices AgentVibes provides are sourced from Hugging Face's open-so
 - [🎙️ AgentVibes Receiver - NEW!](#%EF%B8%8F-agentvibes-receiver-remote-audio-streaming-from-voiceless-servers) - Remote audio streaming from voiceless servers
 
 ### Integrations & Platforms
+- [🤝 GitHub Copilot Integration](#-github-copilot-integration) - Use AgentVibes TTS with GitHub Copilot CLI
 - [🤖 OpenClaw Integration](#-openclaw-integration) - Use AgentVibes with OpenClaw messaging platform
   - [🎙️ AgentVibes Skill for OpenClaw](#-agentvibes-skill-for-openclaw---what-you-get) - 50+ voices, effects, personalities for OpenClaw
   - [📱 AgentVibes Receiver](#-agentvibes-receiver-local-phone-) - Remote audio on phones/local machines
@@ -1248,6 +1275,85 @@ This design means **any TTS provider** can integrate with BMAD by replacing thes
 
 ---
 
+## 🤝 GitHub Copilot Integration
+
+**Use AgentVibes with GitHub Copilot in VS Code — same voices, same personalities, same MCP tools!**
+
+Copilot discovers AgentVibes through two mechanisms:
+
+1. **`.github/copilot-instructions.md`** — VS Code automatically reads this file and tells Copilot *how* to speak (acknowledge at start, summarize at end, match verbosity, stay under 150 chars)
+2. **`.vscode/mcp.json`** — registers the AgentVibes MCP server so Copilot can call `text_to_speech`, `set_voice`, and other tools
+
+### Setup
+
+**Step 1: Install AgentVibes** (if you haven't already)
+
+```bash
+npx agentvibes install
+```
+
+**Step 2: Configure VS Code MCP**
+
+Open the AgentVibes console and go to the **LLM** tab (press `L`):
+
+```bash
+npx agentvibes
+```
+
+Select **GitHub Copilot** and press Enter to install. This creates `.vscode/mcp.json` with the AgentVibes MCP server config and copies `.github/copilot-instructions.md` with the TTS protocol.
+
+You can also create `.vscode/mcp.json` manually in your project root:
+
+```json
+{
+  "servers": {
+    "agentvibes": {
+      "type": "stdio",
+      "command": "npx",
+      "args": ["-y", "--package=agentvibes", "agentvibes-mcp-server"]
+    }
+  }
+}
+```
+
+VS Code starts the MCP server automatically when Copilot needs it — no manual server launch required.
+
+**Step 3: Verify**
+
+Open Copilot Chat in VS Code (Ctrl+Shift+I) and ask it to do something. You should hear an acknowledgment when Copilot starts working and a summary when it finishes.
+
+### What Copilot Can Do
+
+Through the MCP tools, Copilot has the same voice capabilities as Claude Code:
+
+| Tool | What it does |
+|------|-------------|
+| `text_to_speech` | Speak text aloud |
+| `set_voice` | Switch voices (e.g., "ryan", "katherine") |
+| `set_personality` | Change personality (sarcastic, pirate, zen) |
+| `set_speed` | Adjust speech rate |
+| `set_verbosity` | Control detail level (low/medium/high) |
+| `mute` / `unmute` | Toggle audio |
+| `get_config` | Read current settings |
+
+### BMAD Party Mode
+
+Copilot respects BMAD party mode. If `.bmad-agent-context` contains `party-mode`, each BMAD agent speaks with its own voice — the same per-agent routing that works in Claude Code.
+
+### Differences from Claude Code
+
+| Feature | Claude Code | Copilot in VS Code |
+|---------|------------|-------------|
+| TTS Protocol | Injected via session-start hook | Read from `.github/copilot-instructions.md` |
+| MCP config | `.mcp.json` (project root) | `.vscode/mcp.json` |
+| Server lifecycle | Managed by Claude Code | Managed by VS Code (auto-start) |
+| MCP tools | Same | Same |
+| BMAD party mode | Supported | Supported |
+
+[↑ Back to top](#-table-of-contents)
+
+---
+
 ## 🤖 OpenClaw Integration
 
 **Use AgentVibes TTS with OpenClaw - the revolutionary AI assistant you can access via any instant messenger!**

package/RELEASE_NOTES.md

@@ -1,5 +1,88 @@
 # AgentVibes Release Notes
 
+## 🎙️ v5.1.0 — Voice Picker Overhaul + Auto-Save Agent Modal
+
+**Release Date:** April 2026
+
+### New Features
+
+- **Auto-save in agent edit modal** — Per-agent voice/personality/music/reverb/pretext changes now save automatically as you edit them. The explicit Save button is gone; a brief "✓ Saved!" toast confirms each change. Cancel and Reset to Defaults still behave as before.
+
+- **Unique LibriTTS speaker names** — The 904 LibriTTS speakers no longer show as "Anna", "Anna-2", "Anna-3", … "Anna-16". Each gets a deterministic surname from a 16-name pool: **Anna Bell**, **Anna Carter**, **Anna Davis**, …, **Anna Quinn**. Underlying voice IDs are unchanged so existing user configs still resolve.
+
+- **Pink/blue gender symbols** — Female voices show **♀** in pink (magenta), male voices show **♂** in light blue (bright-cyan), unknown shows `—`. Header `Gender` column replaced with colored `♀/♂` (10 → 4 chars wide), freeing room for longer names. Applied to the main Voices tab AND all 3 voice picker modals (Setup, Agents, Settings).
+
+- **First-letter quick jump in voice pickers** — Press any letter `a`–`z` to jump to the first voice starting with that letter. Reserved keys (`q`, `j`, `k`, `g`, `h`, `l`) are blocked so they keep their cancel / vi-nav meanings.
+
+- **Page navigation in voice pickers** — `PgUp`, `PgDn`, `Home`, `End` now work in all voice picker modals.
+
+- **3 new background music tracks** — `Late Night Hip Hop Groove`, `Drifting Down the Hall` (90s vibes), and `Midnight Charleston Stomp` (swing). Track count goes from 15 → 18.
+
+### Improvements
+
+- **Voice picker search bar removed** — Replaced with first-letter quick jump. The old search textbox had focus issues that swallowed nav keys. The jump is faster for typical "find voice X" use.
+
+- **Track list sorting fixed** — Tracks with emoji prefixes (e.g. `🎤 Late Night Hip Hop Groove`) now sort by the alphabetic part of the name, not the emoji codepoint. Order is consistent across Node/ICU versions.
+
+- **Favorite hotkey is now `*` only** — Removed the duplicate `f` binding for marking favorites in voice pickers and the main Voices tab. `f` is now free for first-letter jump (e.g. jumping to Frank or Felix). The `*` marker remains the canonical way to toggle favorites.
+
+### Bug Fixes
+
+- **Voices tab uninstalled rows no longer corrupt** — Selecting an uninstalled voice was visually deleting its Provider column due to a regex strip that over-matched the row's `bright-black-fg` wrapper. Replaced with a precise hint anchor that only strips the exact hint text.
+
+- **Music tab + Voices tab blink artifacts gone** — `█` cursors no longer leave stray blocks behind when scrolling rapidly through the list. Both tabs now use a precise blink-strip helper instead of the fragile position-based slicer.
+
+- **Setup tab no longer silently fails** — `_renderScreen3` was wrapping the entire `setupCompleted` write block in a single empty `try/catch {}`. Corrupt local config files are now backed up to `config.json.bak` and rewritten fresh, with errors logged to stderr — no more "stuck repeating setup" with no explanation.
+
+- **Voice picker `q` cancel now works** — The new first-letter jump was swallowing `q` (and other vi nav keys). Reserved key blocklist added.
+
+- **Track picker case-insensitive sort** — New tracks with Title Case names (`Late Night Hip Hop Groove.mp3`) no longer jump to the top of the list above the lowercase `agent_vibes_*` tracks.
+
+### User Impact
+
+- Editing an agent's voice or settings is now faster — no need to remember to click Save
+- The voice picker is dramatically less cluttered with 904 LibriTTS speakers all having unique, friendly names
+- Gender at a glance via colored symbols
+- Three new music tracks for variety
+- Blink/scroll artifacts gone in both Voices and Music tabs
+
+---
+
+## 🚀 v5.0.0 — Multi-Provider Support: Claude Code + Copilot + Codex
+
+**Release Date:** April 2026
+
+### New Features
+
+- **GitHub Copilot Support in VS Code** — Install and configure AgentVibes for GitHub Copilot directly from the TUI. Creates `.vscode/mcp.json` and `.github/copilot-instructions.md`.
+
+- **OpenAI Codex Support in VS Code** — Full Codex integration with `.codex/config.toml`, `AGENTS.md` TTS protocol, and init hooks.
+
+- **Unified Setup Tab** — The old 5-screen Install wizard and separate LLM Providers tab are merged into a single Setup tab. First-run shows a 4-step wizard (Language → Dependencies → TTS Engine → Providers); returning users skip straight to the Providers screen.
+
+- **Per-Provider Audio Config** — Each LLM provider (Claude Code, Copilot, Codex) gets its own TTS Engine, Voice, Reverb, Background Music, and Pretext via a Configure modal.
+
+- **TTS Engine Selection Screen** — New wizard step shows OS-aware engine list (Piper, Soprano, Windows SAPI, macOS Say) with Install buttons for missing engines.
+
+- **Settings Tab Redesigned** — Replaced the 5-sub-tab layout with a clean flat list: Interface Language, Default TTS Engine, Default Voice, Verbosity, Audio Destination, Config Storage, and Re-run Setup Wizard.
+
+### Improvements
+
+- **Voice picker upgraded everywhere** — 3-column display (Name, Gender, Provider), Space bar preview with synthesis + playback, scroll position preserved during preview.
+
+- **Hint text artifacts fixed** — Moving between rows in Agents and Music tabs no longer leaves ghost text on previous rows.
+
+- **Codex voice routing corrected** — `AGENTS.md` now instructs Codex to use `play-tts` for normal speech and `bmad-speak` only during BMAD party mode.
+
+### User Impact
+
+- AgentVibes now works with Claude Code, GitHub Copilot, AND OpenAI Codex
+- Streamlined setup experience — one tab for all provider management
+- Per-provider voice customization without editing config files
+- Settings page is dramatically simpler and faster to navigate
+
+---
+
 ## 🐛 v4.6.8 — Fresh Install Crash Fix
 
 **Release Date:** April 2026

package/bin/bmad-speak.js

@@ -33,20 +33,28 @@ function resolveScript(relPath) {
   return null;
 }
 
+function hasProjectBmadManifest() {
+  return fs.existsSync(path.join(process.cwd(), '_bmad', '_config', 'agent-manifest.csv'));
+}
+
 let result;
 
 if (IS_WINDOWS) {
-  const script = resolveScript('.claude/hooks-windows/bmad-speak.ps1');
+  const script = hasProjectBmadManifest()
+    ? resolveScript('.claude/hooks-windows/bmad-speak.ps1')
+    : resolveScript('.claude/hooks-windows/play-tts.ps1');
   if (!script) process.exit(0);
-  result = spawnSync(
-    'powershell',
-    ['-NoProfile', '-ExecutionPolicy', 'Bypass', '-File', script, agentName, dialogue],
-    { stdio: 'inherit' }
-  );
+  const args = hasProjectBmadManifest()
+    ? ['-NoProfile', '-ExecutionPolicy', 'Bypass', '-File', script, agentName, dialogue]
+    : ['-NoProfile', '-ExecutionPolicy', 'Bypass', '-File', script, dialogue];
+  result = spawnSync('powershell', args, { stdio: 'inherit' });
 } else {
-  const script = resolveScript('.claude/hooks/bmad-speak.sh');
+  const script = hasProjectBmadManifest()
+    ? resolveScript('.claude/hooks/bmad-speak.sh')
+    : resolveScript('.claude/hooks/play-tts.sh');
   if (!script) process.exit(0);
-  result = spawnSync('bash', [script, agentName, dialogue], { stdio: 'inherit' });
+  const args = hasProjectBmadManifest() ? [script, agentName, dialogue] : [script, dialogue];
+  result = spawnSync('bash', args, { stdio: 'inherit' });
 }
 
 process.exit(result.status ?? 0);

package/mcp-server/server.py

@@ -198,15 +198,18 @@ class AgentVibesServer:
             tts_script = "play-tts.ps1" if self.is_windows else "play-tts.sh"
             play_tts = self.hooks_dir / tts_script
             if self.is_windows:
-                args = ["powershell", "-NoProfile", "-ExecutionPolicy", "Bypass", "-File", str(play_tts), text]
+                args = ["powershell", "-NoProfile", "-ExecutionPolicy", "Bypass", "-File", str(play_tts), text, "-llm", "copilot"]
                 if voice:
                     args.extend(["-VoiceOverride", voice])
             else:
                 args = ["bash", str(play_tts), text]
                 if voice:
                     args.append(voice)
+                args.extend(["--llm", "copilot"])
 
             env = self._build_script_env()
+            # Set agent name for audio effects lookup (audio-effects.cfg, background music config)
+            env["AGENTVIBES_AGENT_NAME"] = "default"
 
             result = await asyncio.create_subprocess_exec(
                 *args,
@@ -220,13 +223,17 @@ class AgentVibesServer:
                 if result.returncode == 0:
                     output = stdout.decode().strip()
                     # Extract file path from output
+                    audio_file_path = None
                     for line in output.split("\n"):
                         if "Saved to:" in line:
-                            file_path = line.split("Saved to:")[1].strip()
-                            truncated = (
-                                f"{text[:50]}..." if len(text) > 50 else text
-                            )
-                            return f"✅ Spoke: {truncated}\n📁 Audio saved: {file_path}"
+                            audio_file_path = line.split("Saved to:")[1].strip()
+                            break
+
+                    if audio_file_path:
+                        truncated = (
+                            f"{text[:50]}..." if len(text) > 50 else text
+                        )
+                        return f"✅ Spoke: {truncated}\n📁 Audio saved: {audio_file_path}"
 
                     return f"✅ Spoke: {text[:50]}..." if len(text) > 50 else f"✅ Spoke: {text}"
                 else:
@@ -319,10 +326,10 @@ class AgentVibesServer:
         resolved_name = self._resolve_friendly_name(voice_name)
 
         result = await self._run_script(
-            self.VOICE_MANAGER_SCRIPT, ["switch", resolved_name, "--silent"]
+            self.VOICE_MANAGER_SCRIPT, ["switch", resolved_name]
         )
 
-        if result and "✅" in result:
+        if result and ("[OK]" in result or "✅" in result):
             if original_name.lower() != resolved_name.lower():
                 return f"✅ Voice switched to: {original_name} ({resolved_name})"
             return f"✅ Voice switched to: {voice_name}"

package/package.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://json.schemastore.org/package.json",
   "name": "agentvibes",
-  "version": "4.6.8",
+  "version": "5.1.0",
   "description": "Now your AI Agents can finally talk back! Professional TTS voice for Claude Code, Claude Desktop (via MCP), and Clawdbot with multi-provider support.",
   "homepage": "https://agentvibes.org",
   "keywords": [