agentvibes 5.3.0 → 5.5.0

package/RELEASE_NOTES.md

@@ -1,1976 +1,1172 @@
-# AgentVibes Release Notes
-
-## 🎯 v5.3.0 — Take Control of Remote Voices
-
-**Release Date:** April 2026
-
-If you're using AgentVibes to send voice announcements from a server to
-your phone, laptop, or another machine, this release puts you in the
-driver's seat. Every call can now pick its own voice, background music,
-intro phrase, reverb, volume, and speed — right from the command line,
-for just that one message.
-
-### ✨ What's New
-
-#### You can now customize every announcement individually
-
-Before, if you wanted a different voice or music for one specific
-message, you had to change a config file (and remember to change it
-back). Now you just add a flag to the command.
-
-Want Winston to speak in his British accent with jazz playing for this
-one deploy notification? Easy:
-
-```bash
-bash .claude/hooks/play-tts-ssh-remote.sh \
-  --text "Deploy complete" \
-  --voice "en_US-ryan-high" \
-  --pretext "Winston here" \
-  --music "Late Night Hip Hop Groove.mp3" \
-  --volume 0.25
-```
-
-Anything you don't specify falls back to your normal settings. Want to
-skip the intro phrase just this once? Pass `--pretext ""` and it stays
-silent before the message.
-
-**Available flags:**
-- `--voice` — which Piper voice to use
-- `--pretext` — the intro phrase before the message (pass `""` to skip it)
-- `--music` — background music track (filenames with spaces now work!)
-- `--volume` — how loud the background music is (0.0 to 1.0)
-- `--effects` — sound effects chain like reverb
-- `--speed` — how fast the voice speaks
-- `--provider` — which TTS engine to use
-- `--agent` — which agent personality to use
-
-The old way of calling the script still works, so nothing you've already
-set up will break.
-
-### 🛠 Reliability Fixes
-
-- **Long messages and special characters no longer get cut off.** On
-  Windows, long announcements or text with quotes, apostrophes, or
-  emoji were getting mangled before they reached the voice engine.
-  Fixed — your message now arrives exactly as you sent it, no matter
-  how long or weird.
-
-- **Voice announcements now work on Windows servers with no monitor.**
-  Windows refuses to play audio in the "service" session that SSH
-  normally uses. A small background helper now runs in your regular
-  user session and picks up announcements from a queue, so audio plays
-  correctly even on headless servers.
-
-- **Voice preview in the TUI works on remote servers.** Before, if you
-  previewed a voice from a server with no speakers, it would try to
-  play locally (and fail). Now it correctly streams to whatever remote
-  device you've configured.
-
-- **No more double intro phrases.** If you set a pretext on both the
-  sending server and the receiving machine, you used to hear it twice.
-  The sender's version wins now — the receiver won't add its own on top.
-
-- **Remote streaming settings now actually stick.** A recent change
-  accidentally caused remote-streaming setups (`ssh-remote`,
-  `agentvibes-receiver`) to get overridden and fall back to local
-  playback. Fixed.
-
-- **Long announcements don't get killed mid-sentence.** The safety
-  timeout that stops stuck audio was too aggressive for long messages.
-  It's now generous enough to handle paragraph-length announcements.
-
-- **Cleaner installer state** — when you install AgentVibes for Claude
-  Code, it now writes its TTS provider file explicitly instead of
-  relying on implicit state.
-
-### 🧪 Testing
-
-55 new tests make sure BMAD party mode keeps working: each agent gets
-their unique voice and music, agents don't accidentally share the same
-Piper speaker ID, and the installer always points party mode at the
-cross-platform entry point.
-
----
-
-## 🎯 v5.2.1 — Multi-LLM Identity & Install Polish
-
-**Release Date:** April 2026
-
-Polished LLM routing for Copilot/Codex and refined the setup experience.
-
-### ✨ What's New
-
-#### Multi-LLM Identity Routing
-
-- **GitHub Copilot now gets its own voice, pretext, and background music** — fully distinct from Claude Code and Codex. Say hello to "Copilot here" with bossa nova on the way in.
-
-- **Per-tool MCP configs with explicit identity** — each AI tool (`.vscode/mcp.json`, `.codex/config.toml`, `~/.copilot/mcp-config.json`) sets its own `AGENTVIBES_LLM` so routing is deterministic.
-
-- **`get_config` MCP tool now returns the detected LLM** — so the calling assistant can confirm its routing and respond with the right voice out of the gate.
-
-- **Linux compatibility refinements** — CRLF line endings, permissions, and transport provider override handling.
-
-#### Setup Flow Improvements
-
-- **Keyboard navigation flow** — pressing Enter through the Install buttons (Claude → Copilot → Codex) now jumps to **Claude Configure** next, letting you walk all three Configures in order before landing on Default.
-
-- **Down-arrow skips the Default row** from Install/Remove columns (consistent with info-box dismiss nav).
-
-- **Partial install success messaging** — if file copies succeed but MCP config needs a nudge, you'll see a clear warning instead of a generic failure.
-
-#### Defaults
-
-- **Claude Code default background music** set to Chillwave (`agent_vibes_chillwave_v2_loop.mp3`).
-
-#### Under the Hood
-
-- LLM key validation tightened for safer env var handling.
-- Improved error logging for edge cases in Copilot CLI config writes.
-- Known limitation documented: if you launch VS Code from a Claude-Code-started terminal, `CLAUDECODE=1` can leak — workaround is to `unset CLAUDECODE` first.
-
----
-
-## 🎯 v5.2.0 — Remote Voice Preview + Caveman Mode + Voice Ratings
-
-**Release Date:** April 2026
-
-This release adds remote TTS preview support, a new ultra-terse verbosity mode, and thumbs up/down voice ratings across the TUI.
-
-### New Features
-
-- **Caveman verbosity mode** — New `caveman` verbosity level for ultra-terse TTS output. Fragments instead of sentences. Set via `/agent-vibes:verbosity caveman` or the MCP `set_verbosity` tool. Auto-downloads a voice on fresh install if none are present.
-
-- **Thumbs up/down voice ratings** — Replace the old star favorites with 👍/👎 ratings. Press `+` to thumbs up, `-` to thumbs down in both the Voices tab and the voice picker (Setup tab). Ratings persist across sessions and are shared between all voice selection UIs.
-
-- **Remote voice preview** — Voice preview in the TUI Voices tab, voice picker, and voice browser now works on headless servers. When the active provider is `ssh-remote` or `agentvibes-receiver`, preview routes through `play-tts.sh` to play audio on the remote receiver instead of requiring local Piper + audio player. Platform-aware: uses PowerShell on Windows, bash on Linux.
-
-- **SSH receiver provider routing** — `ssh-remote` and `agentvibes-receiver` are now first-class providers in `play-tts.sh`. Both the `speak_text()` function and the main routing case statement support them, eliminating "Unknown provider" errors.
-
-### Fixes
-
-- **Auto-patch LibriTTS speaker names** — Voice download now auto-patches LibriTTS speaker names so multi-speaker voices work correctly out of the box.
-
-- **Voice validation regex hardened** — The VOICE parameter regex in `play-tts-ssh-remote.sh` and `play-tts-agentvibes-receiver.sh` now allows `::` (multi-speaker), `.` (locale), and spaces (speaker names) without accepting backslash (injection risk). Linux and Windows receiver templates updated to match.
-
-- **`base64` cross-platform compatibility** — `play-tts-agentvibes-receiver.sh` now probes for GNU `base64 -w 0`, falls back to BSD `-b 0`, then `tr -d '\n'`. Fixes script abort on macOS/BSD systems.
-
-- **Audio effects double-processing fix** — `play-tts-piper.ps1` skips its own audio-processor call when `AGENTVIBES_NO_PLAY` is set, preventing reverb/music from being applied twice.
-
-- **Exit code leak fix** — `play-tts.ps1` now exits with code 0 explicitly, preventing native command exit codes (piper, ffmpeg, sox) from leaking through and causing false TTS failure reports.
-
-- **Windows receiver-tab platform support** — Tailscale IP detection, local IP via PowerShell, sshd_config reading, and clipboard copy all work natively on Windows now.
-
-- **`llm:default` audio effects row** — New default row in `audio-effects.cfg` ensures remote receivers get reverb, music, and pretext even without a per-LLM config entry.
-
-- **Preview sample text** — Changed from "Here is a preview of your audio settings" to avoid Piper pronunciation glitch on the word "preview".
-
----
-
-## 🛡️ v5.1.4 — TTS Resilience Overhaul + Default LLM Provider + Per-Client Routing
-
-**Release Date:** April 2026
-
-This release closes a long-running cluster of bugs around per-LLM TTS routing, parallel audio playback, stuck-process deadlocks, and stale-audio replay. It also adds a new "Default" provider slot in the Setup tab for fallback audio configuration, and switches to a per-client config scheme that correctly routes Claude Code, GitHub Copilot (Chat + CLI), and OpenAI Codex to their own voices and pretexts.
-
-### New Features
-
-- **Default LLM provider** — New entry at the bottom of Setup → Providers. Config-only (no install/remove buttons) with a Configure button that opens the standard per-LLM audio modal. When any tool calls TTS without identifying its LLM, the `llm:default` row in `audio-effects.cfg` provides the fallback voice, pretext, music, reverb, and engine. Empty pretext by default — users opt in.
-
-- **Per-LLM background music now auto-enables** — Setting a `bg_track` on any per-LLM Configure modal now actually plays that track. Previously you also had to toggle the global `backgroundMusic.enabled` flag, which made the per-LLM bg track field silently dead.
-
-- **Copilot CLI support** — `installCopilotMcp` now writes BOTH `.vscode/mcp.json` (for VS Code Copilot Chat) AND `~/.copilot/mcp-config.json` (for the standalone `copilot` CLI, which is a different product). Fresh installs support both tools automatically.
-
-### Per-Client Routing Architecture
-
-Previously `AGENTVIBES_LLM=claude-code` was set in `.mcp.json`, which broke Copilot CLI because Copilot CLI also reads `.mcp.json` with precedence over its own `~/.copilot/mcp-config.json` — so it adopted Claude Code's env and mis-routed.
-
-The new architecture splits per-LLM identification per-client:
-
-- `.mcp.json` (project) has **no `AGENTVIBES_LLM` env block**
-- `~/.copilot/mcp-config.json` sets `AGENTVIBES_LLM=copilot` for GitHub Copilot CLI
-- `~/.codex/config.toml` sets `AGENTVIBES_LLM=codex` for OpenAI Codex
-- The MCP server (`mcp-server/server.py`) **auto-detects Claude Code** via the `CLAUDECODE=1` env var that Claude Code sets on every subprocess it spawns. Copilot CLI and Codex don't set this var, so each client reliably routes to its own config.
-
-Upgrade path: re-run the AgentVibes installer in any existing project. The new `installClaudeMcp` auto-strips any stale `AGENTVIBES_LLM` env block from existing `.mcp.json` files so Copilot CLI stops mis-routing.
-
-### TTS Resilience Overhaul (`play-tts.ps1`)
-
-- **Cross-process playback mutex** — `AgentVibesPlaybackLock` (named mutex) serializes playback across all callers: Claude Code hooks, MCP `text_to_speech`, direct CLI, party mode. No more overlapping or parallel audio when multiple LLMs run in the same project.
-
-- **Self-healing on mutex timeout** — When the 15-second mutex acquisition fails, the new code queries `Win32_Process` for any stuck `play-tts.ps1` process older than 20 seconds, calls `Stop-Process -Force` on each, and logs the kill to stderr. The next TTS call succeeds immediately — no manual `taskkill` needed.
-
-- **25-second script watchdog** — A sibling PowerShell job force-kills `play-tts.ps1` after 25 seconds regardless of where it's stuck (SoundPlayer deadlock, locked audio device, zombie ffmpeg). Guarantees forward progress.
-
-- **Hard error on mutex timeout** — Replaces the old "play anyway" fallback which just stacked more stuck processes behind the first. Now exits cleanly with code 2 and a diagnostic message.
-
-- **Exact filename capture from provider stdout** — `play-tts.ps1` parses the `[OK] Saved to: <path>` line from `play-tts-piper.ps1` and uses that exact file. Replaces the old "pick most recent `tts-*.wav` in the audio dir" heuristic which silently replayed stale audio from earlier sessions whenever synthesis failed. Root cause of the "Codex speaks Claude Code's audio" bug.
-
-- **Synthesis-failure hard error** — When the provider doesn't produce an output file, `play-tts.ps1` exits with code 3 and a loud error instead of falling back to any older file in the cache.
-
-- **Scratch file rename** — Reverb post-processing now writes to `av-reverbed-scratch.wav` and mix post-processing writes to `av-mixed-scratch.wav`. Fixed names outside the `tts-XXXXXXXX` random namespace so the file lookup can never pick them up as a synth input. Eliminates the compound `tts-xxx.wav.effected-mixed-mixed-mixed-mixed.wav` chain that accumulated on every run.
-
-- **Per-LLM voice overrides explicit `VoiceOverride`** — LLMs call `get_config` at session start and echo the returned voice back on every `text_to_speech` call as an explicit MCP parameter. With the old "explicit wins" priority, that global voice overrode per-LLM routing. Now the `llm:<key>` voice row always wins when `-llm` is set.
-
-- **`$LlmBgTrack` and `$LlmBgVolume` parsed from cfg row** — Previously fields 2 and 3 of the `llm:<key>` line were ignored in the PowerShell version (the bash version already parsed them).
-
-- **Per-LLM `bg_track` force-enables `$BgEnabled`** — When a per-LLM row specifies a background track, it's auto-enabled regardless of the global `backgroundMusic.enabled` flag.
-
-- **ASCII-only encoding** — Removed em-dash `—` and right-arrow `→` from `play-tts.ps1`. PowerShell on some Windows locales loads scripts in CP1252 and choked on the UTF-8 bytes with a misleading "Missing closing `}`" parse error.
-
-- **`lessac-medium` → `lessac-high` default for codex** — `en_US-lessac-medium` silently fails to synthesize on some Windows Piper installs (loads the model, exits with empty output). `lessac-high` works reliably and is the new default in `DEFAULT_LLM_CONFIGS.codex` and the packaged `audio-effects.cfg`.
-
-### UX Improvements
-
-- **Setup → Providers Install confirmation** — Pressing Enter to dismiss the post-install confirmation page now advances focus to the NEXT provider row's same-column button (Install → Install, Configure → Configure), instead of snapping back to Claude Code's Install. Installing all three providers is now a natural Enter-Enter-Enter flow.
-
-### Testing & Hardening
-
-- **30 new/updated regression tests** in `test/unit/llm-provider-mcp-routing.test.js` and `test/unit/windows-tts.test.js` covering:
-  - Default LLM provider config shape + setup-tab wiring
-  - `CLAUDECODE` auto-detection in `server.py`
-  - `.mcp.json` template MUST NOT contain `AGENTVIBES_LLM`
-  - `~/.copilot/mcp-config.json` writer in `installCopilotMcp`
-  - `play-tts.ps1` PowerShell-parseable (catches em-dash / encoding regressions via `[System.Management.Automation.Language.Parser]`)
-
-### How to Update
-
-```
-npm cache clean --force
-npx --yes agentvibes@5.1.4
-```
-
-If you have any existing project with AgentVibes installed, re-run the installer there once so the per-client config migration takes effect. The `.mcp.json` in each project will be auto-upgraded to strip the stale `AGENTVIBES_LLM` env block, and `~/.copilot/mcp-config.json` will be created if you have the Copilot provider enabled.
-
----
-
-## 🛡️ v5.1.3 — Hardening Pass (Adversarial Review Followup)
-
-**Release Date:** April 2026
-
-This release addresses HIGH/MEDIUM findings from a parallel adversarial review (Blind Hunter + Edge Case Hunter) of the v5.1.2 hotfix.
-
-### Bug Fixes
-
-- **Existing `.mcp.json` is now auto-migrated** — v5.1.2's installer detected an existing `.mcp.json` and printed instructions to fix it manually. Result: anyone upgrading from v5.1.0/v5.1.1 was still broken after `npm i -g agentvibes@5.1.2`. v5.1.3 detects an existing config, merges the `AGENTVIBES_LLM` env var (and the rest of the agentvibes server entry) in-place, and shows a green "✅ MCP Configuration Updated" message. Existing user `env` keys are preserved.
-
-- **`AGENTVIBES_LLM` is now validated** before being forwarded to the child shell. The bash version of `play-tts.sh` already validated `^[a-zA-Z0-9_-]+$` and rejected weird values; the Python `mcp-server/server.py` and the Windows `play-tts.ps1` did not. Now all three do, with the same regex. Invalid values are logged to stderr and treated as "unset" so TTS still works (falls back to default config).
-
-### Testing & Hardening
-
-- **`npm pack` content guard now hard-fails on `packError`** instead of silently returning early. The previous v5.1.2 implementation had `if (packError) return;` in 6 of 8 tests, which meant a real `npm pack` failure surfaced as one failed test and SIX false-greens — defeating the entire purpose of the publish guard.
-
-- **Dirty-tree check now uses `git status --porcelain`** instead of `git diff HEAD`. The previous implementation only caught modified-but-uncommitted tracked files; it missed brand-new untracked files in publishable directories (e.g. a `play-tts.ps1.new` someone forgot to delete). The v5.1.0 disaster could just as easily have shipped via an untracked stray file.
-
-- **`npm pack` exec timeout** — added explicit 60-second timeout with `SIGKILL` so the publish guard can't hang CI on a stuck registry call.
-
-- **`git status` failures hard-fail** instead of silently skipping. A missing git binary in CI was previously treated as "no findings" — that's exactly the opposite of what a security guard should do. Now it fails loudly with a message explaining the guard requires git.
-
-- **3 new regression tests** for the new validations and migration logic.
-
-### How to Update
-
-```
-npm cache clean --force
-npx --yes agentvibes@5.1.3
-```
-
-If you upgraded from v5.1.0 or v5.1.1, **re-run the AgentVibes installer once** so the new in-place migration touches your existing `.mcp.json` / `.codex/config.toml` / `.vscode/mcp.json`.
-
----
-
-## 🔀 v5.1.2 — MCP Per-LLM Routing Hotfix
-
-**Release Date:** April 2026
-
-### Bug Fixes
-
-- **MCP server no longer hardcodes `-llm copilot`** — `mcp-server/server.py` was passing `-llm copilot` for every caller, regardless of whether the actual client was Codex, Copilot, or Claude Code. This meant per-LLM voice/pretext/music routing in `audio-effects.cfg` was effectively broken: all three providers fell through to the same `llm:copilot` lookup (or worse, the global default config). The server now reads `AGENTVIBES_LLM` from the environment and forwards that as the `-llm` value, falling back to no `-llm` flag if unset.
-
-- **MCP launcher templates set `AGENTVIBES_LLM`** — Each provider's MCP config now sets the env var so the server knows which LLM is calling:
-  - Codex: `.codex/config.toml` → `env = { AGENTVIBES_LLM = "codex" }`
-  - Copilot: `.vscode/mcp.json` → `"env": { "AGENTVIBES_LLM": "copilot" }`
-  - Claude Code: `.mcp.json` → `"env": { "AGENTVIBES_LLM": "claude-code" }`
-
-  Both freshly-installed configs and the manual-instruction snippets shown by the installer include the env var.
-
-### Testing & Hardening
-
-- **24 new regression tests** in `test/unit/llm-provider-mcp-routing.test.js` and `test/unit/npm-pack-contents.test.js` enforce the contract end-to-end:
-  - Each MCP launcher template MUST set `AGENTVIBES_LLM` to the correct value
-  - `mcp-server/server.py` MUST read `AGENTVIBES_LLM` from env and MUST NOT hardcode `-llm copilot`
-  - `play-tts.ps1` MUST declare its `$llm` parameter, contain per-LLM lookup logic, export `AGENTVIBES_LLM_KEY`, and be at least 400 lines (catches the v5.1.0 mass-deletion regression)
-  - `play-tts.sh` MUST parse `--llm` and do per-LLM lookup against `audio-effects.cfg`
-  - **Working tree must be clean** before publishing (the v5.1.0 disaster guard — `npm publish` packs the working tree, not the git tag, so any uncommitted local edits get shipped)
-
-- The contract tests are **mutation-tested**: temporarily replacing `play-tts.ps1` with a broken stub fires 4 assertions with clear messages, proving the exact v5.1.0 regression would now block publish.
-
-### How to Update
-
-```
-npm cache clean --force
-npx --yes agentvibes@5.1.2
-```
-
-If you already have AgentVibes installed for one or more providers, re-run the installer's per-provider configure step (or manually add `"env": { "AGENTVIBES_LLM": "<your-llm>" }` to your `.mcp.json` / `.vscode/mcp.json` / `.codex/config.toml`).
-
----
-
-## 🩹 v5.1.1 — Windows TTS Hook Hotfix
-
-**Release Date:** April 2026
-
-### Bug Fixes
-
-- **Windows `play-tts.ps1` `-llm` parameter restored** — A regression caused the npm-published v5.1.0 package to ship a `play-tts.ps1` that lacked the `-llm` parameter and the per-LLM config lookup. Per-LLM TTS routing on Windows failed with `A parameter cannot be found that matches parameter name 'llm'`. This affected:
-  - **Setup tab Preview button** for any provider configured with per-LLM voice/effects
-  - **agentvibes MCP `text_to_speech` tool** when called from Codex / Copilot / Claude Code
-  - Any code path that invokes `play-tts.ps1 ... -llm <provider>`
-
-  The git tag `v5.1.0` had the correct file all along — only the npm tarball was affected, because `npm publish` packs the working tree (which contained an uncommitted local regression) instead of the git tag.
-
-### How to Update
-
-If you installed v5.1.0 from npm and hit the `-llm` error, clear your npx cache and reinstall:
-
-```
-npm cache clean --force
-npx --yes agentvibes@5.1.1
-```
-
-### Note for Maintainers
-
-To prevent this kind of working-tree-vs-tag drift in future releases, the release workflow should run `npm publish` from a freshly checked-out clone of the tag, not from the development working directory.
-
----
-
-## 🎙️ 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
-
-### Bug Fixes
-
-- **Settings tab no longer crashes on fresh install** — `parseMultiSpeaker()` called `.includes()` on a null voice ID when no voice was configured yet. Added a null guard that returns a safe default object. Reported by a user hitting this immediately after the install wizard completed.
-
-- **macOS /var symlink in replay test** — Fixed test assertion that failed on macOS where `/var` is a symlink to `/private/var`, causing replay path comparisons to break.
-
-- **BMAD voices pretext parsing** — `bmad-voices.md` pretext lines are now correctly parsed and markdown is stripped more thoroughly before TTS synthesis.
-
-### User Impact
-
-- New users no longer crash when navigating to Settings after a fresh install
-- Test suite passes reliably on macOS
-
----
-
-## 🎙️ v4.6.7 — Party Mode TTS Fixes
-
-**Release Date:** April 2026
-
-### Bug Fixes
-
-- **Party mode pretext now spoken** — Agent pretexts (e.g. "John, Product Manager here") were silently dropped. The pre-synthesis step ran on the response text only; when `bmad-speak.ps1` later prepended the pretext, `play-tts.ps1` skipped synthesis because a pre-synthesized WAV was already set. Pre-synthesis now includes the pretext so it's always spoken.
-
-- **Markdown stripped in party mode** — Agents were speaking asterisks, hashes, and backticks literally. `bmad-party-speak.ps1` now strips bold, italic, headings, inline code, links, and bullet markers before TTS — consistent with `bmad-speak.ps1`.
-
-- **Session start hook outputs proper JSON + BMAD routing (Windows)** — `session-start-tts.ps1` was using plain `Write-Output` text which isn't reliably injected as `additionalContext`. Now outputs `hookSpecificOutput` JSON (matching the bash version) and includes BMAD agent voice routing instructions so party mode / single agent mode TTS works correctly on Windows.
-
-- **PreToolUse hook no longer errors on grep/regex commands** — The git-push guard hook used `-match` (regex) which crashed on commands containing `\*`, `\|`, etc. Switched to `-like` wildcard and wrapped in `try/catch` so it exits silently on any non-matching command.
-
-### User Impact
-
-- Party mode agents speak their configured pretext introductions
-- No more spoken asterisks or markdown syntax in party mode
-- Windows TTS now reliably activates on session start
-
----
-
-## 🧭 v4.6.6 — Natural TUI Navigation
-
-**Release Date:** April 2026
-
-### Improvements
-
-The Settings TUI now navigates the way you'd expect. Arrow keys flow naturally top-to-bottom through the interface — main header → sub-tab bar → content → footer — with no surprising jumps or dead ends.
-
-- **Down from the main header** lands on the active sub-tab bar item, so you're always oriented before pressing further.
-- **Down through content** moves row-by-row within the current tab, reaching Full Preview and the Save buttons at the bottom. It no longer auto-jumps to the next sub-tab.
-- **Up from the first content row** returns you to the current sub-tab (not always Voice).
-- **Left/Right in the sub-tab bar** is the natural way to switch between Voice, Effects, Personality, Output, and Language tabs.
-- **Left/Right in the footer** moves between Full Preview, Save Globally, Save Locally, and Cancel Changes.
-- **Language tab** now has a proper scrollable list — arrow down to enter it, Up/Down to select a language, Enter to apply, Down past the last item to reach the footer.
-- **Readme tab** falls back to the AgentVibes package README when opened in a project folder that has no README of its own.
-- **Escape from the installer** now navigates cleanly to Settings instead of getting stuck.
-
----
-
-## 🔧 v4.6.5 — Patch Release
-
-**Release Date:** April 2026
-
-### Improvements
-
-- **TUI non-interactive hint** — Installer header now shows a two-tone hint on row 2: `Skip this TUI?` (white) + `npx agentvibes install --non-interactive` (yellow), with a sponsor link on the right. Useful for CI/CD pipelines and scripted installs.
-
-- **`--non-interactive` routing fix** — `npx agentvibes install --non-interactive` (and `--yes`/`-y`) now correctly bypasses the TUI and routes directly to the CLI installer. Previously the TUI would still launch.
-
-- **Sponsor link in TUI header** — Row 2 of the header now shows a `♡ Sponsor this Developer  github.com/sponsors/paulpreibisch` link on the right side, alongside the non-interactive hint on the left.
-
-- **Consistent line endings via `.gitattributes`** — Added `.gitattributes` to enforce `LF` for shell scripts, JS, JSON, and markdown, and `CRLF` for PowerShell scripts. Prevents spurious `bin/` file modifications showing as dirty on Windows due to CRLF noise.
-
-- **`bin/` execute bits restored** — Re-indexed `bin/agent-vibes`, `bin/agentvibes-voice-browser.js`, `bin/mcp-server.js`, and `bin/test-bmad-pr` to ensure consistent line endings; verified all four retain `chmod +x` (`100755` mode).
-
-
-### User Impact
-
-- No functional changes — infrastructure and tooling improvements only
-
----
-
-## 🐛 v4.6.4 — Patch Release
-
-**Release Date:** April 2026
-
-### Bug Fixes
-
-- **CI green on macOS** — `mktemp` with a file extension suffix (e.g. `tts-XXXXXX.wav`) works on Linux (GNU mktemp) but silently fails on macOS (BSD mktemp), which requires `XXXXXX` to be at the very end of the template. Fixed all 12 occurrences across `play-tts-piper.sh`, `play-tts-soprano.sh`, `play-tts-macos.sh`, and `audio-processor.sh` by creating the temp file without extension then renaming to add `.wav`/`.aiff`.
-
-- **macOS path symlink in test 213** — macOS transparently resolves `/var/folders/...` to `/private/var/folders/...` via a symlink. The voice-manager replay test was comparing against the unresolved path and failing. Fixed by using `cd && pwd` to get the real path before asserting.
-
-- **`bmad-party-speak.sh` execute permission** — The bash hook lost its `+x` bit, causing `party mode scripts ship with the package` CI test to fail. Restored `100755` mode via `git update-index`.
-
-### User Impact
-
-- No user-facing changes — CI infrastructure fixes only
-
----
-
-## 🐛 v4.6.3 — Patch Release
-
-**Release Date:** April 2026
-
-### Bug Fixes
-
-- **Party mode pre-synthesis uses wrong speaker for every agent** — `bmad-party-speak.ps1` extracted the trailing number from the display name suffix (e.g. `14` from `Yara-14`) and passed it directly as the piper `--speaker` index. The display suffix is a human-readable disambiguator, not the model index. `Yara-14` is speaker 860, but the bug caused piper to speak as speaker 14 ("Ivy") instead. Every configured agent was silently playing a completely different voice. The fix looks up the full speaker name in `speaker_id_map` from the `.onnx.json` file, matching what `play-tts-piper.ps1` already does correctly. Fixes [#165](https://github.com/paulpreibisch/AgentVibes/issues/165).
-
-### Testing
-
-- **New cross-platform test: `bmad-party-speak-speaker-id.test.js`** — 15 tests covering correct `speaker_id_map` lookup for all 8 configured agents, plain names without suffixes (e.g. `Evan`), missing model graceful degradation, and a regression suite that verifies the correct index disagrees with the naive suffix extraction for every agent.
-
-### User Impact
-
-- Party mode agents now speak with their correct configured voices (the voices shown in the BMAD Agents tab)
-- No configuration changes needed — the fix is automatic
-
----
-
-## 🐛 v4.6.2 — Patch Release
-
-**Release Date:** April 2026
-
-### Bug Fixes
-
-- **BMAD party mode: agents now speak with their unique voices** — The party mode `SKILL.md` was missing TTS wiring entirely. It now creates `.bmad-agent-context` on activation, calls `bmad-speak.ps1` sequentially per agent after each round, and cleans up on exit. When BMAD and AgentVibes are both installed, AgentVibes' skill now correctly overrides the BMAD version.
-
-- **LibriTTS speaker IDs resolved correctly on Windows** — `play-tts-piper.ps1` was extracting the speaker index via a regex on the voice name suffix (e.g. `Holly-7` → `7`). That suffix is a disambiguation counter, not the Piper speaker index. `Holly-7` is actually speaker 322. The script now looks up the real index from `voice-assignments.json`, with a fallback to the patched `.onnx.json`.
-
-- **LibriTTS `parseMultiSpeaker` fallback for unpatched models** — `voices-tab.js` now falls back to `voice-assignments.json` when the `.onnx.json` speaker map hasn't been patched yet with friendly names, preventing silent fallback to speaker 0 (often male) on fresh installs.
-
-- **Agent pretext spoken on Windows** — `bmad-speak.ps1` never read or applied the agent's configured pretext. It now reads `pretext` from the voice map profile, and falls back to the default `"DisplayName, Title here."` computed from the agent manifest — matching the behaviour of `bmad-speak.sh` and `AgentVoiceStore.getDefaultPretext()`.
-
-### User Impact
-
-- Party mode agents will introduce themselves by role before speaking and use their individually configured voices throughout the conversation
-- LibriTTS multi-speaker voices now reliably map to the correct speaker on first install (no manual patching required)
-- No breaking changes — all fixes are silent fallbacks or missing behaviours being added
-
----
-
-## ✨ v4.6.0 — Minor Release
-
-**Release Date:** April 2026
-
-### New Features
-
-- **BMAD party mode TTS auto-installs for all platforms** — The installer now automatically copies `bmad-party-speak.sh` (Linux/macOS/WSL) or `bmad-party-speak.ps1` (Windows) to `~/.claude/hooks/` and registers a `PostToolUse` hook in `~/.claude/settings.json`. Party mode agents now speak out of the box in any BMAD project without manual setup. Both scripts are included in critical hooks so `npx agentvibes update` keeps them fresh.
-
-### Bug Fixes
-
-- **Background music volume default** — All volume defaults lowered from 70% to 20% across the UI (settings tab, agents tab, music tab, track picker) and scripts (`audio-processor.sh`, `bmad-speak.sh`, `bmad-speak.ps1`). New installs and newly configured agents default to a much more reasonable level.
-- **bmad-speak volume inheritance** — `bmad-speak.sh` and `bmad-speak.ps1` now read the global `background-music-volume.txt` config file as the fallback volume instead of a hardcoded value.
-- **Installer wizard left arrow** — Pressing ← on the completion screen (screen 5) to move from Done-Quit to Done-Customize More no longer jumps back to the installation step.
-
-### Tests
-
-- 29 new tests: volume default regression guards across all affected files, `configurePartyModeHook` installer coverage (idempotency, settings.json registration, script copying, hook preservation), and a regression test for the screen 5 navigation fix.
-
----
-
-## 🐛 v4.5.7 — Patch Release
-
-**Release Date:** April 2026
-
-### Bug Fixes
-
-- **Background music volume default** — All volume defaults lowered from 70% to 20% across the UI (settings tab, agents tab, music tab, track picker). New installs and newly configured agents will default to a much more reasonable background music level.
-- **bmad-speak volume inheritance** — `bmad-speak.sh` and `bmad-speak.ps1` now read the global `background-music-volume.txt` config file as the fallback volume instead of a hardcoded value. Per-agent background music volume now correctly inherits the global setting when no explicit per-agent override is saved.
-
----
-
-## 🐛 v4.5.1 — Patch Release
-
-**Release Date:** April 2026
-
-### Bug Fix
-
-- **Music tab preview** — Pressing Space on a track in the Music tab now plays correctly
-  when running `npx agentvibes` from a fresh directory. Previously, if `.claude/audio/tracks/`
-  didn't exist in the current working directory, the track list showed built-in tracks but
-  Space did nothing (the player was spawned against a non-existent path). Now falls back to
-  the package-bundled tracks directory automatically.
-
----
-
-## 🌍 v4.5.0 — "Speak Every Language" Release
-
-**Release Date:** April 2026
-
-Full multilingual TUI support across all 9 languages, complete Windows security hardening, and zero failing tests.
-
-### 🌍 Multilingual TUI — 9 Languages
-
-Every screen, tab, button, and label in the `npx agentvibes` TUI is now fully translated:
-
-- **English, Spanish, French, German, Portuguese, Japanese, Korean, Chinese (Simplified), Italian**
-- Language selection on first launch (Screen 0 of the installer wizard)
-- Language sub-tab in Settings — switch language live without restarting
-- All tab bar labels, button text, footer hints, and status messages translated
-- BMAD tab and SSH Receiver tab fully localized
-- Per-language i18n files (`src/i18n/en.js`, `es.js`, `fr.js`, ...) with English fallback
-
-### 🪟 Windows Security & Bug Fixes
-
-- **Temp filenames** — All `Date.now()` temp filenames replaced with `randomUUID()` across JS and PowerShell (unpredictable, prevents temp file hijacking)
-- **Shell injection** — `execSync('which ...', { shell: true })` replaced with `spawnSync` (no shell expansion)
-- **Music player** — Hardcoded `ffplay` on Windows replaced with `detectMp3Player()` (respects user's installed player)
-- **Boolean coercion** — `isWindowsTerminal` now correctly returns `true/false` instead of leaking `WT_SESSION` UUID string
-- **Network mount detection** — `.match()` result properly coerced to boolean
-
-### 🎙️ Cross-Platform BMAD Speak
-
-BMAD (Build More Architect Dreams) is an AI multi-agent framework where specialized agents — Architect, PM, Developer, QA, and Analyst — collaborate to build software. With this release, every agent in a BMAD party mode session now speaks aloud with their own unique voice, personality, and music on Windows — making each role instantly recognizable.
-
-## 🐛 v4.5.1 — Patch Release
-
-**Release Date:** April 2026
-
-### Bug Fix
-
-- **Music tab preview** — Pressing Space on a track in the Music tab now plays correctly
-  when running `npx agentvibes` from a fresh directory. Previously, if `.claude/audio/tracks/`
-  didn't exist in the current working directory, the track list showed built-in tracks but
-  Space did nothing (the player was spawned against a non-existent path). Now falls back to
-  the package-bundled tracks directory automatically.
-
----
-
-- `bin/bmad-speak.js` — cross-platform entry point for BMAD agent speech
-- `.claude/hooks-windows/bmad-speak.ps1` — native Windows BMAD speak with per-agent personality routing
-
-### 🧪 Test Suite
-
-- 600 tests, 0 failures
-- Full cross-platform coverage (Windows path separators, chmod skip, provider file restore)
-
----
-
-## 🎉 v4.4.0 — "Full Platform Parity" Release
-
-**Release Date:** March 2026
-
-The biggest AgentVibes release since the TUI launched in v4.0. Three headline features: **BMAD Party Mode** gives every agent their own voice and music, **Windows Parity** brings full feature support to native Windows, and the **SSH Receiver** lets you hear your headless server speak on your local machine. Plus two rounds of adversarial code review hardening with 17+ security fixes.
-
-### 🪟 Windows Parity — First-Class Windows Support
-
-AgentVibes now runs natively on Windows with full feature parity:
-
-- **Background music on Windows** — New `background-music-manager.ps1`, full port of the Linux bash manager
-- **play-tts.ps1** reads `audio-effects.cfg` with per-agent track support, same as Linux
-- **ffmpeg auto-install** via `winget` during installer, PATH auto-refresh (no shell restart)
-- **Piper TTS native** — `piper.exe` resolved from `LOCALAPPDATA`, no WSL/bash needed
-- **Voice selection works** — `play-tts-windows-piper.ps1` reads `tts-voice.txt` set by TUI
-- **Multi-speaker models** — voices like libritts-high pass `--speaker` flag to Piper on Windows
-- **Windows SSH Receiver** — `setup-ssh-receiver.ps1` + `agentvibes-receiver.ps1` templates with hardened `sshd_config`
-- **TUI color contrast** — fixed for Windows Terminal (green focus, layout consistency)
-- **Music preview overlap** — switching tracks kills previous player via `taskkill`
-- **MCP Server** — strips `\r\n` line endings, accepts `[OK]` markers for PowerShell 5.1 compatibility
-
-### 🎭 BMAD Party Mode — Every Agent Has Its Own Voice
-
-When BMAD's party mode runs a multi-agent discussion, every agent speaks with their own individually configured voice, background music, reverb, and personality — making the Architect, PM, Developer, QA, and Analyst immediately recognizable.
-
-**Per-agent configuration:**
-- 🎙️ **Voice** — 914 voices, gender-aware auto-assign
-- 🎵 **Background Music** — Unique ambient track per agent (cinematic, lo-fi, jazz...)
-- 🎚️ **Music Volume** — Per-agent level, or bulk-set all at once
-- 🎛️ **Reverb** — none / room / hall / cathedral / studio
-- 💬 **Pretext** — Custom intro phrase ("Winston says:..." before every line)
-- 🎭 **Personality** — sarcastic, dramatic, pirate, cheerful, and more
-- 🔇 **No overlap** — speech lock serializes agents (mkdir-based, portable across platforms)
-- ✨ **Markdown stripped** — asterisks, emojis, and formatting removed before TTS
-
-### 🎛️ BMad Tab — Full Visual Agent Configurator
-
-New **BMad Tab** (`B` key) in `npx agentvibes` for managing every agent visually:
-
-- Voice, Gender, Provider, Reverb, Music, Vol, and Pretext columns
-- Voice names auto-beautified: `16Speakers::Rose_Ibex` → `Rose Ibex`
-- `Space` to preview with full profile (animated braille spinner while playing)
-- `Enter` to configure, `A` to auto-assign, `B` for bulk edit, `X` to reset
-
-### 🖥️ SSH Receiver Tab — Hear Your Headless Server
-
-New **Receiver Tab** streams TTS from voiceless remote servers to your local machine over TCP — perfect for cloud dev boxes, WSL2, and SSH sessions. Multi-provider TTS support, color-coded log columns, and platform-aware setup guide.
-
-### ⚡ Performance & UX
-- **TTS latency reduced ~1s** — batched 6 Node.js calls into 1, inotifywait queue worker, background cache cleanup
-- **ANSI colors restored** to TTS banner via `AGENTVIBES_WAV_OUTPATH` sidecar file
-- **Banner toggle** — hide TTS info without muting: `touch ~/.agentvibes/banner-disabled`
-- **`bin/agent-vibes` routes to blessed TUI** instead of old CLI installer
-- **Global hooks updated on upgrade** — `~/.claude/hooks/` synced automatically (#141)
-- **Markdown stripping in stop hook** — no more "asterisk asterisk" spoken aloud
-
-### 🔌 Windows MCP Parity — 27/27 Tools Working (#157)
-
-All MCP tools now work natively on Windows (previously 12 silently failed):
-
-- **6 new PowerShell scripts** — personality-manager, speed-manager, language-manager, learn-manager, verbosity-manager, clean-audio-cache
-- **Unified provider naming** — `piper` and `sapi` on all platforms (no more `windows-piper`/`windows-sapi`)
-- **replay command** added to voice-manager for Windows
-- **28 new tests** — script parity, effects round-trip, provider management, naming consistency
-- **Feature-platform matrix** — `docs/feature-platform-matrix.md` tracks all 85 features across 4 platforms
-- **Adversarial review** found 24 issues, 10 fixed in this release
-
-**HIGH bug fixes:**
-- ffmpeg stderr redirected to temp file instead of literal `"NUL"` file on disk
-- `AGENTVIBES_NO_PLAY` env var properly cleaned up on error/kill paths
-- `PIPER_SPEAKER` env var no longer leaks between voice switches
-- Provider config now uses project-local `.claude` directory
-- Text over-sanitization fixed — `$50 (USD)` no longer becomes `50 USD`
-
-### 🔧 Code Hardening (Adversarial Review)
-
-Three rounds of adversarial code review with 27+ fixes across HIGH and MEDIUM severity:
-
-**Round 3 (v4.4.0 — Windows parity):**
-- 6 CRITICAL missing script gaps closed (#157)
-- 4 HIGH severity bugs fixed (NUL redirect, env var leaks, config scope, speaker leak)
-- 3 MEDIUM fixes (text sanitization, error handling, provider naming)
-
-**Round 2 (v4.4.0 — agents tab):**
-- Temp file leak on piper error — cleanup added to error handler (#151)
-- Orphaned player process — generation counter gates player spawn (#152)
-- Duplicate piper resolution — extracted `_resolvePiperBin()` shared helper (#153)
-- Race condition in process handoff — double generation check (#154)
-- Voice reuse modulo bug — per-group round-robin counter (#155)
-- Ambiguous gender names removed from hardcoded map (#156)
-
-**Round 1 (earlier commits):**
-- 11 receiver-tab security findings addressed
-- All HIGH and MEDIUM issues from agents-tab review fixed
-- Portable speech lock (mkdir instead of flock)
-- Path traversal prevention, credential masking, resource cleanup throughout
-
-### 🛡️ Quality
-- 639 Node unit tests passing, 213 BATS tests passing
-- 28 new Windows-specific platform parity tests
-- Sonar quality gates validated across all changed files
-- JS syntax verified on all modules
-
-**Previous release:** [v4.0.0](https://github.com/paulpreibisch/AgentVibes/releases/tag/v4.0.0)
-
----
-
-## 🎉 v4.2 — "Party Mode" Release
-
-**Release Date:** March 2026
-
-This is the biggest AgentVibes release since the TUI launched in v4.0. Two headline features: **BMAD Party Mode** gives every agent their own voice and music, and the **SSH Receiver** lets you hear your headless server speak on your local machine.
-
-### 🤖 What is BMAD?
-
-The BMad Method (Build More Architect Dreams) is an AI-driven development framework module within the BMad Method Ecosystem that helps you build software through the whole process from ideation and planning all the way through agentic implementation. It provides specialized AI agents, guided workflows, and intelligent planning that adapts to your project's complexity, whether you're fixing a bug or building an enterprise platform.
-
-### 🎭 BMAD Party Mode — Every Agent Has Its Own Voice
-
-When BMAD's party mode runs a multi-agent discussion, every agent now speaks with their own individually configured voice, background music, reverb, and personality — making the Architect, PM, Developer, QA, and Analyst immediately recognizable the moment they speak.
-
-```bash
-/agent-vibes:bmad-party enable
-```
-
-**Per-agent configuration:**
-- 🎙️ **Voice** — 914 voices to choose from, auto-assigned gender-aware
-- 🎵 **Background Music** — Unique ambient track per agent (cinematic, lo-fi, jazz...)
-- 🎚️ **Music Volume** — Per-agent level, or bulk-set all at once
-- 🎛️ **Reverb** — none / room / hall / cathedral / studio
-- 💬 **Pretext** — Custom intro phrase ("Winston says:..." before every line)
-- 🎭 **Personality** — sarcastic, dramatic, pirate, cheerful, and more
-- 🔇 **No overlap** — speech lock held until audio fully completes
-- ✨ **Markdown stripped** — asterisks and formatting removed before TTS
-
-**Configuration stored in:** `~/.agentvibes/bmad-voice-map.json`
-
-### 🎛️ BMad Tab — Full Visual Agent Configurator
-
-New **BMad Tab** in `npx agentvibes` for managing every agent visually — built with the same polish as the Voices tab:
-
-```bash
-npx agentvibes   # Press B to open BMad Tab
-```
-
-The agent table shows **Voice, Gender, Provider, Reverb, Music, Vol, and Pretext** columns. Voice names are automatically beautified: `16Speakers::Rose_Ibex` → `Rose Ibex`, `en_US-kusal-medium` → `Kusal`.
-
-| Key | Action |
-|-----|--------|
-| `↑↓` / `jk` | Navigate agents |
-| `Space` | Preview agent with full profile (animated spinner while playing) |
-| `Enter` | Configure voice, music, volume, reverb, personality, pretext |
-| `A` | Auto-assign unique voices to all agents (gender-aware, no repeats) |
-| `B` | Bulk Edit — set music / volume / pretext / reverb for all agents |
-| `X` | Reset agent to defaults |
-
-**BMad Tab highlights:**
-- Inline row hints — navigate to any agent and see `[Space] Preview  [Enter] Configure` on the row
-- Animated `⠋⠙⠹⠸` braille spinner while audio plays
-- Gender & Provider columns — same metadata as the Voices tab
-
-### 🖥️ SSH Receiver Tab — Hear Your Headless Server
-
-New **Receiver Tab** streams TTS from voiceless remote servers to your local machine over TCP — perfect for cloud dev boxes (AWS, GCP, Azure), WSL2, and SSH sessions.
-
-```bash
-# On local machine: open TUI → Receiver tab → Start
-npx agentvibes
-
-# Remote server auto-detects the receiver and streams audio to you
-```
-
-### ⚡ TTS Latency Reduced ~1 Second
-
-- **Batched Node.js profile reads** — 6 `node -e` calls collapsed into 1 (~900ms saved per speech)
-- **inotifywait queue worker** — file-event-driven queue, no polling delay
-- **Background cache cleanup** — off the critical path every 10th call
-
-### 🎨 ANSI Colors Restored to Banner
-
-Full ANSI color in the TTS banner (gold voice, cyan reverb, traffic-light cache size), fixed via `AGENTVIBES_WAV_OUTPATH` sidecar file.
-
-### 🔕 Banner Toggle
-
-Hide TTS info banner without muting: `touch ~/.agentvibes/banner-disabled` or say "turn off the TTS banner" via MCP.
-
-### 🛡️ Security
-
-- Adversarial code review — 58 issues identified and addressed
-- Agent ID injection prevention, PID-scoped temp profile files
-- Env-var-based Node.js JSON reads (no shell interpolation)
-
-**Full Changelog**: https://github.com/paulpreibisch/AgentVibes/compare/v4.0.1...v4.2
-
----
-
-## ✨ v3.5.10 - Soprano Detection Fixes & Enhanced Installer Features
-
-**Release Date:** February 14, 2026
-
-### 🎯 Summary
-
-Production release combining critical bug fixes and new installer features. Fixed Soprano TTS detection for pipx installations (the core issue reported by users), resolved 5 execSync API misuse bugs that were breaking Python package detection, and eliminated 100+ lines of code duplication. Introduces new installer features: custom music track support with preview functionality, personality emoji mapping for better visual recognition, and pretext configuration allowing users to customize agent introductions.
-
-### ✨ Key Features & Fixes
-
-**🔧 Critical Bug Fixes:**
-- **Soprano TTS Detection:** Fixed detection when installed via pipx (was showing "not installed" despite working)
-- **execSync API Bugs:** Fixed 5 locations using incorrect API signature (array args with execSync)
-- **Code Duplication:** Eliminated 100+ lines of duplicate code between Soprano and Piper validators
-- **API Consistency:** All provider validation functions now return consistent response structures
-- **Python Package Detection:** Fixed broken Python pip detection that was silently failing
-
-**🎨 New Installer Features:**
-- **Custom Music Tracks:** Users can now upload and preview their own background music
-- **Personality Emojis:** Visual recognition mapping (😊 for none, 🎭 for dramatic, 💁 for sassy, etc.)
-- **Pretext Configuration:** Custom agent introductions (e.g., "FireBot: " prefix for all messages)
-- **Track Preview:** Audio preview with support for ffplay, sox, and mpv players
-
-**🛡️ Security & Quality:**
-- Improved path traversal protection in provider validation
-- Enhanced error handling and logging
-- Reduced code complexity by 29% through deduplication
-- Test coverage improved: 56.61% → 63.67%
-
-### 📊 Technical Details
-
-**Soprano Detection Improvements:**
-- Checks command in PATH first (most reliable for pipx)
-- Falls back to ~/.local/bin directory check
-- Checks pipx venv directory for installation
-- Final fallback to Python pip package detection
-- Consistent error messages showing all checked locations
-
-**Code Quality Metrics:**
-- File size reduced by 145 lines (-29%)
-- Code duplication eliminated (was ~100 lines)
-- Test coverage improved +7.06%
-- All 114 tests passing
-- Provider validator now 63.67% covered (up from 56.61%)
-
-**Installer Enhancements:**
-- Added personality emoji mapping (26 personalities)
-- Custom track upload with validation
-- File type restrictions (.mp3, .wav, .ogg, .m4a)
-- Registry storage in ~/.agentvibes/custom-tracks.json
-- Audio preview before finalizing selection
-
-### 🔒 Security Notes
-
-- All spawnSync calls now use correct array argument form
-- Path traversal prevention maintained in all operations
-- HOME injection protection via os.homedir()
-- No hardcoded credentials introduced
-- Input validation for file uploads and track selection
-
-### 🐛 Known Limitations
-
-- Audio preview requires ffplay, sox, or mpv (feature gracefully degrades if unavailable)
-- Custom track registry is stored locally per user
-- Some legacy bash scripts still lack strict mode (pre-existing, low risk)
-
-### 🙏 Acknowledgments
-
-This release includes fixes identified through adversarial code review, ensuring production-quality reliability and security alignment with CLAUDE.md standards.
-
----
-
-## 🛡️ v3.5.8 - Provider Validation Security & UX Improvements
-
-**Release Date:** February 12, 2026
-
-### 🎯 Summary
-
-Critical security and reliability update for provider detection. Fixes command injection vulnerabilities in validation code, prevents HOME directory injection attacks, and improves UX with explicit provider detection messaging. Soprano TTS installed via pipx is now correctly detected (previously showed "not installed" due to ES module import error). All 8 critical code review issues resolved with comprehensive security hardening and enhanced error reporting.
-
-### ✨ Key Improvements
-
-- **🔐 Security Fixes:** Fixed command injection vulnerability (template strings → array form), prevented HOME injection attacks, added path traversal protection
-- **✅ Provider Detection:** Soprano via pipx now correctly detected; added checkedLocations tracking for transparency
-- **💬 Better Messaging:** Explicit "Detected and selected!" confirmation; detailed error messages showing what was checked
-- **🧪 Test Coverage:** Enhanced tests verify actual detection values, not just types
-- **🐛 Debugging:** Added [DEBUG] logging for troubleshooting provider issues
-
-### 🔴 Critical Fixes
-
-1. **Command Injection Prevention** - All execSync calls now use array form (security: CLAUDE.md)
-2. **HOME Directory Injection** - Switched to os.homedir() instead of process.env.HOME
-3. **Path Traversal Protection** - Added path.resolve() validation for pipx venv directories
-
-### 🟡 Medium Fixes
-
-4. **Pipx Logic Improved** - Tracks checked locations even on success (transparency)
-5. **Silent Failures Eliminated** - Added [DEBUG] error logging for diagnostics
-6. **Test Quality Enhanced** - Verify message content, not just types
-7. **Documentation** - Added JSDoc comments explaining security-critical imports
-8. **Error Differentiation** - Better distinction between different failure types
-
-### 📊 Technical Impact
-
-- Soprano detection now works reliably for both pip and pipx installations
-- Reduced false negatives in provider validation
-- Enhanced security posture aligned with CLAUDE.md security mandates
-- Improved debuggability with explicit error messages
-
----
-
-## 🔧 v3.5.7 - CLI Fix: npx Command Output & Startup Hooks
-
-**Release Date:** February 12, 2026
-
-Fixes critical bug where `npx agent-vibes install` and other commands produced no output, making CLI unusable. Root cause: bin/agent-vibes used dynamic import without passing arguments to installer.js on local execution. Also removed broken hook configurations (pre_compact.py, notification.ts) that didn't exist and caused startup errors in Claude Code settings.
-
-### 🎯 What's Fixed
-
-- **npx agent-vibes now works** - `npx agent-vibes install`, `npx agent-vibes --help`, all commands produce proper output
-- **Startup hook errors gone** - Removed non-existent hook references from settings.json (pre_compact.py, notification.ts)
-- **CLI execution proper** - Both npx and local execution now use execFileSync with proper argument passing
-
-### 🚀 Technical Details
-
-**Before v3.5.7:**
-```javascript
-// bin/agent-vibes (local execution path)
-import('../src/installer.js');  // ❌ No args, doesn't await
-```
-
-**After v3.5.7:**
-```javascript
-// bin/agent-vibes (all execution paths)
-execFileSync('node', [installerPath, ...arguments_], {
-  stdio: 'inherit',
-  cwd: isNpxExecution ? path.dirname(__dirname) : process.cwd(),
-});  // ✅ Passes args, proper I/O
-```
-
----
-
-## 🔧 v3.5.6 - Bug Fix: Bash Hook Parameter Handling
-
-**Release Date:** February 11, 2026
-
-Fixes critical regression in v3.5.5 where bash hooks failed with unbound variable errors when called with optional parameters under strict mode. Affects `play-tts.sh` and `provider-manager.sh`.
-
----
-
-## 📦 v3.5.5 - Native Windows Support: Soprano, Piper & SAPI Providers
-
-**Release Date:** February 12, 2026
-
-### 🎯 Why v3.5.5?
-
-v3.5.5 brings **native Windows support** to AgentVibes with a full-featured PowerShell installer and three TTS providers. Windows users no longer need WSL - AgentVibes runs natively with Soprano (neural), Piper (offline neural), or Windows SAPI (zero-setup) voices. The installer also adds **background music selection** (16 genre tracks), **reverb/audio effects** (via ffmpeg aecho), and **verbosity control** for the TTS experience.
-
-### 🚀 Key Highlights
-
-#### 🖥️ Native Windows TTS (NEW!)
-- **3 providers**: Soprano (ultra-fast neural), Piper (offline neural), Windows SAPI (built-in)
-- **Beautiful PowerShell installer** with figlet banner and interactive setup
-- **8 hook scripts** for complete TTS functionality on Windows
-- **MCP server** auto-resolves `.sh` to `.ps1` on Windows
-- **46 Windows-specific unit tests** with full coverage
-
-#### 🎵 Background Music Selection
-- **16 genre tracks**: Flamenco, Bachata, Bossa Nova, City Pop, Chillwave, and more
-- **Interactive picker** in the installer with descriptions
-- **ffmpeg mixing**: 2s intro, voice over music, 2s fade-out outro
-
-#### 🎛️ Reverb / Audio Effects
-- **5 reverb levels**: Off, Light, Medium, Heavy, Cathedral
-- **ffmpeg aecho filter** (no SOX dependency on Windows)
-- Applied before background music mixing for clean layering
-
-#### 🔊 Verbosity Control
-- **3 levels**: High (full reasoning), Medium (key updates), Low (essential only)
-- Integrates with session-start-tts.ps1 protocol instructions
-
-### 🤖 AI Summary
-
-AgentVibes v3.5.5 delivers native Windows support with a polished PowerShell installer offering three TTS providers (Soprano neural, Piper offline, Windows SAPI), background music selection from 16 genre tracks, reverb effects via ffmpeg aecho filter, and verbosity control. The release includes 8 Windows hook scripts, MCP server platform detection for automatic .sh-to-.ps1 resolution, and 46 new unit tests. Security hardening adds path traversal prevention with regex allowlisting and path containment checks, reverb config allowlist validation, and strict mode compliance across all scripts. Cross-platform test fixes ensure the full 93-test suite passes on both Windows and Unix.
-
----
-
-## ✨ New Features
-
-### Native Windows TTS
-- Full PowerShell installer (`setup-windows.ps1`) with figlet banner and interactive UX
-- Soprano provider (`play-tts-soprano.ps1`) with Gradio WebUI integration
-- Piper provider (`play-tts-windows-piper.ps1`) with auto-download of voices from HuggingFace
-- Windows SAPI provider (`play-tts-windows-sapi.ps1`) with zero-setup built-in voices
-- TTS router (`play-tts.ps1`) with mute support, background music mixing, and reverb
-- Provider manager, voice manager, audio cache utils, and session-start hook scripts
-- MCP server `.sh` to `.ps1` auto-resolution on Windows
-
-### Installer Enhancements
-- Background music selection with 16 genre tracks and interactive picker
-- Reverb/audio effects selection (Off/Light/Medium/Heavy/Cathedral)
-- Verbosity control (High/Medium/Low) for TTS protocol instructions
-- Updated completion screen showing all 4 settings (provider, background, reverb, verbosity)
-
----
-
-## 🐛 Bug Fixes
-
-### Security Fixes
-- Fix path traversal in background music config reader (regex allowlist + path containment)
-- Add allowlist validation for reverb-level.txt config (prevent invalid values)
-- Add `set -euo pipefail` strict mode to `play-tts.sh` for Sonar compliance
-
-### Cross-Platform Fixes
-- Fix self-copy error when setup-windows.ps1 runs from project root
-- Fix test executable permission checks on Windows (skip Unix mode bits)
-- Fix test path separator comparison in uninstall test (use `path.join` not hardcoded `/`)
-
----
-
-## 🏗️ Improvements
-
-### Code Quality
-- Reverb config uses switch-as-allowlist pattern - file content never flows into commands
-- All SoundPlayer instances wrapped in try/finally for resource disposal
-- Environment variable cleanup (`AGENTVIBES_NO_PLAY`) on all exit paths
-- Input validation with regex + range checks for all installer prompts
-
-### Testing
-- 46 new Windows-specific unit tests (hook scripts, providers, security, encoding)
-- 3 cross-platform test fixes for Windows compatibility
-- Full suite: 93 Node tests passing on Windows
-
----
-
-## 📊 Statistics
-
-- **7 commits** since v3.4.1
-- **3,769 lines added**, 211 removed across 24 files
-- **9 new PowerShell scripts** for Windows TTS
-- **93 tests passing** (46 Windows + 47 cross-platform)
-- **24/24 Sonar quality gates** passing
-- **Security score**: All path traversal and injection vectors reviewed
-
----
-
-## 🔧 Technical Details
-
-### Files Added
-- `.claude/hooks-windows/play-tts.ps1`: TTS router with reverb and background music
-- `.claude/hooks-windows/play-tts-soprano.ps1`: Soprano neural TTS provider
-- `.claude/hooks-windows/play-tts-windows-piper.ps1`: Piper offline TTS provider
-- `.claude/hooks-windows/play-tts-windows-sapi.ps1`: Windows SAPI built-in voices
-- `.claude/hooks-windows/provider-manager.ps1`: Provider switching
-- `.claude/hooks-windows/voice-manager-windows.ps1`: Voice browsing and selection
-- `.claude/hooks-windows/audio-cache-utils.ps1`: Cache management
-- `.claude/hooks-windows/session-start-tts.ps1`: Auto-activates TTS on Claude start
-- `setup-windows.ps1`: Full Windows installer with 4 interactive sections
-- `test/unit/windows-tts.test.js`: 46 Windows-specific unit tests
-
-### Breaking Changes
-None - all changes are backward compatible. Existing Unix/macOS installations are unaffected.
-
----
-
-## 🎓 Migration Notes
-
-### For New Windows Users
-1. Run `npx agentvibes install` (Node.js) or `.\setup-windows.ps1` (PowerShell)
-2. Follow the interactive setup
-3. Choose provider (Soprano, Piper, or SAPI)
-4. Select background music, reverb, and verbosity
-5. TTS works automatically in Claude Code sessions
-
-### For Existing Unix/macOS Users
-- No changes required - your setup continues working
-- All Unix bash hooks remain untouched
-- Only `play-tts.sh` gained `set -euo pipefail` (strict mode)
-
----
-
-## 🙏 Acknowledgments
-
-### Project Lead
-- **[@paulpreibisch](https://github.com/paulpreibisch)** (Paul Preibisch) — Creator and maintainer of AgentVibes
-
-### Community Contributors
-- **[@nathanchase](https://github.com/nathanchase)** — For contributing the Soprano TTS provider in v3.4.0, whose ultra-fast neural engine is now one of the three Windows-native providers
-- **[@alexeyv](https://github.com/alexeyv)** — For suggesting native Windows support and recommending Windows SAPI as a zero-dependency provider
-- **[@bmadcode](https://github.com/bmadcode)** (Brian Madison) — Creator of the [BMAD Method](https://github.com/bmadcode/BMAD-METHOD), used daily for planning and building AgentVibes features
-
-### Quality Assurance
-- **Adversarial Security Review**: Path traversal, injection, and resource disposal all validated
-- **Testing**: 93/93 tests passing (100% suite coverage)
-- **Quality Gates**: 24/24 Sonar requirements validated
-- **Co-Authored-By**: Claude Opus 4.6
-
----
-
-**Full Changelog**: https://github.com/paulpreibisch/AgentVibes/compare/v3.4.1...v3.5.5
-
----
-
-## 📦 v3.4.0 - Soprano TTS, Security Hardening & Environment Intelligence
-
-**Release Date:** February 10, 2026
-
-### 🎯 Why v3.4.0?
-
-v3.4.0 introduces **Soprano TTS** - an ultra-fast neural TTS provider with GPU acceleration, comprehensive **security hardening** across the codebase, and **intelligent environment detection** that recognizes PulseAudio tunnels for remote audio scenarios.
-
-### 🚀 Key Highlights
-
-#### ⚡ Soprano TTS Provider (NEW!)
-- **80M parameter neural model** with premium female English voice
-- **20x CPU speed** (vs Piper), **2000x GPU speed** with CUDA
-- **3 synthesis modes**: WebUI (Gradio), API (OpenAI-compatible), CLI (fallback)
-- **Auto-detection**: Checks for running Gradio server, falls back gracefully
-- **<1GB memory footprint** - perfect for low-RAM systems
-- **Provider-aware voice management**: Auto-selects single voice, shows model specs
-- **Thanks to [@nathanchase](https://github.com/nathanchase)** for this contribution! ([see acknowledgments](#-acknowledgments))
-
-#### 🛡️ Security Hardening (9.5/10 Score)
-- **Timeouts on system commands**: Prevents installer hangs (nvidia-smi, sysctl, meminfo)
-- **Bounds checking**: Validates array access before parsing system output
-- **NaN validation**: Prevents crashes from malformed memory/GPU detection
-- **Case-insensitive checks**: PulseAudio tunnel detection handles TCP: and tcp:
-- **Code duplication eliminated**: Extracted PulseAudio helper function (DRY)
-
-#### 🌐 Environment Intelligence
-- **PulseAudio tunnel detection**: Recognizes `PULSE_SERVER=tcp:*` as working audio
-- **Context-aware messaging**:
-  - "🌐 PulseAudio Tunnel Detected!" for SSH + tunnel setups
-  - "🔊 Audio Output Detected!" for local speakers
-  - Distinguishes local/tunnel/hybrid configurations
-- **Smart environment classification**:
-  - DESKTOP: Local audio OR active PulseAudio tunnel
-  - VOICELESS: No audio AND no tunnel
-  - PHONE: Termux/Android devices
-
-#### 🎤 Installer Enhancements
-- **Provider-aware voice pages**: Soprano shows model specs, Piper shows 50+ voices
-- **Auto-selection logic**: Soprano (1 voice) auto-selects, no manual choice needed
-- **GPU-based recommendations**: "Your GPU will run Soprano 2000x faster!"
-- **RAM-based suggestions**: Low memory systems see "Soprano uses <1GB" message
-- **Better RAM display**: Shows "512MB" instead of "0GB" for sub-1GB systems
-
-### 🤖 AI Summary
-
-AgentVibes v3.4.0 brings Soprano TTS - an 80M parameter neural provider offering 20x CPU and 2000x GPU acceleration with sub-1GB memory footprint - plus comprehensive security hardening (timeouts, bounds checking, NaN validation) and intelligent environment detection that recognizes PulseAudio tunnels as working audio for remote scenarios. The enhanced installer provides context-aware messaging distinguishing local speakers from SSH tunnels, GPU-based provider recommendations (Soprano for CUDA users, macOS Say for Apple, Piper for versatility), and provider-specific voice pages that auto-select Soprano's single voice while showcasing model specifications. This release achieves a 9.5/10 security score through systematic defensive programming, making AgentVibes production-ready for enterprise deployments while expanding TTS provider options for diverse hardware configurations.
-
----
-
-## ✨ New Features
-
-### Soprano TTS Provider
-- Add Soprano TTS provider script with 3 synthesis modes (WebUI, API, CLI) (#95)
-- Integrate Soprano into TTS router and provider manager
-- Add soprano-gradio-synth.py helper for WebUI/SSE protocol
-- Provider-aware voice selection page with model specifications
-- Auto-select single Soprano voice with performance details
-
-### Installer Intelligence
-- Add `detectSystemCapabilities()` for GPU/RAM detection
-- Add `hasPulseAudioTunnel()` helper function
-- Context-aware audio detection messaging (tunnel vs local)
-- GPU-based provider ordering (Soprano first for CUDA users)
-- RAM-based recommendations (<4GB systems see Soprano first)
-- Provider-specific intro messages (Soprano vs Piper vs macOS)
-
-### Environment Detection
-- PulseAudio tunnel recognition via PULSE_SERVER env var
-- Case-insensitive TCP protocol detection
-- Smart DESKTOP classification (local audio OR tunnel)
-- Improved VOICELESS detection (no audio AND no tunnel)
-
----
-
-## 🐛 Bug Fixes
-
-### Security Fixes
-- Add 5s timeout to nvidia-smi to prevent GPU detection hangs
-- Add 3s timeout to sysctl/meminfo to prevent memory detection hangs
-- Add bounds checking before parsing sysctl output (macOS)
-- Add bounds checking before parsing /proc/meminfo (Linux)
-- Add NaN validation for parseInt() memory size parsing
-- Fix case sensitivity in PULSE_SERVER detection (handles TCP: and tcp:)
-
-### Test Fixes
-- Fix provider-manager test #90: Add soprano and ssh-remote to cleanup list
-- Ensure zero-provider edge case properly simulates empty state
-
-### User Experience
-- Fix RAM display for <1GB systems (show "512MB" not "0GB")
-- Fix PulseAudio selection triggering wrong setup flow
-- Separate PulseAudio tunnel setup from SSH receiver setup
-
----
-
-## 🏗️ Improvements
-
-### Code Quality
-- Extract PulseAudio detection to helper function (DRY principle)
-- Implement system capabilities caching (eliminates duplicate calls)
-- Add comprehensive error handling in detectSystemCapabilities()
-- Improve code comments for security-critical sections
-
-### Performance
-- Cache system detection results (prevents duplicate nvidia-smi calls)
-- Add timeouts to prevent indefinite hangs
-- Optimize provider detection with early returns
-
-### Documentation
-- Add comprehensive commit message documenting all changes
-- Document security improvements (timeouts, bounds checking, NaN validation)
-- Explain PulseAudio tunnel detection architecture
-- Detail environment classification logic
-
----
-
-## 📊 Statistics
-
-- **91 commits** since v3.3.0
-- **817 lines added** in merge to master
-- **6 files modified** in core integration
-- **260 tests passing** (213 BATS + 47 Node)
-- **Security score**: 7.5/10 → 9.5/10
-- **Test coverage**: 100% pass rate
-
----
-
-## 🔧 Technical Details
-
-### Files Modified
-- `src/installer.js`: +335 lines (security fixes, environment detection, Soprano integration)
-- `test/unit/provider-manager.bats`: +4 lines (fix edge case test)
-- `.claude/hooks/play-tts-soprano.sh`: +320 lines (new provider)
-- `.claude/hooks/soprano-gradio-synth.py`: +139 lines (new helper)
-- `.claude/hooks/provider-manager.sh`: +17 lines (Soprano support)
-- `.claude/hooks/play-tts.sh`: +6 lines (route to Soprano)
-
-### Breaking Changes
-None - all changes are backward compatible.
-
-### Dependencies
-- **New**: `soprano-tts` (Python package, optional)
-- **Recommended**: CUDA-capable GPU for 2000x speedup (optional)
-- **Compatible**: Works on CPU-only systems (20x vs Piper)
-
----
-
-## 🎓 Migration Notes
-
-### For New Users
-1. Run `npx agentvibes install`
-2. Installer auto-detects your hardware (GPU, RAM, platform)
-3. Soprano appears as option if you have working audio
-4. Select Soprano for ultra-fast TTS with GPU acceleration
-
-### For Existing Users
-1. Update: `npx agentvibes update`
-2. Switch provider: `/agent-vibes:provider switch soprano`
-3. Test: `/agent-vibes:sample soprano-default`
-4. Optionally install soprano-tts: `pip install soprano-tts`
-
-### PulseAudio Tunnel Users
-- Installer now auto-detects your tunnel configuration
-- Shows "🌐 PulseAudio Tunnel Detected!" instead of "speakers"
-- Provides DESKTOP mode options (Soprano, Piper, macOS Say)
-- No manual configuration needed
-
----
-
-## 🙏 Acknowledgments
-
-### Special Thanks
-
-**🎉 [@nathanchase](https://github.com/nathanchase)** - For contributing the Soprano TTS Provider integration (PR #95)! Nathan's work brings ultra-fast neural TTS with GPU acceleration to AgentVibes, offering 20x CPU and 2000x GPU performance improvements. The comprehensive integration includes WebUI, API, and CLI synthesis modes with intelligent auto-detection and graceful fallback. Thank you for this outstanding contribution! 🚀
-
-### Quality Assurance
-
-- **Security Review**: Adversarial code review achieved 9.5/10 score
-- **Testing**: All 260 tests pass (100% suite coverage)
-- **Quality Gates**: All Sonar requirements validated
-- **Co-Authored-By**: Claude Sonnet 4.5
-
----
-
-## 📚 Additional Resources
-
-- [Soprano TTS Documentation](https://github.com/paulpreibisch/AgentVibes/blob/master/docs/providers.md#soprano-tts)
-- [PulseAudio Tunnel Setup](https://github.com/paulpreibisch/AgentVibes/blob/master/docs/SSH_REMOTE_SETUP.md)
-- [Security Hardening Guide](https://github.com/paulpreibisch/AgentVibes/blob/master/docs/security-hardening-guide.md)
-- [Provider Comparison](https://github.com/paulpreibisch/AgentVibes/blob/master/docs/providers.md)
-
----
-
-**Full Changelog**: https://github.com/paulpreibisch/AgentVibes/compare/v3.3.0...v3.4.0
-
----
-
-## 📦 v3.3.0 - Remote TTS, Smart Installer, OpenClaw Receiver & Cache Management
-
-**Release Date:** February 5, 2026
-
-### 🎯 Why v3.3.0?
-
-v3.3.0 transforms AgentVibes into a **universal TTS platform** for any environment:
-
-- **SSH-Remote Provider** - Generate TTS on servers, receive audio on your phone/computer
-- **Termux/Android Support** - Native Piper TTS on mobile devices
-- **OpenClaw Integration** - Turn voiceless servers into Siri-like conversational AI
-- **AgentVibes Receiver** - Receive and play audio from remote servers on your device
-- **Smart Installer** - Auto-detects your environment (voiceless, GUI, Termux, SSH)
-- **Intelligent Cache Management** - Real-time tracking and auto-cleanup prevents disk bloat
-
-#### 🌐 Real-World Use Case: OpenClaw + AgentVibes Receiver
-
-You deploy OpenClaw on a voiceless Mac mini (or remote server) where users message you via WhatsApp, Telegram, or Discord. With v3.3.0:
-
-**Before AgentVibes Receiver:**
-- User messages: "Tell me a joke"
-- Mac mini processes request
-- Text response appears in chat
-- 😞 No audio - silent experience
-
-**After AgentVibes Receiver:**
-1. **Install AgentVibes** on your Mac mini (or remote server)
-2. **Install AgentVibes Receiver** on your phone/iPad/laptop
-3. **Connect via Tailscale** (one-time setup)
-4. **User messages:** "Tell me a joke"
-5. **Mac mini generates TTS** with your configured voice
-6. **Audio streams to your device** via SSH tunnel
-7. **Your speakers play:** 🔊 "Why did the AI go to school? To improve its learning model!"
-8. **User in WhatsApp also hears** the audio playing (Siri-like experience)
-
-Result: OpenClaw transforms from **silent text AI** → **Conversational AI with voice**
-
-Perfect for:
-- 🖥️ Mac mini with OpenClaw
-- 🖥️ Remote servers (AWS, DigitalOcean, Linode)
-- 🏗️ Container deployments (Docker)
-- 🔧 WSL (Windows Subsystem for Linux)
-- 📱 Any voiceless environment needing audio
-
-### 🤖 AI Summary
-
-AgentVibes v3.3.0 unleashes the platform across new frontiers: remote servers via SSH-PulseAudio tunneling, Android/Termux environments with native Piper support, and OpenClaw (formerly Clawdbot) multi-agent orchestration. The redesigned smart installer detects your environment (voiceless, GUI, SSH, Termux) and shows only relevant options, plus optional BMAD personality injection for advanced users. Every TTS output now displays real-time cache metrics (file count/size with dynamic colors) plus intelligent size-based auto-cleanup that deletes oldest files when the cache exceeds threshold. The release includes comprehensive TTS queue management to prevent audio overlap, audio effects support across all providers, and full MCP tool integration for programmatic control. This release transforms AgentVibes into a universal TTS platform.
-
-**Key Highlights:**
-- 🌍 **SSH-Remote TTS** - Remote device playback via PulseAudio tunneling (servers, containers, WSL)
-- 📱 **Android/Termux Support** - Native Piper TTS on Android with termux-media-player integration
-- 🤖 **OpenClaw Receiver** (formerly Clawdbot) - AgentVibes Receiver for receiving TTS from voiceless servers
-- 🧠 **Smart Installer** - Voiceless environment detection + personality injection for BMAD
-- 📊 **Real-Time Cache Tracking** - File count and size on every output with dynamic colors
-- 🧹 **Intelligent Auto-Cleanup** - Size-based threshold (15MB default) prevents storage bloat
-- 🎵 **Queue Management** - Prevents TTS audio overlap via centralized queue system
-- ⚙️ **Audio Effects** - Full support across SSH-remote, Termux-ssh, and local providers
-- 📁 **Uninstall Command** - Comprehensive cleanup with full documentation
-- ✅ **96 Commits** - Massive feature expansion with 213 BATS tests passing
-
-### ✨ New Features
-
-#### 🌍 Remote SSH TTS Support
-
-**SSH-Remote Provider:**
-- Play TTS on remote servers via SSH + PulseAudio tunneling
-- Zero-dependency for audio output (uses PulseAudio network tunnel)
-- Perfect for deployed Claude Code on servers, containers, WSL
-- Auto-configuration of PulseAudio TCP module
-- Fallback to local playback if SSH unavailable
-- Full compatibility with all voice selection and audio effects
-
-**SSH-PulseAudio Integration:**
-- Automatic SSH connection detection and setup
-- Secure TCP tunnel for audio stream transmission
-- Support for both interactive and batch TTS operations
-- Persistent audio configuration per SSH session
-
-#### 📱 Android/Termux Support
-
-**Termux-SSH Provider:**
-- Native Piper TTS on Android via Termux environment
-- Uses termux-media-player for audio playback
-- Full voice selection and effects support
-- Automatic temp directory detection
-- Integration with Tailscale for secure remote access
-- Comprehensive setup guide with QR codes
-
-**Android Installation:**
-- Self-contained Termux installer script
-- One-command setup: `curl https://agentvibes.org/install-android | bash`
-- Automatic dependency detection and installation
-- Piper voice download management
-
-#### 🎙️ OpenClaw Integration & AgentVibes Receiver
-
-**What is AgentVibes Receiver?**
-
-AgentVibes Receiver is a **lightweight audio client** that receives and plays TTS audio from remote servers where OpenClaw is installed. It runs on your phone, tablet, or personal computer and connects to voiceless servers via SSH tunnel.
-
-**The Problem It Solves:**
-- OpenClaw running on Mac mini/remote server has no audio output
-- Users message via WhatsApp/Telegram/Discord - get text responses only
-- 😞 No voice = Less engaging AI experience
-
-**AgentVibes Receiver Solution:**
-1. **Lightweight client** runs on your device (phone/tablet/laptop)
-2. **SSH tunnel** securely connects to your voiceless server
-3. **Audio streams** from server to your device via PulseAudio
-4. **Auto-plays** on your speakers when OpenClaw responds
-5. **Siri-like experience** - Text + Voice in one flow
-
-**How It Works:**
-
-```
-┌──────────────────────────────┐
-│ Your Mac mini / Server       │
-│ (OpenClaw + AgentVibes)      │
-│ ├─ No audio output           │
-│ ├─ Generates TTS             │
-│ └─ Sends via SSH tunnel      │
-└──────────────────────────────┘
-            ↓ SSH Tunnel (encrypted)
-┌──────────────────────────────┐
-│ Your Phone / Laptop          │
-│ (AgentVibes Receiver)        │
-│ ├─ Receives audio stream     │
-│ ├─ Plays on speakers         │
-│ └─ You hear OpenClaw speak   │
-└──────────────────────────────┘
-```
-
-**Example Flow:**
-```
-WhatsApp: "Tell me a joke"
-        ↓
-Mac mini: Processes with Claude
-        ↓
-Generates TTS: "Why did the AI... [audio file]"
-        ↓ SSH tunnel
-Your Phone: Plays audio 🔊
-        ↓
-You hear: "Why did the AI go to school?"
-```
-
-**AgentVibes Receiver Features:**
-- **One-Time Setup** - Pair with server via SSH key
-- **Automatic Connection** - Reconnects if interrupted
-- **Real-Time Streaming** - Low latency audio playback
-- **SSH Encryption** - Secure tunnel for audio
-- **Tailscale Support** - Easy VPN for remote servers
-- **Multiple Servers** - Connect to different OpenClaw instances
-- **Voice Control** - Full voice selection on the server side
-- **Cache Metrics** - Monitor audio generation and cleanup
-
-**OpenClaw Skill Integration:**
-- Installed automatically with AgentVibes on OpenClaw server
-- Full feature access:
-  - Voice selection (50+ voices)
-  - Personality/sentiment (sarcastic, flirty, etc.)
-  - Audio effects (reverb, echo, pitch)
-  - Speech speed (0.5x - 3.0x)
-  - Language translation (speak in different languages)
-  - Real-time cache tracking
-  - Automatic cleanup of old audio files
-
-#### 🧠 Smart Installer Enhancements
-
-**Voiceless Environment Detection:**
-- Auto-detects if GUI audio is unavailable (headless servers, containers)
-- Offers SSH-remote TTS as alternative for voiceless environments
-- Prevents installation of unnecessary audio dependencies
-
-**Personality Injection (BMAD):**
-- Interactive prompt during install for BMAD users
-- Optional TTS personality configuration
-- Sentiment/personality selection built into setup flow
-- Skipped for non-BMAD environments
-
-**Provider Auto-Selection:**
-- Intelligent detection of available providers:
-  - macOS Say (macOS systems)
-  - Piper TTS (all systems)
-  - SSH-remote (if SSH available)
-  - Termux-ssh (Android/Termux)
-- Shows only relevant providers in installation
-
-**Better UX:**
-- Clear descriptions of each provider
-- Setup URLs for complex providers (Tailscale)
-- Comprehensive help text for each option
-- Git log integration for recent changes
-
-#### 📊 Real-Time TTS Cache Tracking & Intelligent Auto-Cleanup
-
-**Why Cache Management Matters:**
-- TTS audio files accumulate quickly
-- Server deployments can run out of disk space silently
-- Users have no visibility into cache size or cleanup status
-- Manual cleanup is inconvenient and error-prone
-
-**Cache Display on Every Output:**
-Every time you generate TTS, you see real-time cache metrics:
-```
-💾 Saved to: /home/user/.claude/audio/tts-1770274925.wav 📦 28 20.9MB 🧹[15mb]
-```
-
-What you see:
-- 💾 **Full path** - Clickable file for replay or sharing
-- 📦 **28** - File count in cache
-- **20.9MB** - Total cache size (color-coded):
-  - 🟢 Green: <500MB
-  - 🟡 Yellow: 500MB-3GB
-  - 🔴 Red: >3GB
-- 🧹 **[15mb]** - Auto-cleanup threshold
-
-**Intelligent Size-Based Auto-Cleanup:**
-- Deletes oldest files when cache exceeds threshold (default: 15MB)
-- Silent operation (no blocking prompts)
-- Write-lock protection prevents conflicts with TTS generation
-- Respects active TTS (won't delete while generating)
-
-**Configuration:**
-```bash
-# Per-project threshold override
-echo "50" > .claude/tts-auto-clean-threshold.txt  # 50MB limit
-
-# Or disable cleanup
-echo "0" > .claude/tts-auto-clean-threshold.txt   # Disable
-```
-
-**Manual Cleanup:**
-```bash
-# Non-interactive cleanup
-/agent-vibes:clean
-
-# Or programmatically via MCP
-await agent_vibes.clean_audio_cache()
-```
-
-#### 🎵 TTS Queue Management
-
-**Overlap Prevention:**
-- Centralized queue system for TTS operations
-- Prevents simultaneous audio playback
-- Critical for Clawdbot multi-agent scenarios
-- Atomic queue operations ensure consistency
-
-**Queue Integration:**
-- Automatic in OpenClaw Receiver
-- Optional in standalone environments
-- Fallback to sequential playback
-
-#### ⚙️ Audio Effects Across All Providers
-
-**Effects Support:**
-- Reverb, echo, pitch, EQ available
-- SSH-remote provider: Full effects support
-- Termux-ssh provider: Full effects support
-- All local providers: Unchanged effects behavior
-
-**Configuration:**
-- Per-session override via environment variables
-- Project-local settings via config files
-- Persistent across TTS operations
-
-#### 📁 Comprehensive Uninstall Command
-
-**`/agent-vibes:uninstall` Skill:**
-- Complete removal of AgentVibes and dependencies
-- Interactive prompts for user confirmation
-- Option to preserve configuration
-- Detailed removal logs
-- Full documentation included
-
-### 🐛 Bug Fixes
-
-- **TTS Overlap** - Fixed audio overlapping via queue management
-- **Termux Audio** - Proper detection and use of termux-media-player
-- **SSH Detection** - Improved SSH environment detection logic
-- **Race Conditions** - Write-lock mechanism prevents cleanup conflicts
-- **Temp Directory** - Proper Termux temp directory handling
-- **Color Codes** - Fixed GOLD color (256-color \033[38;5;226m)
-- **Stat Compatibility** - BSD/GNU stat detection with proper output suppression
-- **Syntax Validation** - Fixed installer.js syntax errors
-- **Coverage Testing** - Proper coverage file generation for CI/CD
-
-### 🔒 Security & Quality
-
-- **No Hardcoded Credentials** - All secure operations use environment variables
-- **SSH Safety** - Secure PulseAudio tunnel authentication
-- **Atomic Operations** - Queue and receiver use atomic file operations
-- **Input Validation** - All external inputs validated
-- **Pre-existing Limitations** - TTS scripts lack `set -euo pipefail` (pre-existing)
-- **Sonar Compliance** - Security hotspots resolved, quality gates passing
-- **Test Coverage** - 213 BATS tests + 47 Node unit tests
-
-### ✅ Testing & Validation
-
-- **213 BATS Tests** - Core functionality validation
-- **47 Node Tests** - JavaScript/installer validation
-- **Cross-Platform** - Piper, macOS, SSH-remote, Termux-ssh
-- **Environment Tests** - Voiceless, GUI, SSH, Termux detection
-- **Audio Effects** - All providers tested
-- **Backwards Compatible** - No breaking changes to existing code
-
----
-
-## 📦 v3.2.0 - Clawdbot Integration: AI Assistants on Any Messenger
-
-**Release Date:** January 27, 2026
-
-### 🎯 Why v3.2.0?
-
-This minor release adds **native Clawdbot integration** to AgentVibes, bringing professional TTS to the revolutionary AI assistant you can access via any instant messenger. Clawdbot connects Claude AI to WhatsApp, Telegram, Discord, and more—and now with AgentVibes, your Clawdbot can speak with 50+ professional voices in 30+ languages. This release also includes SonarCloud quality gate improvements and CI/CD workflow enhancements.
-
-### 🤖 AI Summary
-
-AgentVibes v3.2.0 introduces seamless integration with Clawdbot, the revolutionary AI assistant accessible via any instant messenger. With this release, Clawdbot users get professional TTS with 50+ voices, remote SSH audio support for server deployments, and zero-configuration setup—just install AgentVibes and the Clawdbot skill is ready. The release also includes quality improvements: SonarCloud workflow fixes, enhanced documentation for disabling quality gate checks, and improved test coverage validation.
-
-**Key Highlights:**
-- 🤖 **Clawdbot Integration** - Native TTS support for Clawdbot AI assistant framework
-- 💬 **Messenger Platforms** - Works with WhatsApp, Telegram, Discord via Clawdbot
-- 🔊 **Remote SSH Audio** - Perfect for Clawdbot on remote servers with PulseAudio tunneling
-- 📦 **Simple Install** - Just `npx agentvibes install` and it works
-- 🛡️ **SonarCloud Fixes** - Quality gate workflow improvements and documentation
-- ✅ **Full Test Coverage** - All 213 BATS + 47 Node tests passing
-
-### ✨ New Features
-
-**Clawdbot Skill (`.clawdbot/`):**
-- New `.clawdbot/` directory with skill integration files
-- `README.md` - Clawdbot integration overview and setup guide
-- `skill/SKILL.md` - Comprehensive skill documentation with voice selection, background music, effects, personalities, and remote SSH audio setup
-- Automatically distributed via npm package
-- Zero-configuration when AgentVibes is installed
-
-**README Updates:**
-- Added "🤖 Clawdbot Integration" section with full documentation
-- Updated header to include Clawdbot alongside Claude Code, Claude Desktop, and Warp Terminal
-- Added Clawdbot to Quick Links table
-- Clawdbot description: "A revolutionary AI assistant you can access via any instant messenger"
-- Website link: https://clawd.bot
-
-**package.json Updates:**
-- Added `.clawdbot/` to npm files array for distribution
-- Added `clawdbot` to keywords for npm discoverability
-- Updated description to mention Clawdbot support
-
-### 🐛 Bug Fixes
-
-- **SonarCloud Quality Gate** - Disabled quality gate status reporting to GitHub to prevent false CI failures
-- **Coverage File Generation** - Ensured coverage file is generated before SonarCloud scan
-- **CLI Test Coverage** - Added CLI tests and excluded CLI entry point from coverage requirements
-- **macOS Runner** - Removed macos-15-large runner to avoid GitHub billing limits
-- **Piper Voice Test** - Updated installation test to match current voice list
-
-### 📚 Documentation
-
-- Added step-by-step SonarCloud dashboard configuration guide
-- Added guide to disable SonarCloud GitHub App checks
-- Comprehensive Clawdbot integration documentation with SSH audio examples
-
-### 🔒 Security & Quality
-
-- ✅ All Sonar quality gates validated
-- ✅ No hardcoded credentials in changes
-- ✅ New Clawdbot files are documentation only (no executable code)
-- ✅ All 213 BATS + 47 Node tests passing
-
-### 📊 Changes Summary
-
-- **Files Added:** 2 (`.clawdbot/README.md`, `.clawdbot/skill/SKILL.md`)
-- **Files Modified:** 2 (`README.md`, `package.json`)
-- **Commits Since v3.1.0:** 11 (5 fixes, 4 docs, 1 test, 1 debug)
-
-### 🎯 User Impact
-
-**For Clawdbot Users:**
-- Professional TTS with 50+ voices in 30+ languages
-- Works on remote servers with SSH audio tunneling
-- Zero API costs—Piper TTS is free and offline
-- Automatic integration when AgentVibes is installed
-
-**For Existing Users:**
-- Zero breaking changes
-- All existing features work exactly the same
-- Clawdbot support is additive
-
-### 🚀 Migration Notes
-
-No migration required! This is a fully backward-compatible minor release.
-
-**To Use with Clawdbot:**
-1. Install: `npx agentvibes install`
-2. Speak: `npx agentvibes speak "Hello!"`
-
-### 📦 Full Changelog
-
-**Feature Commits:**
-- `(pending)` feat: Add Clawdbot integration
-
-**Bug Fix Commits:**
-- `5cd97d52` fix: Disable SonarCloud quality gate status reporting to GitHub
-- `12f822e6` fix: Disable quality gate failure in SonarCloud workflow
-- `0d26ccc2` fix: Ensure coverage file is generated before SonarCloud scan
-- `c2465508` fix: Add CLI tests and exclude CLI entry point from coverage
-- `c673afe1` fix: Remove macos-15-large runner to avoid GitHub billing limits
-- `92271732` fix: Update Piper installation test to match current voice list
-
-**Documentation Commits:**
-- `f72dd977` docs: Add guide to disable SonarCloud GitHub App checks
-- `6587519b` docs: Add step-by-step SonarCloud dashboard configuration guide
-- `ba765f50` docs: Add SonarCloud quality gate configuration guidance
-
-**Test Commits:**
-- `47f08a79` test: Trigger workflow to verify SonarCloud quality gate fix
-
-**Debug Commits:**
-- `84945d25` debug: Add coverage file verification to SonarCloud workflow
-
----
-
-## 📦 v3.1.0 - Android Native Support: Run Claude Code on Your Phone
-
-**Release Date:** January 22, 2026
-
-### 🎯 Why v3.1.0?
-
-This minor release brings **native Android support** to AgentVibes, enabling developers to run Claude Code with professional TTS voices directly on Android devices via Termux. No SSH required, no remote server needed—just install Termux on your Android phone or tablet and get the full AgentVibes experience locally. This complements the v3.0.0 termux-ssh provider by offering a **complete mobile development solution**: use native Termux for local Android development, or use termux-ssh when connecting to remote servers.
-
-### 🤖 AI Summary
-
-AgentVibes v3.1.0 introduces native Android/Termux support, enabling developers to run Claude Code with professional TTS voices directly on their Android devices. Through automatic detection and a specialized installer, AgentVibes now runs Piper TTS via proot-distro with Debian (solving Android's glibc compatibility issues), uses termux-media-player for audio playback, and includes comprehensive Android-specific documentation. Perfect for developers who want to code on-the-go with their Android phone or tablet using the full power of Claude Code and AgentVibes.
-
-**Key Highlights:**
-- 🤖 **Native Android Support** - Run Claude Code with TTS directly on Android devices via Termux
-- 📦 **Automatic Termux Detection** - AgentVibes auto-detects Android and runs specialized installation
-- 🎯 **Proot-Distro Integration** - Solves glibc compatibility with proot Debian environment
-- 🔊 **Android Audio Playback** - Uses termux-media-player for native Android audio routing
-- 📚 **Comprehensive Documentation** - Complete Android setup guide with troubleshooting and F-Droid instructions
-- ✅ **Full Test Coverage** - All 213 BATS + 38 Node tests passing with Android compatibility
-
-### ✨ New Features
-
-**Termux Installer (`.claude/hooks/termux-installer.sh`):**
-- New 224-line installer specifically for Android/Termux environments
-- Automatically installs proot-distro with Debian (for glibc compatibility)
-- Downloads and configures Piper TTS binary in proot environment
-- Creates `/usr/bin/piper` wrapper that routes through proot
-- Installs audio dependencies: ffmpeg, sox, bc, termux-api
-- Interactive voice selection with 50+ language options
-- Validates Termux environment before proceeding
-
-**Termux Detection (`src/installer.js`):**
-- New `isTermux()` function checks for `/data/data/com.termux` directory
-- New `detectAndNotifyTermux()` displays Android detection messages
-- Auto-configures piper provider with Termux-compatible voice
-- Shows Termux-specific installation instructions
-- Piper installer automatically redirects to termux-installer.sh on Android
-
-**Audio Processor Updates (`.claude/hooks/audio-processor.sh`):**
-- Detects Termux environment for temp directory selection
-- Uses `${PREFIX}/tmp` on Termux, `/tmp` on standard systems
-- Ensures audio effects work correctly on Android
-- Cross-platform compatibility maintained
-
-**Piper Installer Updates (`.claude/hooks/piper-installer.sh`):**
-- Auto-detects Termux and redirects to specialized installer
-- Shows clear message when routing to Termux-specific setup
-
-**Android Audio Playback (`.claude/hooks/play-tts-piper.sh`):**
-- Detects Android/Termux environment
-- Uses `termux-media-player` instead of `paplay` on Android
-- Audio routes through Android's native media system
-
-### 📚 Documentation
-
-**New Android Setup Section (`README.md`):**
-- Added "🤖 Android / Termux" section to System Requirements
-- Complete 3-step installation guide for Android devices
-- Explanation of why Termux needs special handling (bionic vs glibc)
-- Requirements: Termux app from F-Droid, Termux:API, Android 7.0+
-- Audio playback architecture explanation
-- Setup verification commands
-- Troubleshooting table for common issues
-- Clear explanation of why F-Droid version is required (not Google Play)
-- Updated Quick Links table with direct link to Android setup
-
-### 🐛 Bug Fixes
-
-- **Test #90 Fix** - Added termux-ssh provider to test cleanup list for "no providers found" edge case
-- **Temp Directory Detection** - Fixed audio-processor.sh defaulting to Termux paths on non-Termux systems
-- **Sonar Compliance** - Added `set -euo pipefail` strict mode to termux-installer.sh for security
-
-### 🔒 Security & Quality
-
-- ✅ All Sonar quality gates validated
-- ✅ Strict mode (`set -euo pipefail`) on all new bash scripts
-- ✅ No hardcoded credentials
-- ✅ Proper variable quoting and input validation
-- ✅ Cross-platform temp directory handling
-- ✅ All 213 BATS + 38 Node tests passing
-
-### 📊 Changes Summary
-
-- **Files Modified:** 7
-- **Lines Added:** +391
-- **Lines Removed:** -8
-- **New Files:** 1 (termux-installer.sh)
-- **Commits:** 8 (5 fixes, 1 feature, 1 docs, 1 merge)
-
-### 🎯 User Impact
-
-**For Android Users:**
-- Can now run Claude Code directly on Android phones/tablets
-- Full TTS support with 50+ voices and languages
-- No remote server required for basic usage
-- Works offline after initial voice downloads
-
-**For Developers:**
-- Complete mobile development solution (native + SSH)
-- Native Termux for local Android development
-- Termux-SSH provider for remote server connections
-- Seamless integration with existing AgentVibes workflows
-
-**For Existing Users:**
-- Zero breaking changes
-- All existing features work exactly the same
-- New Android support is opt-in via Termux installation
-
-### 🚀 Migration Notes
-
-No migration required! This is a fully backward-compatible minor release.
-
-**To Try Android Support:**
-1. Install [Termux from F-Droid](https://f-droid.org/en/packages/com.termux/)
-2. Install [Termux:API](https://f-droid.org/en/packages/com.termux.api/)
-3. In Termux: `pkg install nodejs-lts`
-4. Run: `npx agentvibes install`
-
-AgentVibes will auto-detect Termux and run the specialized installer.
-
-### 📦 Full Changelog
-
-**Feature Commits:**
-- `e9d4cf95` feat: Add Android/Termux support for Piper TTS
-
-**Bug Fix Commits:**
-- `aa4d3cdd` fix: Add termux-ssh provider to test #90 cleanup list
-- `c1b00c6d` fix: Use termux-media-player for audio playback on Android
-- `f96ab89a` fix: Properly detect Termux environment for temp directory
-- `e2efeb06` fix: Add strict mode to termux-installer.sh for Sonar compliance
-
-**Documentation Commits:**
-- `701a9412` docs: Add comprehensive Android/Termux setup section to README
-
-**Merge Commits:**
-- `a5d3f546` Merge feature/android-termux into master
-- `95f04e70` Merge origin/master into feature/pulseaudio-reverse-tunnel
-
----
-
-## 📦 v3.0.0 - Cross-Platform Remote Audio: Termux SSH Provider
-
-**Release Date:** January 8, 2026
-
-### 🎯 Why v3.0.0?
-
-This major release marks a significant milestone in AgentVibes' evolution, introducing **mobile-first interactive AI conversations**. The termux-ssh provider enables a revolutionary workflow: **have fully interactive, hands-free conversations with Claude Code using just your mobile device**—whether you're coding locally on your laptop with audio routed to your phone, or working on remote servers from anywhere in the world. This architectural breakthrough represents a new paradigm: **"Code with your hands, converse with your voice."**
-
-### 🤖 AI Summary
-
-AgentVibes v3.0.0 introduces the termux-ssh TTS provider, enabling **true mobile-first interactive conversations with Claude Code**. Route TTS audio to your Android device via SSH—whether coding locally on your laptop or on remote servers—and have hands-free, voice-driven conversations with Claude using just your phone. This major release includes comprehensive Tailscale VPN setup documentation for internet-wide access, full MCP server integration, and transforms how developers interact with AI assistants. Perfect for developers who want to experience AI conversations naturally through their mobile device while their hands stay on the keyboard.
-
-**Key Highlights:**
-- 📱 **Mobile-First AI Conversations** - Have fully interactive, hands-free conversations with Claude Code using just your Android device
-- 💻 **Local + Remote Development** - Works for both local coding (laptop → phone audio) and remote server development
-- 🌐 **Tailscale Integration** - Comprehensive guide for internet-wide device access without port forwarding or firewall configuration
-- 🔧 **Enhanced Installer** - Interactive SSH host configuration with validation and clear use-case guidance
-- 🎯 **Full MCP Compatibility** - Complete integration with all MCP commands and workflows
-- 🛡️ **Quality Gates Integration** - Automated security validation in release process
-
-### 🎥 Demo Video
-
-**Watch it in action:** [Mobile-First AI Conversations with Claude Code](https://youtu.be/ngLiA_KQtTA?si=wTwS4CJidIxWqLIP)
-
-See the termux-ssh provider in action—fully interactive, hands-free conversations with Claude Code using just your Android device.
-
-### ✨ New Features
-
-**Termux SSH TTS Provider (`.claude/hooks/play-tts-termux-ssh.sh`):**
-- New TTS provider for Android via SSH connection
-- Routes text to `termux-tts-speak` on remote Android device
-- Configuration priority: env var → project → global
-- Secure quote escaping for safe text transmission
-- 196 lines of fully documented code
-
-**Installer Updates (`src/installer.js`):**
-- Added termux-ssh to provider selection menu
-- Interactive SSH host alias configuration with validation
-- Saves host alias to `.claude/termux-ssh-host.txt`
-- Clear use case description: "Only choose if your project is on a remote server and you want audio sent to your Android device"
-- Documentation link to TERMUX_SETUP.md
-
-**TTS Router Updates (`.claude/hooks/play-tts.sh`):**
-- Added termux-ssh provider routing in two locations
-- Full integration with existing provider detection
-- Supports mixed-provider mode (Piper + Termux)
-
-**MCP Server Integration (`mcp-server/server.py`):**
+# AgentVibes Release Notes
+
+## 🎵 v5.5.0 — Per-LLM Audio Routing & Windows Installer Resilience
+
+**Released:** 2026-04-27
+
+### 🆕 Per-LLM Audio Routing
+Each LLM (Claude Code, Copilot, Codex) can now have its own voice, pretext, reverb, and
+background-music settings. The MCP server passes `--llm <key>` to both `play-tts.sh`
+(Linux/macOS) and `play-tts.ps1` (Windows), and the scripts look up `llm:<key>` rows in
+`audio-effects.cfg`. Default rows for `claude-code`, `copilot`, and `codex` ship out of the
+box; configure them via **Setup → Default → Configure** in the TUI.
+
+### 🐛 Windows Installer Crash Fix
+Fixed `spinner.info is not a function` error that crashed AgentVibes **reinstalls** on Windows
+when users had an older global install. All 10 file-copy functions in the installer now wrap
+their spinner with `createRobustSpinner()` so stale callers can never cause a crash regardless
+of which methods they expose.
+
+### 🎶 Windows Background Music Parity
+Windows TTS playback now prefers `ffplay` (sinc resampling, no artefacts) over the low-quality
+WinMM `SoundPlayer` resampler. The new `Invoke-AudioPlay` helper handles the fallback
+transparently — if `ffplay` is unavailable, `SoundPlayer` is used as before.
+
+### 🎉 Party Mode Cross-Platform Entry Point
+BMAD party mode step files and the Copilot skill now consistently reference
+`node bin/bmad-speak.js` — the single cross-platform entry point that delegates to
+`bmad-speak.ps1` on Windows and `bmad-speak.sh` elsewhere.
+
+### 🔧 Other Fixes
+- `play-tts.sh` now accepts a named `--llm <key>` flag in addition to the `LLM_PROVIDER` env var
+- `mcp-server/server.py` routes `AGENTVIBES_LLM` → `CLAUDECODE=1` → `AGENTVIBES_MCP_FALLBACK`
+  priority chain and forwards the resolved key as `-llm`/`--llm` to TTS scripts
+- Added `audio-effects.cfg` rows for `llm:claude-code`, `llm:copilot`, `llm:codex`
+- Added `command-routing.test.js` and `ConfigService` unit tests
+- npm pack content guard now catches untracked publishable files
+
+### 📊 Technical
+- 231 tests passing (0 failures)
+
+---
+
+## 🎛️ v5.4.0 — TUI Installer, Spinner Fix & Dependency Cleanup
+
+**Released:** 2026-04-22
+
+### ✨ New
+- **TUI Installer**: Interactive terminal UI for guided installation — browse voices, configure providers, enable BMAD party mode, all from a beautiful terminal interface
+- **Cross-Platform Spinner Fix**: Resolved `spinner.info is not a function` crash on WSL/Linux that blocked installation
+
+### 🐛 Bug Fixes
+- **Removed circular self-dependency**: `package.json` was depending on `agentvibes@^3.5.9` (itself), causing npm to shadow the fixed binary with the old buggy one — the silent cause of the spinner crash on repeat installs
+- **Restored background music volume fallback**: `audio-processor.sh` `bg_volume="0.20"` fallback lost in merge was restored
+- **Fixed PROJECT_ROOT detection in `play-tts.sh`**: Walk-up logic was going 2 levels too far, causing TTS to use global `~/.agentvibes` config instead of project config
+
+### 🔧 Technical
+- 706/738 tests passing
+
+---
+
+## ✨ v3.5.10 - Soprano Detection Fixes & Enhanced Installer Features
+
+**Release Date:** February 14, 2026
+
+### 🎯 Summary
+
+Production release combining critical bug fixes and new installer features. Fixed Soprano TTS detection for pipx installations (the core issue reported by users), resolved 5 execSync API misuse bugs that were breaking Python package detection, and eliminated 100+ lines of code duplication. Introduces new installer features: custom music track support with preview functionality, personality emoji mapping for better visual recognition, and pretext configuration allowing users to customize agent introductions.
+
+### ✨ Key Features & Fixes
+
+**🔧 Critical Bug Fixes:**
+- **Soprano TTS Detection:** Fixed detection when installed via pipx (was showing "not installed" despite working)
+- **execSync API Bugs:** Fixed 5 locations using incorrect API signature (array args with execSync)
+- **Code Duplication:** Eliminated 100+ lines of duplicate code between Soprano and Piper validators
+- **API Consistency:** All provider validation functions now return consistent response structures
+- **Python Package Detection:** Fixed broken Python pip detection that was silently failing
+
+**🎨 New Installer Features:**
+- **Custom Music Tracks:** Users can now upload and preview their own background music
+- **Personality Emojis:** Visual recognition mapping (😊 for none, 🎭 for dramatic, 💁 for sassy, etc.)
+- **Pretext Configuration:** Custom agent introductions (e.g., "FireBot: " prefix for all messages)
+- **Track Preview:** Audio preview with support for ffplay, sox, and mpv players
+
+**🛡️ Security & Quality:**
+- Improved path traversal protection in provider validation
+- Enhanced error handling and logging
+- Reduced code complexity by 29% through deduplication
+- Test coverage improved: 56.61% → 63.67%
+
+### 📊 Technical Details
+
+**Soprano Detection Improvements:**
+- Checks command in PATH first (most reliable for pipx)
+- Falls back to ~/.local/bin directory check
+- Checks pipx venv directory for installation
+- Final fallback to Python pip package detection
+- Consistent error messages showing all checked locations
+
+**Code Quality Metrics:**
+- File size reduced by 145 lines (-29%)
+- Code duplication eliminated (was ~100 lines)
+- Test coverage improved +7.06%
+- All 114 tests passing
+- Provider validator now 63.67% covered (up from 56.61%)
+
+**Installer Enhancements:**
+- Added personality emoji mapping (26 personalities)
+- Custom track upload with validation
+- File type restrictions (.mp3, .wav, .ogg, .m4a)
+- Registry storage in ~/.agentvibes/custom-tracks.json
+- Audio preview before finalizing selection
+
+### 🔒 Security Notes
+
+- All spawnSync calls now use correct array argument form
+- Path traversal prevention maintained in all operations
+- HOME injection protection via os.homedir()
+- No hardcoded credentials introduced
+- Input validation for file uploads and track selection
+
+### 🐛 Known Limitations
+
+- Audio preview requires ffplay, sox, or mpv (feature gracefully degrades if unavailable)
+- Custom track registry is stored locally per user
+- Some legacy bash scripts still lack strict mode (pre-existing, low risk)
+
+### 🙏 Acknowledgments
+
+This release includes fixes identified through adversarial code review, ensuring production-quality reliability and security alignment with CLAUDE.md standards.
+
+---
+
+## 🛡️ v3.5.8 - Provider Validation Security & UX Improvements
+
+**Release Date:** February 12, 2026
+
+### 🎯 Summary
+
+Critical security and reliability update for provider detection. Fixes command injection vulnerabilities in validation code, prevents HOME directory injection attacks, and improves UX with explicit provider detection messaging. Soprano TTS installed via pipx is now correctly detected (previously showed "not installed" due to ES module import error). All 8 critical code review issues resolved with comprehensive security hardening and enhanced error reporting.
+
+### ✨ Key Improvements
+
+- **🔐 Security Fixes:** Fixed command injection vulnerability (template strings → array form), prevented HOME injection attacks, added path traversal protection
+- **✅ Provider Detection:** Soprano via pipx now correctly detected; added checkedLocations tracking for transparency
+- **💬 Better Messaging:** Explicit "Detected and selected!" confirmation; detailed error messages showing what was checked
+- **🧪 Test Coverage:** Enhanced tests verify actual detection values, not just types
+- **🐛 Debugging:** Added [DEBUG] logging for troubleshooting provider issues
+
+### 🔴 Critical Fixes
+
+1. **Command Injection Prevention** - All execSync calls now use array form (security: CLAUDE.md)
+2. **HOME Directory Injection** - Switched to os.homedir() instead of process.env.HOME
+3. **Path Traversal Protection** - Added path.resolve() validation for pipx venv directories
+
+### 🟡 Medium Fixes
+
+4. **Pipx Logic Improved** - Tracks checked locations even on success (transparency)
+5. **Silent Failures Eliminated** - Added [DEBUG] error logging for diagnostics
+6. **Test Quality Enhanced** - Verify message content, not just types
+7. **Documentation** - Added JSDoc comments explaining security-critical imports
+8. **Error Differentiation** - Better distinction between different failure types
+
+### 📊 Technical Impact
+
+- Soprano detection now works reliably for both pip and pipx installations
+- Reduced false negatives in provider validation
+- Enhanced security posture aligned with CLAUDE.md security mandates
+- Improved debuggability with explicit error messages
+
+---
+
+## 🔧 v3.5.7 - CLI Fix: npx Command Output & Startup Hooks
+
+**Release Date:** February 12, 2026
+
+Fixes critical bug where `npx agent-vibes install` and other commands produced no output, making CLI unusable. Root cause: bin/agent-vibes used dynamic import without passing arguments to installer.js on local execution. Also removed broken hook configurations (pre_compact.py, notification.ts) that didn't exist and caused startup errors in Claude Code settings.
+
+### 🎯 What's Fixed
+
+- **npx agent-vibes now works** - `npx agent-vibes install`, `npx agent-vibes --help`, all commands produce proper output
+- **Startup hook errors gone** - Removed non-existent hook references from settings.json (pre_compact.py, notification.ts)
+- **CLI execution proper** - Both npx and local execution now use execFileSync with proper argument passing
+
+### 🚀 Technical Details
+
+**Before v3.5.7:**
+```javascript
+// bin/agent-vibes (local execution path)
+import('../src/installer.js');  // ❌ No args, doesn't await
+```
+
+**After v3.5.7:**
+```javascript
+// bin/agent-vibes (all execution paths)
+execFileSync('node', [installerPath, ...arguments_], {
+  stdio: 'inherit',
+  cwd: isNpxExecution ? path.dirname(__dirname) : process.cwd(),
+});  // ✅ Passes args, proper I/O
+```
+
+---
+
+## 🔧 v3.5.6 - Bug Fix: Bash Hook Parameter Handling
+
+**Release Date:** February 11, 2026
+
+Fixes critical regression in v3.5.5 where bash hooks failed with unbound variable errors when called with optional parameters under strict mode. Affects `play-tts.sh` and `provider-manager.sh`.
+
+---
+
+## 📦 v3.5.5 - Native Windows Support: Soprano, Piper & SAPI Providers
+
+**Release Date:** February 12, 2026
+
+### 🎯 Why v3.5.5?
+
+v3.5.5 brings **native Windows support** to AgentVibes with a full-featured PowerShell installer and three TTS providers. Windows users no longer need WSL - AgentVibes runs natively with Soprano (neural), Piper (offline neural), or Windows SAPI (zero-setup) voices. The installer also adds **background music selection** (16 genre tracks), **reverb/audio effects** (via ffmpeg aecho), and **verbosity control** for the TTS experience.
+
+### 🚀 Key Highlights
+
+#### 🖥️ Native Windows TTS (NEW!)
+- **3 providers**: Soprano (ultra-fast neural), Piper (offline neural), Windows SAPI (built-in)
+- **Beautiful PowerShell installer** with figlet banner and interactive setup
+- **8 hook scripts** for complete TTS functionality on Windows
+- **MCP server** auto-resolves `.sh` to `.ps1` on Windows
+- **46 Windows-specific unit tests** with full coverage
+
+#### 🎵 Background Music Selection
+- **16 genre tracks**: Flamenco, Bachata, Bossa Nova, City Pop, Chillwave, and more
+- **Interactive picker** in the installer with descriptions
+- **ffmpeg mixing**: 2s intro, voice over music, 2s fade-out outro
+
+#### 🎛️ Reverb / Audio Effects
+- **5 reverb levels**: Off, Light, Medium, Heavy, Cathedral
+- **ffmpeg aecho filter** (no SOX dependency on Windows)
+- Applied before background music mixing for clean layering
+
+#### 🔊 Verbosity Control
+- **3 levels**: High (full reasoning), Medium (key updates), Low (essential only)
+- Integrates with session-start-tts.ps1 protocol instructions
+
+### 🤖 AI Summary
+
+AgentVibes v3.5.5 delivers native Windows support with a polished PowerShell installer offering three TTS providers (Soprano neural, Piper offline, Windows SAPI), background music selection from 16 genre tracks, reverb effects via ffmpeg aecho filter, and verbosity control. The release includes 8 Windows hook scripts, MCP server platform detection for automatic .sh-to-.ps1 resolution, and 46 new unit tests. Security hardening adds path traversal prevention with regex allowlisting and path containment checks, reverb config allowlist validation, and strict mode compliance across all scripts. Cross-platform test fixes ensure the full 93-test suite passes on both Windows and Unix.
+
+---
+
+## ✨ New Features
+
+### Native Windows TTS
+- Full PowerShell installer (`setup-windows.ps1`) with figlet banner and interactive UX
+- Soprano provider (`play-tts-soprano.ps1`) with Gradio WebUI integration
+- Piper provider (`play-tts-windows-piper.ps1`) with auto-download of voices from HuggingFace
+- Windows SAPI provider (`play-tts-windows-sapi.ps1`) with zero-setup built-in voices
+- TTS router (`play-tts.ps1`) with mute support, background music mixing, and reverb
+- Provider manager, voice manager, audio cache utils, and session-start hook scripts
+- MCP server `.sh` to `.ps1` auto-resolution on Windows
+
+### Installer Enhancements
+- Background music selection with 16 genre tracks and interactive picker
+- Reverb/audio effects selection (Off/Light/Medium/Heavy/Cathedral)
+- Verbosity control (High/Medium/Low) for TTS protocol instructions
+- Updated completion screen showing all 4 settings (provider, background, reverb, verbosity)
+
+---
+
+## 🐛 Bug Fixes
+
+### Security Fixes
+- Fix path traversal in background music config reader (regex allowlist + path containment)
+- Add allowlist validation for reverb-level.txt config (prevent invalid values)
+- Add `set -euo pipefail` strict mode to `play-tts.sh` for Sonar compliance
+
+### Cross-Platform Fixes
+- Fix self-copy error when setup-windows.ps1 runs from project root
+- Fix test executable permission checks on Windows (skip Unix mode bits)
+- Fix test path separator comparison in uninstall test (use `path.join` not hardcoded `/`)
+
+---
+
+## 🏗️ Improvements
+
+### Code Quality
+- Reverb config uses switch-as-allowlist pattern - file content never flows into commands
+- All SoundPlayer instances wrapped in try/finally for resource disposal
+- Environment variable cleanup (`AGENTVIBES_NO_PLAY`) on all exit paths
+- Input validation with regex + range checks for all installer prompts
+
+### Testing
+- 46 new Windows-specific unit tests (hook scripts, providers, security, encoding)
+- 3 cross-platform test fixes for Windows compatibility
+- Full suite: 93 Node tests passing on Windows
+
+---
+
+## 📊 Statistics
+
+- **7 commits** since v3.4.1
+- **3,769 lines added**, 211 removed across 24 files
+- **9 new PowerShell scripts** for Windows TTS
+- **93 tests passing** (46 Windows + 47 cross-platform)
+- **24/24 Sonar quality gates** passing
+- **Security score**: All path traversal and injection vectors reviewed
+
+---
+
+## 🔧 Technical Details
+
+### Files Added
+- `.claude/hooks-windows/play-tts.ps1`: TTS router with reverb and background music
+- `.claude/hooks-windows/play-tts-soprano.ps1`: Soprano neural TTS provider
+- `.claude/hooks-windows/play-tts-windows-piper.ps1`: Piper offline TTS provider
+- `.claude/hooks-windows/play-tts-windows-sapi.ps1`: Windows SAPI built-in voices
+- `.claude/hooks-windows/provider-manager.ps1`: Provider switching
+- `.claude/hooks-windows/voice-manager-windows.ps1`: Voice browsing and selection
+- `.claude/hooks-windows/audio-cache-utils.ps1`: Cache management
+- `.claude/hooks-windows/session-start-tts.ps1`: Auto-activates TTS on Claude start
+- `setup-windows.ps1`: Full Windows installer with 4 interactive sections
+- `test/unit/windows-tts.test.js`: 46 Windows-specific unit tests
+
+### Breaking Changes
+None - all changes are backward compatible. Existing Unix/macOS installations are unaffected.
+
+---
+
+## 🎓 Migration Notes
+
+### For New Windows Users
+1. Run `npx agentvibes install` (Node.js) or `.\setup-windows.ps1` (PowerShell)
+2. Follow the interactive setup
+3. Choose provider (Soprano, Piper, or SAPI)
+4. Select background music, reverb, and verbosity
+5. TTS works automatically in Claude Code sessions
+
+### For Existing Unix/macOS Users
+- No changes required - your setup continues working
+- All Unix bash hooks remain untouched
+- Only `play-tts.sh` gained `set -euo pipefail` (strict mode)
+
+---
+
+## 🙏 Acknowledgments
+
+### Project Lead
+- **[@paulpreibisch](https://github.com/paulpreibisch)** (Paul Preibisch) — Creator and maintainer of AgentVibes
+
+### Community Contributors
+- **[@nathanchase](https://github.com/nathanchase)** — For contributing the Soprano TTS provider in v3.4.0, whose ultra-fast neural engine is now one of the three Windows-native providers
+- **[@alexeyv](https://github.com/alexeyv)** — For suggesting native Windows support and recommending Windows SAPI as a zero-dependency provider
+- **[@bmadcode](https://github.com/bmadcode)** (Brian Madison) — Creator of the [BMAD Method](https://github.com/bmadcode/BMAD-METHOD), used daily for planning and building AgentVibes features
+
+### Quality Assurance
+- **Adversarial Security Review**: Path traversal, injection, and resource disposal all validated
+- **Testing**: 93/93 tests passing (100% suite coverage)
+- **Quality Gates**: 24/24 Sonar requirements validated
+- **Co-Authored-By**: Claude Opus 4.6
+
+---
+
+**Full Changelog**: https://github.com/paulpreibisch/AgentVibes/compare/v3.4.1...v3.5.5
+
+---
+
+## 📦 v3.4.0 - Soprano TTS, Security Hardening & Environment Intelligence
+
+**Release Date:** February 10, 2026
+
+### 🎯 Why v3.4.0?
+
+v3.4.0 introduces **Soprano TTS** - an ultra-fast neural TTS provider with GPU acceleration, comprehensive **security hardening** across the codebase, and **intelligent environment detection** that recognizes PulseAudio tunnels for remote audio scenarios.
+
+### 🚀 Key Highlights
+
+#### ⚡ Soprano TTS Provider (NEW!)
+- **80M parameter neural model** with premium female English voice
+- **20x CPU speed** (vs Piper), **2000x GPU speed** with CUDA
+- **3 synthesis modes**: WebUI (Gradio), API (OpenAI-compatible), CLI (fallback)
+- **Auto-detection**: Checks for running Gradio server, falls back gracefully
+- **<1GB memory footprint** - perfect for low-RAM systems
+- **Provider-aware voice management**: Auto-selects single voice, shows model specs
+- **Thanks to [@nathanchase](https://github.com/nathanchase)** for this contribution! ([see acknowledgments](#-acknowledgments))
+
+#### 🛡️ Security Hardening (9.5/10 Score)
+- **Timeouts on system commands**: Prevents installer hangs (nvidia-smi, sysctl, meminfo)
+- **Bounds checking**: Validates array access before parsing system output
+- **NaN validation**: Prevents crashes from malformed memory/GPU detection
+- **Case-insensitive checks**: PulseAudio tunnel detection handles TCP: and tcp:
+- **Code duplication eliminated**: Extracted PulseAudio helper function (DRY)
+
+#### 🌐 Environment Intelligence
+- **PulseAudio tunnel detection**: Recognizes `PULSE_SERVER=tcp:*` as working audio
+- **Context-aware messaging**:
+  - "🌐 PulseAudio Tunnel Detected!" for SSH + tunnel setups
+  - "🔊 Audio Output Detected!" for local speakers
+  - Distinguishes local/tunnel/hybrid configurations
+- **Smart environment classification**:
+  - DESKTOP: Local audio OR active PulseAudio tunnel
+  - VOICELESS: No audio AND no tunnel
+  - PHONE: Termux/Android devices
+
+#### 🎤 Installer Enhancements
+- **Provider-aware voice pages**: Soprano shows model specs, Piper shows 50+ voices
+- **Auto-selection logic**: Soprano (1 voice) auto-selects, no manual choice needed
+- **GPU-based recommendations**: "Your GPU will run Soprano 2000x faster!"
+- **RAM-based suggestions**: Low memory systems see "Soprano uses <1GB" message
+- **Better RAM display**: Shows "512MB" instead of "0GB" for sub-1GB systems
+
+### 🤖 AI Summary
+
+AgentVibes v3.4.0 brings Soprano TTS - an 80M parameter neural provider offering 20x CPU and 2000x GPU acceleration with sub-1GB memory footprint - plus comprehensive security hardening (timeouts, bounds checking, NaN validation) and intelligent environment detection that recognizes PulseAudio tunnels as working audio for remote scenarios. The enhanced installer provides context-aware messaging distinguishing local speakers from SSH tunnels, GPU-based provider recommendations (Soprano for CUDA users, macOS Say for Apple, Piper for versatility), and provider-specific voice pages that auto-select Soprano's single voice while showcasing model specifications. This release achieves a 9.5/10 security score through systematic defensive programming, making AgentVibes production-ready for enterprise deployments while expanding TTS provider options for diverse hardware configurations.
+
+---
+
+## ✨ New Features
+
+### Soprano TTS Provider
+- Add Soprano TTS provider script with 3 synthesis modes (WebUI, API, CLI) (#95)
+- Integrate Soprano into TTS router and provider manager
+- Add soprano-gradio-synth.py helper for WebUI/SSE protocol
+- Provider-aware voice selection page with model specifications
+- Auto-select single Soprano voice with performance details
+
+### Installer Intelligence
+- Add `detectSystemCapabilities()` for GPU/RAM detection
+- Add `hasPulseAudioTunnel()` helper function
+- Context-aware audio detection messaging (tunnel vs local)
+- GPU-based provider ordering (Soprano first for CUDA users)
+- RAM-based recommendations (<4GB systems see Soprano first)
+- Provider-specific intro messages (Soprano vs Piper vs macOS)
+
+### Environment Detection
+- PulseAudio tunnel recognition via PULSE_SERVER env var
+- Case-insensitive TCP protocol detection
+- Smart DESKTOP classification (local audio OR tunnel)
+- Improved VOICELESS detection (no audio AND no tunnel)
+
+---
+
+## 🐛 Bug Fixes
+
+### Security Fixes
+- Add 5s timeout to nvidia-smi to prevent GPU detection hangs
+- Add 3s timeout to sysctl/meminfo to prevent memory detection hangs
+- Add bounds checking before parsing sysctl output (macOS)
+- Add bounds checking before parsing /proc/meminfo (Linux)
+- Add NaN validation for parseInt() memory size parsing
+- Fix case sensitivity in PULSE_SERVER detection (handles TCP: and tcp:)
+
+### Test Fixes
+- Fix provider-manager test #90: Add soprano and ssh-remote to cleanup list
+- Ensure zero-provider edge case properly simulates empty state
+
+### User Experience
+- Fix RAM display for <1GB systems (show "512MB" not "0GB")
+- Fix PulseAudio selection triggering wrong setup flow
+- Separate PulseAudio tunnel setup from SSH receiver setup
+
+---
+
+## 🏗️ Improvements
+
+### Code Quality
+- Extract PulseAudio detection to helper function (DRY principle)
+- Implement system capabilities caching (eliminates duplicate calls)
+- Add comprehensive error handling in detectSystemCapabilities()
+- Improve code comments for security-critical sections
+
+### Performance
+- Cache system detection results (prevents duplicate nvidia-smi calls)
+- Add timeouts to prevent indefinite hangs
+- Optimize provider detection with early returns
+
+### Documentation
+- Add comprehensive commit message documenting all changes
+- Document security improvements (timeouts, bounds checking, NaN validation)
+- Explain PulseAudio tunnel detection architecture
+- Detail environment classification logic
+
+---
+
+## 📊 Statistics
+
+- **91 commits** since v3.3.0
+- **817 lines added** in merge to master
+- **6 files modified** in core integration
+- **260 tests passing** (213 BATS + 47 Node)
+- **Security score**: 7.5/10 → 9.5/10
+- **Test coverage**: 100% pass rate
+
+---
+
+## 🔧 Technical Details
+
+### Files Modified
+- `src/installer.js`: +335 lines (security fixes, environment detection, Soprano integration)
+- `test/unit/provider-manager.bats`: +4 lines (fix edge case test)
+- `.claude/hooks/play-tts-soprano.sh`: +320 lines (new provider)
+- `.claude/hooks/soprano-gradio-synth.py`: +139 lines (new helper)
+- `.claude/hooks/provider-manager.sh`: +17 lines (Soprano support)
+- `.claude/hooks/play-tts.sh`: +6 lines (route to Soprano)
+
+### Breaking Changes
+None - all changes are backward compatible.
+
+### Dependencies
+- **New**: `soprano-tts` (Python package, optional)
+- **Recommended**: CUDA-capable GPU for 2000x speedup (optional)
+- **Compatible**: Works on CPU-only systems (20x vs Piper)
+
+---
+
+## 🎓 Migration Notes
+
+### For New Users
+1. Run `npx agentvibes install`
+2. Installer auto-detects your hardware (GPU, RAM, platform)
+3. Soprano appears as option if you have working audio
+4. Select Soprano for ultra-fast TTS with GPU acceleration
+
+### For Existing Users
+1. Update: `npx agentvibes update`
+2. Switch provider: `/agent-vibes:provider switch soprano`
+3. Test: `/agent-vibes:sample soprano-default`
+4. Optionally install soprano-tts: `pip install soprano-tts`
+
+### PulseAudio Tunnel Users
+- Installer now auto-detects your tunnel configuration
+- Shows "🌐 PulseAudio Tunnel Detected!" instead of "speakers"
+- Provides DESKTOP mode options (Soprano, Piper, macOS Say)
+- No manual configuration needed
+
+---
+
+## 🙏 Acknowledgments
+
+### Special Thanks
+
+**🎉 [@nathanchase](https://github.com/nathanchase)** - For contributing the Soprano TTS Provider integration (PR #95)! Nathan's work brings ultra-fast neural TTS with GPU acceleration to AgentVibes, offering 20x CPU and 2000x GPU performance improvements. The comprehensive integration includes WebUI, API, and CLI synthesis modes with intelligent auto-detection and graceful fallback. Thank you for this outstanding contribution! 🚀
+
+### Quality Assurance
+
+- **Security Review**: Adversarial code review achieved 9.5/10 score
+- **Testing**: All 260 tests pass (100% suite coverage)
+- **Quality Gates**: All Sonar requirements validated
+- **Co-Authored-By**: Claude Sonnet 4.5
+
+---
+
+## 📚 Additional Resources
+
+- [Soprano TTS Documentation](https://github.com/paulpreibisch/AgentVibes/blob/master/docs/providers.md#soprano-tts)
+- [PulseAudio Tunnel Setup](https://github.com/paulpreibisch/AgentVibes/blob/master/docs/SSH_REMOTE_SETUP.md)
+- [Security Hardening Guide](https://github.com/paulpreibisch/AgentVibes/blob/master/docs/security-hardening-guide.md)
+- [Provider Comparison](https://github.com/paulpreibisch/AgentVibes/blob/master/docs/providers.md)
+
+---
+
+**Full Changelog**: https://github.com/paulpreibisch/AgentVibes/compare/v3.3.0...v3.4.0
+
+---
+
+## 📦 v3.3.0 - Remote TTS, Smart Installer, OpenClaw Receiver & Cache Management
+
+**Release Date:** February 5, 2026
+
+### 🎯 Why v3.3.0?
+
+v3.3.0 transforms AgentVibes into a **universal TTS platform** for any environment:
+
+- **SSH-Remote Provider** - Generate TTS on servers, receive audio on your phone/computer
+- **Termux/Android Support** - Native Piper TTS on mobile devices
+- **OpenClaw Integration** - Turn voiceless servers into Siri-like conversational AI
+- **AgentVibes Receiver** - Receive and play audio from remote servers on your device
+- **Smart Installer** - Auto-detects your environment (voiceless, GUI, Termux, SSH)
+- **Intelligent Cache Management** - Real-time tracking and auto-cleanup prevents disk bloat
+
+#### 🌐 Real-World Use Case: OpenClaw + AgentVibes Receiver
+
+You deploy OpenClaw on a voiceless Mac mini (or remote server) where users message you via WhatsApp, Telegram, or Discord. With v3.3.0:
+
+**Before AgentVibes Receiver:**
+- User messages: "Tell me a joke"
+- Mac mini processes request
+- Text response appears in chat
+- 😞 No audio - silent experience
+
+**After AgentVibes Receiver:**
+1. **Install AgentVibes** on your Mac mini (or remote server)
+2. **Install AgentVibes Receiver** on your phone/iPad/laptop
+3. **Connect via Tailscale** (one-time setup)
+4. **User messages:** "Tell me a joke"
+5. **Mac mini generates TTS** with your configured voice
+6. **Audio streams to your device** via SSH tunnel
+7. **Your speakers play:** 🔊 "Why did the AI go to school? To improve its learning model!"
+8. **User in WhatsApp also hears** the audio playing (Siri-like experience)
+
+Result: OpenClaw transforms from **silent text AI** → **Conversational AI with voice**
+
+Perfect for:
+- 🖥️ Mac mini with OpenClaw
+- 🖥️ Remote servers (AWS, DigitalOcean, Linode)
+- 🏗️ Container deployments (Docker)
+- 🔧 WSL (Windows Subsystem for Linux)
+- 📱 Any voiceless environment needing audio
+
+### 🤖 AI Summary
+
+AgentVibes v3.3.0 unleashes the platform across new frontiers: remote servers via SSH-PulseAudio tunneling, Android/Termux environments with native Piper support, and OpenClaw (formerly Clawdbot) multi-agent orchestration. The redesigned smart installer detects your environment (voiceless, GUI, SSH, Termux) and shows only relevant options, plus optional BMAD personality injection for advanced users. Every TTS output now displays real-time cache metrics (file count/size with dynamic colors) plus intelligent size-based auto-cleanup that deletes oldest files when the cache exceeds threshold. The release includes comprehensive TTS queue management to prevent audio overlap, audio effects support across all providers, and full MCP tool integration for programmatic control. This release transforms AgentVibes into a universal TTS platform.
+
+**Key Highlights:**
+- 🌍 **SSH-Remote TTS** - Remote device playback via PulseAudio tunneling (servers, containers, WSL)
+- 📱 **Android/Termux Support** - Native Piper TTS on Android with termux-media-player integration
+- 🤖 **OpenClaw Receiver** (formerly Clawdbot) - AgentVibes Receiver for receiving TTS from voiceless servers
+- 🧠 **Smart Installer** - Voiceless environment detection + personality injection for BMAD
+- 📊 **Real-Time Cache Tracking** - File count and size on every output with dynamic colors
+- 🧹 **Intelligent Auto-Cleanup** - Size-based threshold (15MB default) prevents storage bloat
+- 🎵 **Queue Management** - Prevents TTS audio overlap via centralized queue system
+- ⚙️ **Audio Effects** - Full support across SSH-remote, Termux-ssh, and local providers
+- 📁 **Uninstall Command** - Comprehensive cleanup with full documentation
+- ✅ **96 Commits** - Massive feature expansion with 213 BATS tests passing
+
+### ✨ New Features
+
+#### 🌍 Remote SSH TTS Support
+
+**SSH-Remote Provider:**
+- Play TTS on remote servers via SSH + PulseAudio tunneling
+- Zero-dependency for audio output (uses PulseAudio network tunnel)
+- Perfect for deployed Claude Code on servers, containers, WSL
+- Auto-configuration of PulseAudio TCP module
+- Fallback to local playback if SSH unavailable
+- Full compatibility with all voice selection and audio effects
+
+**SSH-PulseAudio Integration:**
+- Automatic SSH connection detection and setup
+- Secure TCP tunnel for audio stream transmission
+- Support for both interactive and batch TTS operations
+- Persistent audio configuration per SSH session
+
+#### 📱 Android/Termux Support
+
+**Termux-SSH Provider:**
+- Native Piper TTS on Android via Termux environment
+- Uses termux-media-player for audio playback
+- Full voice selection and effects support
+- Automatic temp directory detection
+- Integration with Tailscale for secure remote access
+- Comprehensive setup guide with QR codes
+
+**Android Installation:**
+- Self-contained Termux installer script
+- One-command setup: `curl https://agentvibes.org/install-android | bash`
+- Automatic dependency detection and installation
+- Piper voice download management
+
+#### 🎙️ OpenClaw Integration & AgentVibes Receiver
+
+**What is AgentVibes Receiver?**
+
+AgentVibes Receiver is a **lightweight audio client** that receives and plays TTS audio from remote servers where OpenClaw is installed. It runs on your phone, tablet, or personal computer and connects to voiceless servers via SSH tunnel.
+
+**The Problem It Solves:**
+- OpenClaw running on Mac mini/remote server has no audio output
+- Users message via WhatsApp/Telegram/Discord - get text responses only
+- 😞 No voice = Less engaging AI experience
+
+**AgentVibes Receiver Solution:**
+1. **Lightweight client** runs on your device (phone/tablet/laptop)
+2. **SSH tunnel** securely connects to your voiceless server
+3. **Audio streams** from server to your device via PulseAudio
+4. **Auto-plays** on your speakers when OpenClaw responds
+5. **Siri-like experience** - Text + Voice in one flow
+
+**How It Works:**
+
+```
+┌──────────────────────────────┐
+│ Your Mac mini / Server       │
+│ (OpenClaw + AgentVibes)      │
+│ ├─ No audio output           │
+│ ├─ Generates TTS             │
+│ └─ Sends via SSH tunnel      │
+└──────────────────────────────┘
+            ↓ SSH Tunnel (encrypted)
+┌──────────────────────────────┐
+│ Your Phone / Laptop          │
+│ (AgentVibes Receiver)        │
+│ ├─ Receives audio stream     │
+│ ├─ Plays on speakers         │
+│ └─ You hear OpenClaw speak   │
+└──────────────────────────────┘
+```
+
+**Example Flow:**
+```
+WhatsApp: "Tell me a joke"
+        ↓
+Mac mini: Processes with Claude
+        ↓
+Generates TTS: "Why did the AI... [audio file]"
+        ↓ SSH tunnel
+Your Phone: Plays audio 🔊
+        ↓
+You hear: "Why did the AI go to school?"
+```
+
+**AgentVibes Receiver Features:**
+- **One-Time Setup** - Pair with server via SSH key
+- **Automatic Connection** - Reconnects if interrupted
+- **Real-Time Streaming** - Low latency audio playback
+- **SSH Encryption** - Secure tunnel for audio
+- **Tailscale Support** - Easy VPN for remote servers
+- **Multiple Servers** - Connect to different OpenClaw instances
+- **Voice Control** - Full voice selection on the server side
+- **Cache Metrics** - Monitor audio generation and cleanup
+
+**OpenClaw Skill Integration:**
+- Installed automatically with AgentVibes on OpenClaw server
+- Full feature access:
+  - Voice selection (50+ voices)
+  - Personality/sentiment (sarcastic, flirty, etc.)
+  - Audio effects (reverb, echo, pitch)
+  - Speech speed (0.5x - 3.0x)
+  - Language translation (speak in different languages)
+  - Real-time cache tracking
+  - Automatic cleanup of old audio files
+
+#### 🧠 Smart Installer Enhancements
+
+**Voiceless Environment Detection:**
+- Auto-detects if GUI audio is unavailable (headless servers, containers)
+- Offers SSH-remote TTS as alternative for voiceless environments
+- Prevents installation of unnecessary audio dependencies
+
+**Personality Injection (BMAD):**
+- Interactive prompt during install for BMAD users
+- Optional TTS personality configuration
+- Sentiment/personality selection built into setup flow
+- Skipped for non-BMAD environments
+
+**Provider Auto-Selection:**
+- Intelligent detection of available providers:
+  - macOS Say (macOS systems)
+  - Piper TTS (all systems)
+  - SSH-remote (if SSH available)
+  - Termux-ssh (Android/Termux)
+- Shows only relevant providers in installation
+
+**Better UX:**
+- Clear descriptions of each provider
+- Setup URLs for complex providers (Tailscale)
+- Comprehensive help text for each option
+- Git log integration for recent changes
+
+#### 📊 Real-Time TTS Cache Tracking & Intelligent Auto-Cleanup
+
+**Why Cache Management Matters:**
+- TTS audio files accumulate quickly
+- Server deployments can run out of disk space silently
+- Users have no visibility into cache size or cleanup status
+- Manual cleanup is inconvenient and error-prone
+
+**Cache Display on Every Output:**
+Every time you generate TTS, you see real-time cache metrics:
+```
+💾 Saved to: /home/user/.claude/audio/tts-1770274925.wav 📦 28 20.9MB 🧹[15mb]
+```
+
+What you see:
+- 💾 **Full path** - Clickable file for replay or sharing
+- 📦 **28** - File count in cache
+- **20.9MB** - Total cache size (color-coded):
+  - 🟢 Green: <500MB
+  - 🟡 Yellow: 500MB-3GB
+  - 🔴 Red: >3GB
+- 🧹 **[15mb]** - Auto-cleanup threshold
+
+**Intelligent Size-Based Auto-Cleanup:**
+- Deletes oldest files when cache exceeds threshold (default: 15MB)
+- Silent operation (no blocking prompts)
+- Write-lock protection prevents conflicts with TTS generation
+- Respects active TTS (won't delete while generating)
+
+**Configuration:**
+```bash
+# Per-project threshold override
+echo "50" > .claude/tts-auto-clean-threshold.txt  # 50MB limit
+
+# Or disable cleanup
+echo "0" > .claude/tts-auto-clean-threshold.txt   # Disable
+```
+
+**Manual Cleanup:**
+```bash
+# Non-interactive cleanup
+/agent-vibes:clean
+
+# Or programmatically via MCP
+await agent_vibes.clean_audio_cache()
+```
+
+#### 🎵 TTS Queue Management
+
+**Overlap Prevention:**
+- Centralized queue system for TTS operations
+- Prevents simultaneous audio playback
+- Critical for Clawdbot multi-agent scenarios
+- Atomic queue operations ensure consistency
+
+**Queue Integration:**
+- Automatic in OpenClaw Receiver
+- Optional in standalone environments
+- Fallback to sequential playback
+
+#### ⚙️ Audio Effects Across All Providers
+
+**Effects Support:**
+- Reverb, echo, pitch, EQ available
+- SSH-remote provider: Full effects support
+- Termux-ssh provider: Full effects support
+- All local providers: Unchanged effects behavior
+
+**Configuration:**
+- Per-session override via environment variables
+- Project-local settings via config files
+- Persistent across TTS operations
+
+#### 📁 Comprehensive Uninstall Command
+
+**`/agent-vibes:uninstall` Skill:**
+- Complete removal of AgentVibes and dependencies
+- Interactive prompts for user confirmation
+- Option to preserve configuration
+- Detailed removal logs
+- Full documentation included
+
+### 🐛 Bug Fixes
+
+- **TTS Overlap** - Fixed audio overlapping via queue management
+- **Termux Audio** - Proper detection and use of termux-media-player
+- **SSH Detection** - Improved SSH environment detection logic
+- **Race Conditions** - Write-lock mechanism prevents cleanup conflicts
+- **Temp Directory** - Proper Termux temp directory handling
+- **Color Codes** - Fixed GOLD color (256-color \033[38;5;226m)
+- **Stat Compatibility** - BSD/GNU stat detection with proper output suppression
+- **Syntax Validation** - Fixed installer.js syntax errors
+- **Coverage Testing** - Proper coverage file generation for CI/CD
+
+### 🔒 Security & Quality
+
+- **No Hardcoded Credentials** - All secure operations use environment variables
+- **SSH Safety** - Secure PulseAudio tunnel authentication
+- **Atomic Operations** - Queue and receiver use atomic file operations
+- **Input Validation** - All external inputs validated
+- **Pre-existing Limitations** - TTS scripts lack `set -euo pipefail` (pre-existing)
+- **Sonar Compliance** - Security hotspots resolved, quality gates passing
+- **Test Coverage** - 213 BATS tests + 47 Node unit tests
+
+### ✅ Testing & Validation
+
+- **213 BATS Tests** - Core functionality validation
+- **47 Node Tests** - JavaScript/installer validation
+- **Cross-Platform** - Piper, macOS, SSH-remote, Termux-ssh
+- **Environment Tests** - Voiceless, GUI, SSH, Termux detection
+- **Audio Effects** - All providers tested
+- **Backwards Compatible** - No breaking changes to existing code
+
+---
+
+## 📦 v3.2.0 - Clawdbot Integration: AI Assistants on Any Messenger
+
+**Release Date:** January 27, 2026
+
+### 🎯 Why v3.2.0?
+
+This minor release adds **native Clawdbot integration** to AgentVibes, bringing professional TTS to the revolutionary AI assistant you can access via any instant messenger. Clawdbot connects Claude AI to WhatsApp, Telegram, Discord, and more—and now with AgentVibes, your Clawdbot can speak with 50+ professional voices in 30+ languages. This release also includes SonarCloud quality gate improvements and CI/CD workflow enhancements.
+
+### 🤖 AI Summary
+
+AgentVibes v3.2.0 introduces seamless integration with Clawdbot, the revolutionary AI assistant accessible via any instant messenger. With this release, Clawdbot users get professional TTS with 50+ voices, remote SSH audio support for server deployments, and zero-configuration setup—just install AgentVibes and the Clawdbot skill is ready. The release also includes quality improvements: SonarCloud workflow fixes, enhanced documentation for disabling quality gate checks, and improved test coverage validation.
+
+**Key Highlights:**
+- 🤖 **Clawdbot Integration** - Native TTS support for Clawdbot AI assistant framework
+- 💬 **Messenger Platforms** - Works with WhatsApp, Telegram, Discord via Clawdbot
+- 🔊 **Remote SSH Audio** - Perfect for Clawdbot on remote servers with PulseAudio tunneling
+- 📦 **Simple Install** - Just `npx agentvibes install` and it works
+- 🛡️ **SonarCloud Fixes** - Quality gate workflow improvements and documentation
+- ✅ **Full Test Coverage** - All 213 BATS + 47 Node tests passing
+
+### ✨ New Features
+
+**Clawdbot Skill (`.clawdbot/`):**
+- New `.clawdbot/` directory with skill integration files
+- `README.md` - Clawdbot integration overview and setup guide
+- `skill/SKILL.md` - Comprehensive skill documentation with voice selection, background music, effects, personalities, and remote SSH audio setup
+- Automatically distributed via npm package
+- Zero-configuration when AgentVibes is installed
+
+**README Updates:**
+- Added "🤖 Clawdbot Integration" section with full documentation
+- Updated header to include Clawdbot alongside Claude Code, Claude Desktop, and Warp Terminal
+- Added Clawdbot to Quick Links table
+- Clawdbot description: "A revolutionary AI assistant you can access via any instant messenger"
+- Website link: https://clawd.bot
+
+**package.json Updates:**
+- Added `.clawdbot/` to npm files array for distribution
+- Added `clawdbot` to keywords for npm discoverability
+- Updated description to mention Clawdbot support
+
+### 🐛 Bug Fixes
+
+- **SonarCloud Quality Gate** - Disabled quality gate status reporting to GitHub to prevent false CI failures
+- **Coverage File Generation** - Ensured coverage file is generated before SonarCloud scan
+- **CLI Test Coverage** - Added CLI tests and excluded CLI entry point from coverage requirements
+- **macOS Runner** - Removed macos-15-large runner to avoid GitHub billing limits
+- **Piper Voice Test** - Updated installation test to match current voice list
+
+### 📚 Documentation
+
+- Added step-by-step SonarCloud dashboard configuration guide
+- Added guide to disable SonarCloud GitHub App checks
+- Comprehensive Clawdbot integration documentation with SSH audio examples
+
+### 🔒 Security & Quality
+
+- ✅ All Sonar quality gates validated
+- ✅ No hardcoded credentials in changes
+- ✅ New Clawdbot files are documentation only (no executable code)
+- ✅ All 213 BATS + 47 Node tests passing
+
+### 📊 Changes Summary
+
+- **Files Added:** 2 (`.clawdbot/README.md`, `.clawdbot/skill/SKILL.md`)
+- **Files Modified:** 2 (`README.md`, `package.json`)
+- **Commits Since v3.1.0:** 11 (5 fixes, 4 docs, 1 test, 1 debug)
+
+### 🎯 User Impact
+
+**For Clawdbot Users:**
+- Professional TTS with 50+ voices in 30+ languages
+- Works on remote servers with SSH audio tunneling
+- Zero API costs—Piper TTS is free and offline
+- Automatic integration when AgentVibes is installed
+
+**For Existing Users:**
+- Zero breaking changes
+- All existing features work exactly the same
+- Clawdbot support is additive
+
+### 🚀 Migration Notes
+
+No migration required! This is a fully backward-compatible minor release.
+
+**To Use with Clawdbot:**
+1. Install: `npx agentvibes install`
+2. Speak: `npx agentvibes speak "Hello!"`
+
+### 📦 Full Changelog
+
+**Feature Commits:**
+- `(pending)` feat: Add Clawdbot integration
+
+**Bug Fix Commits:**
+- `5cd97d52` fix: Disable SonarCloud quality gate status reporting to GitHub
+- `12f822e6` fix: Disable quality gate failure in SonarCloud workflow
+- `0d26ccc2` fix: Ensure coverage file is generated before SonarCloud scan
+- `c2465508` fix: Add CLI tests and exclude CLI entry point from coverage
+- `c673afe1` fix: Remove macos-15-large runner to avoid GitHub billing limits
+- `92271732` fix: Update Piper installation test to match current voice list
+
+**Documentation Commits:**
+- `f72dd977` docs: Add guide to disable SonarCloud GitHub App checks
+- `6587519b` docs: Add step-by-step SonarCloud dashboard configuration guide
+- `ba765f50` docs: Add SonarCloud quality gate configuration guidance
+
+**Test Commits:**
+- `47f08a79` test: Trigger workflow to verify SonarCloud quality gate fix
+
+**Debug Commits:**
+- `84945d25` debug: Add coverage file verification to SonarCloud workflow
+
+---
+
+## 📦 v3.1.0 - Android Native Support: Run Claude Code on Your Phone
+
+**Release Date:** January 22, 2026
+
+### 🎯 Why v3.1.0?
+
+This minor release brings **native Android support** to AgentVibes, enabling developers to run Claude Code with professional TTS voices directly on Android devices via Termux. No SSH required, no remote server needed—just install Termux on your Android phone or tablet and get the full AgentVibes experience locally. This complements the v3.0.0 termux-ssh provider by offering a **complete mobile development solution**: use native Termux for local Android development, or use termux-ssh when connecting to remote servers.
+
+### 🤖 AI Summary
+
+AgentVibes v3.1.0 introduces native Android/Termux support, enabling developers to run Claude Code with professional TTS voices directly on their Android devices. Through automatic detection and a specialized installer, AgentVibes now runs Piper TTS via proot-distro with Debian (solving Android's glibc compatibility issues), uses termux-media-player for audio playback, and includes comprehensive Android-specific documentation. Perfect for developers who want to code on-the-go with their Android phone or tablet using the full power of Claude Code and AgentVibes.
+
+**Key Highlights:**
+- 🤖 **Native Android Support** - Run Claude Code with TTS directly on Android devices via Termux
+- 📦 **Automatic Termux Detection** - AgentVibes auto-detects Android and runs specialized installation
+- 🎯 **Proot-Distro Integration** - Solves glibc compatibility with proot Debian environment
+- 🔊 **Android Audio Playback** - Uses termux-media-player for native Android audio routing
+- 📚 **Comprehensive Documentation** - Complete Android setup guide with troubleshooting and F-Droid instructions
+- ✅ **Full Test Coverage** - All 213 BATS + 38 Node tests passing with Android compatibility
+
+### ✨ New Features
+
+**Termux Installer (`.claude/hooks/termux-installer.sh`):**
+- New 224-line installer specifically for Android/Termux environments
+- Automatically installs proot-distro with Debian (for glibc compatibility)
+- Downloads and configures Piper TTS binary in proot environment
+- Creates `/usr/bin/piper` wrapper that routes through proot
+- Installs audio dependencies: ffmpeg, sox, bc, termux-api
+- Interactive voice selection with 50+ language options
+- Validates Termux environment before proceeding
+
+**Termux Detection (`src/installer.js`):**
+- New `isTermux()` function checks for `/data/data/com.termux` directory
+- New `detectAndNotifyTermux()` displays Android detection messages
+- Auto-configures piper provider with Termux-compatible voice
+- Shows Termux-specific installation instructions
+- Piper installer automatically redirects to termux-installer.sh on Android
+
+**Audio Processor Updates (`.claude/hooks/audio-processor.sh`):**
+- Detects Termux environment for temp directory selection
+- Uses `${PREFIX}/tmp` on Termux, `/tmp` on standard systems
+- Ensures audio effects work correctly on Android
+- Cross-platform compatibility maintained
+
+**Piper Installer Updates (`.claude/hooks/piper-installer.sh`):**
+- Auto-detects Termux and redirects to specialized installer
+- Shows clear message when routing to Termux-specific setup
+
+**Android Audio Playback (`.claude/hooks/play-tts-piper.sh`):**
+- Detects Android/Termux environment
+- Uses `termux-media-player` instead of `paplay` on Android
+- Audio routes through Android's native media system
+
+### 📚 Documentation
+
+**New Android Setup Section (`README.md`):**
+- Added "🤖 Android / Termux" section to System Requirements
+- Complete 3-step installation guide for Android devices
+- Explanation of why Termux needs special handling (bionic vs glibc)
+- Requirements: Termux app from F-Droid, Termux:API, Android 7.0+
+- Audio playback architecture explanation
+- Setup verification commands
+- Troubleshooting table for common issues
+- Clear explanation of why F-Droid version is required (not Google Play)
+- Updated Quick Links table with direct link to Android setup
+
+### 🐛 Bug Fixes
+
+- **Test #90 Fix** - Added termux-ssh provider to test cleanup list for "no providers found" edge case
+- **Temp Directory Detection** - Fixed audio-processor.sh defaulting to Termux paths on non-Termux systems
+- **Sonar Compliance** - Added `set -euo pipefail` strict mode to termux-installer.sh for security
+
+### 🔒 Security & Quality
+
+- ✅ All Sonar quality gates validated
+- ✅ Strict mode (`set -euo pipefail`) on all new bash scripts
+- ✅ No hardcoded credentials
+- ✅ Proper variable quoting and input validation
+- ✅ Cross-platform temp directory handling
+- ✅ All 213 BATS + 38 Node tests passing
+
+### 📊 Changes Summary
+
+- **Files Modified:** 7
+- **Lines Added:** +391
+- **Lines Removed:** -8
+- **New Files:** 1 (termux-installer.sh)
+- **Commits:** 8 (5 fixes, 1 feature, 1 docs, 1 merge)
+
+### 🎯 User Impact
+
+**For Android Users:**
+- Can now run Claude Code directly on Android phones/tablets
+- Full TTS support with 50+ voices and languages
+- No remote server required for basic usage
+- Works offline after initial voice downloads
+
+**For Developers:**
+- Complete mobile development solution (native + SSH)
+- Native Termux for local Android development
+- Termux-SSH provider for remote server connections
+- Seamless integration with existing AgentVibes workflows
+
+**For Existing Users:**
+- Zero breaking changes
+- All existing features work exactly the same
+- New Android support is opt-in via Termux installation
+
+### 🚀 Migration Notes
+
+No migration required! This is a fully backward-compatible minor release.
+
+**To Try Android Support:**
+1. Install [Termux from F-Droid](https://f-droid.org/en/packages/com.termux/)
+2. Install [Termux:API](https://f-droid.org/en/packages/com.termux.api/)
+3. In Termux: `pkg install nodejs-lts`
+4. Run: `npx agentvibes install`
+
+AgentVibes will auto-detect Termux and run the specialized installer.
+
+### 📦 Full Changelog
+
+**Feature Commits:**
+- `e9d4cf95` feat: Add Android/Termux support for Piper TTS
+
+**Bug Fix Commits:**
+- `aa4d3cdd` fix: Add termux-ssh provider to test #90 cleanup list
+- `c1b00c6d` fix: Use termux-media-player for audio playback on Android
+- `f96ab89a` fix: Properly detect Termux environment for temp directory
+- `e2efeb06` fix: Add strict mode to termux-installer.sh for Sonar compliance
+
+**Documentation Commits:**
+- `701a9412` docs: Add comprehensive Android/Termux setup section to README
+
+**Merge Commits:**
+- `a5d3f546` Merge feature/android-termux into master
+- `95f04e70` Merge origin/master into feature/pulseaudio-reverse-tunnel
+
+---
+
+## 📦 v3.0.0 - Cross-Platform Remote Audio: Termux SSH Provider
+
+**Release Date:** January 8, 2026
+
+### 🎯 Why v3.0.0?
+
+This major release marks a significant milestone in AgentVibes' evolution, introducing **mobile-first interactive AI conversations**. The termux-ssh provider enables a revolutionary workflow: **have fully interactive, hands-free conversations with Claude Code using just your mobile device**—whether you're coding locally on your laptop with audio routed to your phone, or working on remote servers from anywhere in the world. This architectural breakthrough represents a new paradigm: **"Code with your hands, converse with your voice."**
+
+### 🤖 AI Summary
+
+AgentVibes v3.0.0 introduces the termux-ssh TTS provider, enabling **true mobile-first interactive conversations with Claude Code**. Route TTS audio to your Android device via SSH—whether coding locally on your laptop or on remote servers—and have hands-free, voice-driven conversations with Claude using just your phone. This major release includes comprehensive Tailscale VPN setup documentation for internet-wide access, full MCP server integration, and transforms how developers interact with AI assistants. Perfect for developers who want to experience AI conversations naturally through their mobile device while their hands stay on the keyboard.
+
+**Key Highlights:**
+- 📱 **Mobile-First AI Conversations** - Have fully interactive, hands-free conversations with Claude Code using just your Android device
+- 💻 **Local + Remote Development** - Works for both local coding (laptop → phone audio) and remote server development
+- 🌐 **Tailscale Integration** - Comprehensive guide for internet-wide device access without port forwarding or firewall configuration
+- 🔧 **Enhanced Installer** - Interactive SSH host configuration with validation and clear use-case guidance
+- 🎯 **Full MCP Compatibility** - Complete integration with all MCP commands and workflows
+- 🛡️ **Quality Gates Integration** - Automated security validation in release process
+
+### 🎥 Demo Video
+
+**Watch it in action:** [Mobile-First AI Conversations with Claude Code](https://youtu.be/ngLiA_KQtTA?si=wTwS4CJidIxWqLIP)
+
+See the termux-ssh provider in action—fully interactive, hands-free conversations with Claude Code using just your Android device.
+
+### ✨ New Features
+
+**Termux SSH TTS Provider (`.claude/hooks/play-tts-termux-ssh.sh`):**
+- New TTS provider for Android via SSH connection
+- Routes text to `termux-tts-speak` on remote Android device
+- Configuration priority: env var → project → global
+- Secure quote escaping for safe text transmission
+- 196 lines of fully documented code
+
+**Installer Updates (`src/installer.js`):**
+- Added termux-ssh to provider selection menu
+- Interactive SSH host alias configuration with validation
+- Saves host alias to `.claude/termux-ssh-host.txt`
+- Clear use case description: "Only choose if your project is on a remote server and you want audio sent to your Android device"
+- Documentation link to TERMUX_SETUP.md
+
+**TTS Router Updates (`.claude/hooks/play-tts.sh`):**
+- Added termux-ssh provider routing in two locations
+- Full integration with existing provider detection
+- Supports mixed-provider mode (Piper + Termux)
+
+**MCP Server Integration (`mcp-server/server.py`):**