verbalcoding 0.2.7 → 0.2.8

package/README.md

@@ -34,7 +34,7 @@ VerbalCoding turns a Discord voice channel into a hands-free control surface for
 | What you get | Why it feels good |
 |---|---|
 | Voice-first agent control | Talk to Hermes Agent, Claude Code, Codex, Gemini CLI, OpenCode, OpenClaw, or any custom CLI harness. |
-| Local-first speech loop | Discord voice capture → `whisper.cpp` STT → agent → chunked TTS playback. |
+| On-device speech loop | Discord voice capture → local `whisper-cli` transcription → agent → chunked TTS playback. |
 | Shared voice + text context | Voice turns and `!ask` text commands can reuse the same supported agent session. |
 | Barge-in and sensitivity modes | Interrupt playback naturally and switch between normal and conservative/noisy environments. |
 | Multilingual voice presets | Switch STT, progress language, and TTS voice together with `vc language ko/en/auto`. |
@@ -69,23 +69,10 @@ vc doctor
 ./run.sh
 ```
 
-`vc setup --yes` and `./scripts/install.sh --yes` bootstrap local prerequisites where possible: Node/npm dependencies, `ffmpeg`, `whisper-cli`, the default whisper.cpp model, a local `.venv-tts` Edge TTS helper, and the short `vc` shell command for clone installs. They support macOS/Homebrew plus common Linux package managers (`apt`, `dnf`, `pacman`); rerun with `--no-wizard` for dependency-only setup or `--skip-system` if you want to install OS packages yourself.
+`vc setup --yes` bootstraps local prerequisites from the npm package. `./scripts/install.sh --yes` does the same for GitHub clone installs. Both cover Node/npm dependencies, `ffmpeg`, `whisper-cli`, the default whisper.cpp model, a local `.venv-tts` Edge TTS helper, and setup wizard configuration where possible. They support macOS/Homebrew plus common Linux package managers (`apt`, `dnf`, `pacman`); rerun with `--no-wizard` for dependency-only setup or `--skip-system` if you want to install OS packages yourself.
 
 Need a clean install walkthrough? Start with [Fresh Install](docs/FRESH_INSTALL.md).
 
-## How It Works
-
-```mermaid
-flowchart LR
-  A[Discord voice] --> B["@discordjs/voice"]
-  B --> C[PCM cleanup + gates]
-  C --> D["whisper.cpp STT"]
-  D --> E["CLI agent adapter"]
-  E --> F["Concise answer"]
-  F --> G["Chunked TTS"]
-  G --> H["Discord playback"]
-```
-
 ## Supported Agent Backends
 
 | Backend | Default command | Session support |
@@ -107,12 +94,6 @@ flowchart LR
 | [Configuration](docs/CONFIGURATION.md) | `.env`, agent backends, MCP, TTS backends, operational notes |
 | [Multi-Instance](docs/MULTI_INSTANCE.md) | One permanent Discord voice room per project |
 | [Release Notes](docs/RELEASE.md) | Current capabilities and pre-release checklist |
-| [한국어 문서](docs/i18n/README.ko.md) | npm 설치, 사용법, 설정, 멀티 인스턴스 한국어 가이드 |
-| [日本語 docs](docs/i18n/README.ja.md) | npm install, usage, configuration, multi-instance guide in Japanese |
-| [中文文档](docs/i18n/README.zh.md) | npm 安装、使用、配置和多实例中文指南 |
-| [Español docs](docs/i18n/README.es.md) | Instalación npm, uso, configuración y multiinstancia en español |
-| [Français docs](docs/i18n/README.fr.md) | Installation npm, utilisation, configuration et multi-instance en français |
-| [Русская документация](docs/i18n/README.ru.md) | npm установка, использование, конфигурация и мульти-инстансы на русском |
 
 ## Tiny Command Map
 
@@ -128,11 +109,15 @@ vc start                  # start the default bridge
 
 In Discord:
 
-```text
-!join        !ask <prompt>       !verbose on/off
-!latency     !sensitivity normal !sensitivity conservative
-!session new <name> <workdir> [context] --voice <voice-channel>
-```
+| Command | What it does |
+|---|---|
+| `!join` | Join your current voice channel. |
+| `!ask <prompt>` | Send text to the same agent backend. |
+| `!verbose on\|off` | Show/speak short progress updates. |
+| `!latency` | Summarize recent voice/STT/agent/TTS latency. |
+| `!sensitivity normal` | Use normal indoor barge-in sensitivity. |
+| `!sensitivity conservative` | Use stricter noisy/outdoor sensitivity. |
+| `!session new <name> <workdir> [context] --voice <voice-channel>` | Bind a project session to a voice room. |
 
 ## Requirements
 
@@ -140,7 +125,7 @@ In Discord:
 |---|---|
 | Runtime | Node.js 20+, npm; install script can install via Homebrew/apt/dnf/pacman |
 | Audio | `ffmpeg`; install script can install it |
-| STT | `whisper.cpp` / `whisper-cli`; install script uses Homebrew on macOS or local Linux build fallback |
+| Speech recognition | Local `whisper-cli` from whisper.cpp; install script uses Homebrew on macOS or local Linux build fallback |
 | TTS | Edge TTS CLI; install script creates `.venv-tts` if needed |
 | Discord | Bot token, Message Content intent, voice permissions |
 | Agent | At least one authenticated CLI harness, Hermes Agent by default |

package/app-node/cli_install.test.mjs

@@ -63,6 +63,8 @@ test('bootstrap script installs cross-platform prerequisites and local model hel
 
   assert.match(script, /brew install/);
   assert.match(script, /apt-get install/);
+  assert.match(script, /has_cmd node \|\| packages\+\=\(nodejs\)/);
+  assert.match(script, /has_cmd npm \|\| packages\+\=\(npm\)/);
   assert.match(script, /dnf install/);
   assert.match(script, /pacman -Sy/);
   assert.match(script, /git clone --depth 1 https:\/\/github\.com\/ggml-org\/whisper\.cpp\.git/);
@@ -70,6 +72,19 @@ test('bootstrap script installs cross-platform prerequisites and local model hel
   assert.match(script, /\.venv-tts/);
 });
 
+test('doctor auto-bootstraps fixable prerequisites by default', () => {
+  const doctor = fs.readFileSync(path.join(ROOT, 'scripts', 'doctor.mjs'), 'utf8');
+  const cli = fs.readFileSync(path.join(ROOT, 'scripts', 'cli.mjs'), 'utf8');
+
+  assert.match(doctor, /fixablePrerequisites/);
+  assert.match(doctor, /bootstrap_prereqs\.sh'\), '--yes'/);
+  assert.match(doctor, /VERBALCODING_DOCTOR_AUTO_FIX/);
+  assert.match(doctor, /--no-fix/);
+  assert.match(doctor, /WHISPER_CPP_BIN/);
+  assert.match(doctor, /EDGE_TTS_COMMAND/);
+  assert.match(cli, /doctor\.mjs'\), \.\.\.argv\.slice\(1\)/);
+});
+
 test('Ubuntu Docker smoke script validates clean install without secrets', () => {
   const script = fs.readFileSync(path.join(ROOT, 'scripts', 'docker_ubuntu_smoke.sh'), 'utf8');
 

package/docs/FRESH_INSTALL.md

@@ -32,7 +32,13 @@ cd VerbalCoding
 
 ## 2. Bootstrap dependencies and run the setup wizard
 
-The npm commands above run the same bootstrapper as the clone install. For a clone, run:
+For an npm install, do not run `./scripts/install.sh` directly; there is no repository checkout in your current directory. Use the packaged CLI wrapper instead:
+
+```bash
+vc setup --yes
+```
+
+`vc setup` runs the `scripts/install.sh` bundled inside the installed npm package. Only use `./scripts/install.sh --yes` when you are inside a GitHub clone:
 
 ```bash
 ./scripts/install.sh --yes
@@ -104,7 +110,7 @@ The invite includes bot and slash-command scopes plus text/voice permissions use
 vc doctor
 ```
 
-`vc doctor` is redacted: it reports missing tokens/commands/models without printing secret values. Fix every `✗` item, then rerun it.
+`vc doctor` is redacted: it reports missing tokens/commands/models without printing secret values. When fixable local prerequisites are missing (`ffmpeg`, `whisper-cli`, the default model, or Edge TTS helper), it automatically reruns the packaged bootstrap first. Fix any remaining `✗` items, then rerun it.
 
 Expected success includes:
 

package/docs/assets/figures/verbalcoding-flow.svg

@@ -1,6 +1,6 @@
 <svg width="1200" height="520" viewBox="0 0 1200 520" fill="none" xmlns="http://www.w3.org/2000/svg" role="img" aria-labelledby="title desc">
-  <title id="title">VerbalCoding voice-to-agent flow</title>
-  <desc id="desc">A stylized pipeline from Discord voice to speech recognition, CLI agent, text answer, and TTS playback.</desc>
+  <title id="title">VerbalCoding natural voice loop</title>
+  <desc id="desc">A compact phone-call-like loop: user speaks in Discord, Local STT with whisper-cli transcribes, the CLI agent works, TTS speaks back, and the user can interrupt anytime.</desc>
   <defs>
     <linearGradient id="bg" x1="0" y1="0" x2="1200" y2="520" gradientUnits="userSpaceOnUse">
       <stop stop-color="#0F172A"/>
@@ -19,45 +19,60 @@
   <circle cx="1030" cy="90" r="190" fill="#6366F1" opacity="0.16"/>
   <circle cx="170" cy="430" r="210" fill="#06B6D4" opacity="0.13"/>
   <rect x="70" y="54" width="1060" height="412" rx="32" fill="url(#card)" stroke="#334155" filter="url(#shadow)"/>
+
   <text x="110" y="118" fill="#F8FAFC" font-family="Inter, ui-sans-serif, system-ui" font-size="42" font-weight="800">VerbalCoding</text>
-  <text x="110" y="154" fill="#94A3B8" font-family="Inter, ui-sans-serif, system-ui" font-size="20">Discord voice → local STT → CLI coding agent → spoken answer</text>
+  <text x="110" y="154" fill="#94A3B8" font-family="Inter, ui-sans-serif, system-ui" font-size="20">A natural Discord voice loop for coding agents — speak, listen, interrupt, continue</text>
 
   <g font-family="Inter, ui-sans-serif, system-ui" font-size="17" font-weight="700">
-    <rect x="110" y="220" width="150" height="92" rx="20" fill="#5865F2"/>
-    <text x="185" y="258" fill="white" text-anchor="middle">Discord</text>
-    <text x="185" y="284" fill="#E0E7FF" text-anchor="middle" font-size="14">voice channel</text>
+    <rect x="105" y="220" width="160" height="92" rx="20" fill="#5865F2"/>
+    <text x="185" y="254" fill="white" text-anchor="middle">Discord</text>
+    <text x="185" y="280" fill="#E0E7FF" text-anchor="middle" font-size="14">phone-call voice</text>
 
-    <rect x="305" y="220" width="150" height="92" rx="20" fill="#0891B2"/>
-    <text x="380" y="258" fill="white" text-anchor="middle">whisper.cpp</text>
-    <text x="380" y="284" fill="#CFFAFE" text-anchor="middle" font-size="14">local STT</text>
+    <rect x="305" y="220" width="165" height="92" rx="20" fill="#0891B2"/>
+    <text x="387.5" y="254" fill="white" text-anchor="middle">Local STT</text>
+    <text x="387.5" y="280" fill="#CFFAFE" text-anchor="middle" font-size="14">whisper-cli</text>
 
-    <rect x="500" y="220" width="150" height="92" rx="20" fill="#7C3AED"/>
-    <text x="575" y="258" fill="white" text-anchor="middle">Adapter</text>
-    <text x="575" y="284" fill="#EDE9FE" text-anchor="middle" font-size="14">Hermes / Claude / Codex</text>
+    <rect x="510" y="220" width="165" height="92" rx="20" fill="#7C3AED"/>
+    <text x="592.5" y="254" fill="white" text-anchor="middle">Adapter</text>
+    <text x="592.5" y="280" fill="#EDE9FE" text-anchor="middle" font-size="14">Hermes / Claude / Codex</text>
 
-    <rect x="695" y="220" width="150" height="92" rx="20" fill="#111827" stroke="#475569"/>
-    <text x="770" y="258" fill="white" text-anchor="middle">CLI Agent</text>
-    <text x="770" y="284" fill="#CBD5E1" text-anchor="middle" font-size="14">does the work</text>
+    <rect x="715" y="220" width="165" height="92" rx="20" fill="#111827" stroke="#475569"/>
+    <text x="797.5" y="254" fill="white" text-anchor="middle">CLI Agent</text>
+    <text x="797.5" y="280" fill="#CBD5E1" text-anchor="middle" font-size="14">does the work</text>
 
-    <rect x="890" y="220" width="150" height="92" rx="20" fill="#0EA5E9"/>
-    <text x="965" y="258" fill="white" text-anchor="middle">TTS</text>
-    <text x="965" y="284" fill="#E0F2FE" text-anchor="middle" font-size="14">chunked playback</text>
+    <rect x="920" y="220" width="165" height="92" rx="20" fill="#0EA5E9"/>
+    <text x="1002.5" y="254" fill="white" text-anchor="middle">TTS</text>
+    <text x="1002.5" y="280" fill="#E0F2FE" text-anchor="middle" font-size="14">spoken reply</text>
   </g>
 
   <g stroke="#94A3B8" stroke-width="4" stroke-linecap="round">
-    <path d="M266 266H296"/>
-    <path d="M461 266H491"/>
-    <path d="M656 266H686"/>
-    <path d="M851 266H881"/>
+    <path d="M275 266H295"/>
+    <path d="M480 266H500"/>
+    <path d="M685 266H705"/>
+    <path d="M890 266H910"/>
+  </g>
+  <g fill="#94A3B8" opacity="0.95">
+    <circle cx="285" cy="266" r="4"/>
+    <circle cx="490" cy="266" r="4"/>
+    <circle cx="695" cy="266" r="4"/>
+    <circle cx="900" cy="266" r="4"/>
   </g>
-  <g fill="#94A3B8">
-    <path d="M296 266l-10-7v14l10-7z"/>
-    <path d="M491 266l-10-7v14l10-7z"/>
-    <path d="M686 266l-10-7v14l10-7z"/>
-    <path d="M881 266l-10-7v14l10-7z"/>
+
+  <path d="M1002 330C1002 405 185 405 185 330" stroke="#67E8F9" stroke-width="4" stroke-linecap="round" stroke-dasharray="13 13"/>
+  <g fill="#67E8F9">
+    <circle cx="1002" cy="330" r="5"/>
+    <circle cx="185" cy="330" r="5"/>
+  </g>
+  <text x="594" y="438" fill="#A5F3FC" text-anchor="middle" font-family="Inter, ui-sans-serif, system-ui" font-size="17" font-weight="700">Conversation loop: hear the answer, speak again, or interrupt anytime</text>
+
+  <path d="M185 210C185 178 1002 178 1002 210" stroke="#FBBF24" stroke-width="3" stroke-linecap="round" stroke-dasharray="8 10" opacity="0.9"/>
+  <g fill="#FBBF24" opacity="0.95">
+    <circle cx="185" cy="210" r="4"/>
+    <circle cx="1002" cy="210" r="4"/>
   </g>
+  <text x="594" y="194" fill="#FDE68A" text-anchor="middle" font-family="Inter, ui-sans-serif, system-ui" font-size="15" font-weight="700">Barge-in stays open while the agent is thinking or speaking</text>
 
-  <rect x="150" y="360" width="900" height="54" rx="17" fill="#020617" stroke="#1F2937"/>
-  <text x="182" y="394" fill="#A7F3D0" font-family="SFMono-Regular, ui-monospace, monospace" font-size="18">$ vc language ko &amp;&amp; vc instance start my-project</text>
-  <text x="1045" y="394" fill="#64748B" text-anchor="end" font-family="Inter, ui-sans-serif, system-ui" font-size="15">hands-free coding loop</text>
+  <rect x="150" y="348" width="900" height="54" rx="17" fill="#020617" stroke="#1F2937"/>
+  <text x="182" y="382" fill="#A7F3D0" font-family="SFMono-Regular, ui-monospace, monospace" font-size="18">$ vc language ko &amp;&amp; vc instance start my-project</text>
+  <text x="1045" y="382" fill="#64748B" text-anchor="end" font-family="Inter, ui-sans-serif, system-ui" font-size="15">hands-free coding call</text>
 </svg>

package/docs/i18n/CONFIGURATION.es.md

@@ -1,36 +1,40 @@
-# VerbalCoding Configuración
+# Configuración de VerbalCoding
 
-## Setup Wizard
+## Asistente de configuración
 
-Use upstream Discord-side guides first, then return to VerbalCoding:
+La configuración de la aplicación/bot de Discord no se vuelve a explicar desde cero aquí de forma intencionada. Usa estas guías originales para los pasos del lado de Discord y luego vuelve a la configuración de VerbalCoding:
 
-- Hermes Agent Discord messaging guide: <https://hermes-agent.nousresearch.com/docs/user-guide/messaging/discord>
-- Discord official bot overview: <https://docs.discord.com/developers/bots/overview>
-- Discord official quick start: <https://docs.discord.com/developers/quick-start/getting-started>
+- Guía de mensajería Discord de Hermes Agent: <https://hermes-agent.nousresearch.com/docs/user-guide/messaging/discord>
+- Resumen oficial de bots de Discord: <https://docs.discord.com/developers/bots/overview>
+- Inicio rápido oficial de Discord: <https://docs.discord.com/developers/quick-start/getting-started>
 
 ```bash
-vc setup --yes
-# or from a clone
 ./scripts/install.sh
 ```
 
-The installer asks for the Discord token, allowed users, auto-join voice channel names, transcript channel/thread, CLI harness backend, default voice language, TTS settings, and wake-word behavior. It writes `.env` with mode `0600`.
+El instalador solicita token de Discord, usuarios permitidos, nombres de canales de voz para auto-unión, canal/hilo de transcripción, backend de arnés CLI, idioma de voz predeterminado, ajustes de TTS y comportamiento de palabra de activación. Escribe `.env` con modo `0600`; `.env` está ignorado por git. También enlaza el comando corto de shell `vc`.
 
-## Supported Agent Backends
+Si solo necesitas el comando de shell después de una instalación manual:
 
-Set `AGENT_BACKEND` in `.env`.
+```bash
+npm link
+```
+
+## Backends de agentes compatibles
 
-| Backend | Default command | Notes |
+Define `AGENT_BACKEND` en `.env`.
+
+| Backend | Comando predeterminado | Notas |
 |---|---|---|
-| `hermes` | `hermes chat -Q -q` | Default; supports resume and verbose progress |
-| `claude-code` / `claude` | `claude -p` | Override with `CLAUDE_COMMAND` or `AGENT_COMMAND` |
-| `codex` | `codex exec` | Override with `CODEX_COMMAND` or `AGENT_COMMAND` |
-| `gemini` | `gemini -p` | Override with `GEMINI_COMMAND` or `AGENT_COMMAND` |
-| `opencode` | `opencode run` | Override with `OPENCODE_COMMAND` or `AGENT_COMMAND` |
-| `openclaw` | `openclaw run` | Override with `OPENCLAW_COMMAND` or `AGENT_COMMAND` |
-| `custom` | `AGENT_COMMAND` required | Prompt is appended as final argv |
+| `hermes` | `hermes chat -Q -q` | Predeterminado. Conserva el comportamiento de reanudación de `.verbalcoding-session`. |
+| `claude-code` / `claude` | `claude -p` | Sobrescribe con `CLAUDE_COMMAND` o `AGENT_COMMAND`. |
+| `codex` | `codex exec` | Sobrescribe con `CODEX_COMMAND` o `AGENT_COMMAND`. |
+| `gemini` | `gemini -p` | Sobrescribe con `GEMINI_COMMAND` o `AGENT_COMMAND`. |
+| `opencode` | `opencode run` | Sobrescribe con `OPENCODE_COMMAND` o `AGENT_COMMAND`. |
+| `openclaw` | `openclaw run` | Sobrescribe con `OPENCLAW_COMMAND` o `AGENT_COMMAND`. |
+| `custom` | `AGENT_COMMAND` requerido | El prompt se añade como argumento argv final. |
 
-Generic overrides:
+Sobrescrituras genéricas:
 
 ```bash
 AGENT_BACKEND=custom
@@ -43,23 +47,37 @@ UTTERANCE_IDLE_MS=4500
 LATENCY_LOG_PATH=./.logs/latency.jsonl
 ```
 
-## Example `.env`
+## Contrato del adaptador de agente
+
+El puente de voz habla con cada backend mediante un único contrato de adaptador:
+
+- `run({ text }, signal, plan)` devuelve estado, texto de respuesta final, etiqueta del backend, tiempo transcurrido y metadatos de sesión opcionales.
+- `ask(text, signal, plan)` es el atajo de compatibilidad que devuelve solo el texto de la respuesta final.
+- `capabilities` declara si el backend admite reanudación de sesión, progreso en streaming y cancelación.
+- Hermes es el adaptador de referencia: reanudación, streaming de progreso detallado, cancelación y recuperación de respuesta final desde archivos de sesión de Hermes.
+
+Los nuevos backends deberían implementar el mismo contrato y mantener el comportamiento de voz/STT/TTS fuera del adaptador.
+
+## Ejemplo de `.env`
 
 ```bash
 DISCORD_BOT_TOKEN="***"
 DISCORD_ALLOWED_USERS="123456789012345678"
 AUTO_JOIN_VOICE_CHANNELS="일반,General,general"
 TRANSCRIPT_CHANNEL_ID="123456789012345678"
+
 AGENT_BACKEND="hermes"
 STT_ENGINE="whisper_cpp"
 WHISPER_CPP_BIN="whisper-cli"
 WHISPER_CPP_MODEL="./models/ggml-small-q5_1.bin"
+
 TTS_BACKEND="edge"
 TTS_VOICE_TYPE="korean_female"
 TTS_VOICE="ko-KR-SunHiNeural"
 TTS_RATE="+10%"
 TTS_MAX_CHARS="495"
 TTS_VOLUME="1.0"
+
 REQUIRE_WAKE_WORD="0"
 MIN_UTTERANCE_SECONDS="1.0"
 UTTERANCE_IDLE_MS="4500"
@@ -69,39 +87,60 @@ AGENT_VERBOSE_PROGRESS="0"
 LATENCY_LOG_PATH="./.logs/latency.jsonl"
 ```
 
-## TTS Voice Selection
+## Selección de voz TTS
+
+Los preajustes de idioma y la selección de voz están separados:
 
-`vc language ko|en|auto` changes STT language, progress language, and default TTS voice. Live commands such as “남자 한국어 목소리로 바꿔”, “여자 한국어 목소리로 바꿔”, `change voice to Korean female`, and `switch speaker to English` change only the speaker/voice type.
+- `vc language ko|en|auto` cambia el idioma STT, el idioma de progreso y la voz predeterminada para ese idioma.
+- Comandos de voz en vivo como “남자 한국어 목소리로 바꿔”, “여자 한국어 목소리로 바꿔”, `change voice to Korean female` y `switch speaker to English` cambian solo el hablante/tipo de voz.
+- `!voice-test <text>` reproduce una muestra rápida con el backend y la voz actualmente seleccionados.
 
-Default Edge catalog:
+La selección de voz se guarda por defecto en `config/tts-voices.json`. Sobrescribe la ruta con `TTS_VOICE_CONFIG`. El puente en ejecución vuelve a leer/aplicar la selección de voz antes de sintetizar, por lo que los comandos de voz surten efecto sin reinicio completo.
 
-| `TTS_VOICE_TYPE` | `TTS_VOICE` | Language |
+Catálogo Edge predeterminado:
+
+| `TTS_VOICE_TYPE` | `TTS_VOICE` | Idioma |
 |---|---|---|
-| `korean_male` | `ko-KR-InJoonNeural` | Korean |
-| `korean_female` | `ko-KR-SunHiNeural` | Korean |
-| `korean_multilingual_male` | `ko-KR-HyunsuMultilingualNeural` | Korean |
-| `english_male` | `en-US-GuyNeural` | English |
-| `english_female` | `en-US-AriaNeural` | English |
+| `korean_male` | `ko-KR-InJoonNeural` | Coreano |
+| `korean_female` | `ko-KR-SunHiNeural` | Coreano |
+| `korean_multilingual_male` | `ko-KR-HyunsuMultilingualNeural` | Coreano |
+| `english_male` | `en-US-GuyNeural` | Inglés |
+| `english_female` | `en-US-AriaNeural` | Inglés |
+
+Sobrescritura manual persistente:
+
+```bash
+TTS_BACKEND="edge"
+TTS_VOICE_TYPE="korean_male"
+TTS_VOICE="ko-KR-InJoonNeural"
+TTS_VOICE_CONFIG="config/tts-voices.json"
+```
 
-Backend-specific voice options:
+Para OpenVoice, SpeechSwift o Supertonic, mantén los ajustes de voz/referencia específicos del backend en las secciones siguientes; el mismo archivo de catálogo de voces aún puede rastrear el tipo de voz activo.
 
-| Backend | Settings | Voice choices |
+Opciones de voz específicas de backend:
+
+| Backend | Ajustes | Opciones de voz |
 |---|---|---|
-| Edge | `TTS_VOICE_TYPE`, `TTS_VOICE` | Built-in types plus any `edge-tts --list-voices` voice |
-| Supertonic | `SUPERTONIC_VOICE`, `SUPERTONIC_LANGUAGE` | `M1`–`M5`, `F1`–`F5`; `ko`, `en`, `es`, `pt`, `fr` |
-| OpenVoice | `OPENVOICE_REF_AUDIO`, `OPENVOICE_STYLE`, `OPENVOICE_LANGUAGE` | User-provided permitted reference WAV |
-| SpeechSwift / CosyVoice | `SPEECHSWIFT_REF_AUDIO`, `SPEECHSWIFT_ENGINE`, `SPEECHSWIFT_SPEAKER`, `SPEECHSWIFT_MODEL_ID` | Reference-sample voice or backend speaker/model ID |
+| Edge | `TTS_VOICE_TYPE`, `TTS_VOICE` | Tipos integrados anteriores, más cualquier voz devuelta por `edge-tts --list-voices` |
+| Supertonic | `SUPERTONIC_VOICE`, `SUPERTONIC_LANGUAGE` | `M1`–`M5`, `F1`–`F5`; idioma `ko`, `en`, `es`, `pt`, `fr` |
+| OpenVoice | `OPENVOICE_REF_AUDIO`, `OPENVOICE_STYLE`, `OPENVOICE_LANGUAGE` | WAV de referencia permitido proporcionado por el usuario; el estilo predeterminado es `default` |
+| SpeechSwift / CosyVoice | `SPEECHSWIFT_REF_AUDIO`, `SPEECHSWIFT_ENGINE`, `SPEECHSWIFT_SPEAKER`, `SPEECHSWIFT_MODEL_ID` | Voces de muestra de referencia para CosyVoice, o IDs de hablante/modelo admitidos por el backend |
 
-## Utterance Segmentation
+## Segmentación de emisiones
 
-`UTTERANCE_IDLE_MS` controls how long the bridge waits after speech before starting STT. Default is `4500` ms.
+`UTTERANCE_IDLE_MS` controla cuánto espera el puente después de un segmento de habla antes de decidir que el usuario terminó y empezar STT. El valor predeterminado es `4500` ms para conservar instrucciones habladas más largas con pausas naturales. Los valores menores se sienten más rápidos para comandos cortos, pero pueden dividir dictado largo; los valores mayores son más seguros para habla reflexiva.
 
 ```bash
-UTTERANCE_IDLE_MS="4500"
-UTTERANCE_IDLE_MS="6000"
+UTTERANCE_IDLE_MS="4500"  # balanced default
+UTTERANCE_IDLE_MS="6000"  # safer for long dictation with pauses
 ```
 
-## MCP Server
+## Servidor MCP
+
+VerbalCoding incluye un servidor MCP stdio para que Hermes Agent o cualquier cliente MCP pueda controlar el puente mediante herramientas en lugar de depender de skills o comandos de shell de forma libre.
+
+Ejemplo de configuración de Hermes:
 
 ```yaml
 mcp_servers:
@@ -112,39 +151,89 @@ mcp_servers:
     connect_timeout: 30
 ```
 
-Tools: `status`, `doctor`, `set_auto_restart`, `set_language`, `start`, `stop`, and `restart`.
+Herramientas MCP expuestas:
+
+| Herramienta | Propósito |
+|---|---|
+| `status` | Informar estado del puente/configuración sin secretos |
+| `doctor` | Ejecutar la comprobación doctor con secretos redactados |
+| `set_auto_restart` | Habilitar/deshabilitar el reinicio automático del bot de voz al hacer commit |
+| `set_language` | Actualizar juntos STT/progreso/TTS |
+| `start`, `stop`, `restart` | Controlar el puente de voz de Discord |
 
-## Optional OpenVoice TTS
+## TTS OpenVoice opcional
+
+Edge TTS sigue siendo el valor predeterminado y la alternativa. Para probar clonación de voz local con OpenVoice V2:
 
 ```bash
 ./scripts/setup_openvoice.sh
+# Download checkpoints_v2_0417.zip from OpenVoice docs and extract under vendor/OpenVoice/checkpoints_v2/
+mkdir -p voice-samples
+# Put a permitted reference sample at voice-samples/user-reference.wav,
+# or capture one from Discord with !voice-clone capture.
 python3 integrations/openvoice/synth.py --openvoice-dir vendor/OpenVoice --ref-audio voice-samples/user-reference.wav --text '안녕하세요. 버벌코딩 목소리 복제 테스트입니다.' --output /tmp/verbalcoding-openvoice-smoke.wav
 ```
 
+Luego define:
+
 ```bash
 TTS_BACKEND="openvoice"
 OPENVOICE_REF_AUDIO="./voice-samples/user-reference.wav"
 OPENVOICE_PROGRESS="0"
 ```
 
-Only clone voices you own or have permission to use. OpenVoice falls back to Edge on failure.
+Clona solo voces que poseas o tengas permiso para usar. Si OpenVoice falla o agota el tiempo, VerbalCoding vuelve a Edge TTS.
 
-## Optional Supertonic TTS
+## TTS Supertonic opcional
 
 ```bash
 ./scripts/setup_supertonic.sh
 supertonic tts '안녕하세요. 수퍼토닉 테스트입니다.' --lang ko --voice M1 --steps 2 --speed 1.0 -o /tmp/verbalcoding-supertonic.wav
 ```
 
-## Optional SpeechSwift / CosyVoice TTS
+Luego define:
+
+```bash
+TTS_BACKEND="supertonic"
+SUPERTONIC_COMMAND="./.venv-supertonic/bin/supertonic"
+SUPERTONIC_VOICE="M1"
+SUPERTONIC_LANGUAGE="ko"
+SUPERTONIC_STEPS="2"
+SUPERTONIC_SPEED="1.0"
+SUPERTONIC_PROGRESS="0"
+```
+
+Si Supertonic falta, falla o agota el tiempo, VerbalCoding vuelve a Edge TTS.
+
+## TTS SpeechSwift / CosyVoice opcional
+
+En Apple Silicon, `speech-swift` es un backend local para clonación de voz coreana con CosyVoice/Qwen3-TTS nativo de MLX.
 
 ```bash
 brew tap soniqo/speech https://github.com/soniqo/speech-swift
 brew install speech
 ```
 
-Recommended env includes `TTS_BACKEND="speechswift"`, `SPEECHSWIFT_MODE="server"`, `SPEECHSWIFT_ENGINE="cosyvoice"`, `SPEECHSWIFT_REF_AUDIO`, and `SPEECHSWIFT_SERVER_URL`. Keep Edge for quick progress prompts.
+Entorno recomendado:
+
+```bash
+TTS_BACKEND="speechswift"
+SPEECHSWIFT_MODE="server"
+SPEECHSWIFT_ENGINE="cosyvoice"
+SPEECHSWIFT_LANGUAGE="korean"
+SPEECHSWIFT_REF_AUDIO="./voice-samples/user-reference.wav"
+SPEECHSWIFT_SERVER_HOST="127.0.0.1"
+SPEECHSWIFT_SERVER_PORT="18080"
+SPEECHSWIFT_SERVER_URL="http://127.0.0.1:18080"
+SPEECHSWIFT_PROGRESS="0"
+```
+
+Mantén Edge para prompts rápidos de progreso/backchannel.
 
-## Operational Notes
+## Notas operativas
 
-Enable Discord Message Content intent, grant voice connect/speak permissions, authenticate the selected CLI harness separately, and avoid reading diffs/log dumps aloud.
+- El bot necesita el intent privilegiado Message Content de Discord habilitado para comandos de texto.
+- El bot necesita permisos de conectar/hablar en el canal de voz.
+- Para Hermes Agent, configura/autentica Hermes normalmente (`hermes setup`, `hermes login`, etc.) en tu perfil predeterminado.
+- Para Claude Code, Codex, Gemini, OpenCode y OpenClaw, instala y autentica esas CLIs por separado.
+- Si una CLI emite salida de diff/código durante un timeout o fallo de señal, el puente evita leerla en voz alta y envía texto detallado en su lugar.