feat(dross): voice Phase 2a — local whisper transcribe + mic (2.12.0)

faster-whisper (small.en, GPU+CPU fallback) on CT 102 → POST
/api/voice/transcribe (multer→whisper client) → mic in the bubble
records (MediaRecorder), uploads, drops the transcript into the input
to review-and-send. Infra scripts in deploy/whisper/. Retention (P2b)
next. NOTE: mic needs a secure context (the https domain), not the LAN IP.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

deploy/whisper/README.md

@@ -0,0 +1,16 @@
+# faster-whisper service (Dross voice STT)
+
+Runs on **CT 102** (the Ollama box, `192.168.1.185`), bare-metal (no Docker), on the
+RTX A2000 with CPU fallback. OpenAI-style `/transcribe` consumed by void-app
+`lib/voice/whisper.js` (`WHISPER_URL=http://192.168.1.185:8001`).
+
+## Install (on CT 102)
+```
+scp deploy/whisper/{server.py,setup.sh} root@192.168.1.185:/opt/whisper_server.py /root/setup.sh
+ssh root@192.168.1.185 'bash /root/setup.sh && install -m644 /opt/whisper_server.py /opt/whisper/server.py && systemctl enable --now whisper'
+curl http://192.168.1.185:8001/health   # {"ok":true,"model":"small.en","device":"cuda"}
+```
+- venv at `/opt/whisper/venv`; model `small.en` (env `WHISPER_MODEL`); CUDA libs via
+  `nvidia-cublas-cu12`/`nvidia-cudnn-cu12` pip wheels (LD_LIBRARY_PATH in the unit).
+- GPU → CPU fallback is in `server.py` `load()`.
+- **CT 102 disk was expanded +20G** (was 89% full) before install.

deploy/whisper/server.py

@@ -0,0 +1,35 @@
+import os, tempfile
+from fastapi import FastAPI, UploadFile, File, HTTPException
+from faster_whisper import WhisperModel
+
+MODEL = os.environ.get("WHISPER_MODEL", "small.en")
+app = FastAPI()
+model = None
+device_used = None
+
+def load():
+    global model, device_used
+    try:
+        model = WhisperModel(MODEL, device="cuda", compute_type="int8_float16")
+        device_used = "cuda"
+    except Exception:
+        model = WhisperModel(MODEL, device="cpu", compute_type="int8")
+        device_used = "cpu"
+
+load()
+
+@app.get("/health")
+def health():
+    return {"ok": True, "model": MODEL, "device": device_used}
+
+@app.post("/transcribe")
+async def transcribe(file: UploadFile = File(...)):
+    data = await file.read()
+    if not data:
+        raise HTTPException(400, "empty audio")
+    with tempfile.NamedTemporaryFile(suffix=".bin") as f:
+        f.write(data); f.flush()
+        segments, info = model.transcribe(f.name, beam_size=1, vad_filter=True)
+        text = "".join(s.text for s in segments).strip()
+    return {"text": text, "language": info.language,
+            "duration": round(info.duration, 2), "device": device_used}

deploy/whisper/setup.sh

@@ -0,0 +1,26 @@
+#!/usr/bin/env bash
+set -e
+export DEBIAN_FRONTEND=noninteractive
+apt-get update -qq
+apt-get install -y -qq python3-pip python3-venv ffmpeg >/dev/null
+mkdir -p /opt/whisper
+python3 -m venv /opt/whisper/venv
+/opt/whisper/venv/bin/pip install -q --upgrade pip
+/opt/whisper/venv/bin/pip install -q faster-whisper fastapi "uvicorn[standard]" python-multipart nvidia-cublas-cu12 nvidia-cudnn-cu12
+SITE=/opt/whisper/venv/lib/python3.12/site-packages
+cat > /etc/systemd/system/whisper.service <<UNIT
+[Unit]
+Description=faster-whisper transcription server (Dross voice)
+After=network.target
+[Service]
+Type=simple
+WorkingDirectory=/opt/whisper
+Environment=WHISPER_MODEL=small.en
+Environment=LD_LIBRARY_PATH=${SITE}/nvidia/cublas/lib:${SITE}/nvidia/cudnn/lib
+ExecStart=/opt/whisper/venv/bin/uvicorn server:app --host 0.0.0.0 --port 8001
+Restart=on-failure
+[Install]
+WantedBy=multi-user.target
+UNIT
+systemctl daemon-reload
+echo "deps+unit installed"

lib/api/index.js

@@ -39,6 +39,7 @@ import { router as backupsRouter } from './routes/backups.js';
 import { router as kuttRouter } from './routes/kutt.js';
 import { router as themeRouter } from './routes/theme.js';
 import { router as drossRouter } from './routes/dross.js';
+import { router as voiceRouter } from './routes/voice.js';
 
 export function mountApi(app) {
   const api = Router();
@@ -75,6 +76,7 @@ export function mountApi(app) {
   api.use('/kutt', kuttRouter);
   api.use('/theme', themeRouter);
   api.use('/dross', drossRouter);
+  api.use('/voice', voiceRouter);
   api.use('/pending-changes', pendingChangesRouter);
   api.use('/audit', auditRouter);
   api.use('/search', searchRouter);

lib/api/routes/voice.js

@@ -0,0 +1,24 @@
+import { Router } from 'express';
+import multer from 'multer';
+import { asyncWrap } from '../errors.js';
+import { requireOwner } from '../cap.js';
+import * as whisper from '../../voice/whisper.js';
+export const router = Router();
+
+// In-memory upload; clips are small voice notes. 25 MB ceiling.
+const upload = multer({ storage: multer.memoryStorage(), limits: { fileSize: 25 * 1024 * 1024 } });
+
+// POST /api/voice/transcribe — owner-only. multipart field `audio`. Returns { text }.
+// (Phase 2b will optionally persist the clip + transcript when keepClips is on.)
+router.post('/transcribe', requireOwner, upload.single('audio'), asyncWrap(async (req, res) => {
+  if (!req.file || !req.file.buffer?.length) {
+    return res.status(400).json({ error: { code: 'no_audio', message: 'no audio supplied' } });
+  }
+  try {
+    const r = await whisper.transcribe(
+      req.file.buffer, req.file.originalname || 'clip.webm', req.file.mimetype || 'audio/webm');
+    res.json({ text: r.text, duration: r.duration ?? null });
+  } catch {
+    res.status(503).json({ error: { code: 'stt_unavailable', message: 'transcription service unavailable' } });
+  }
+}));

lib/voice/whisper.js

@@ -0,0 +1,22 @@
+// Thin client for the local faster-whisper service on CT 102 (the Ollama box).
+// GPU with CPU fallback lives in the service itself; here we just POST the audio
+// buffer and return the transcript. LAN-only endpoint.
+const WHISPER_URL = process.env.WHISPER_URL || 'http://192.168.1.185:8001';
+
+export async function transcribe(buffer, filename = 'clip.webm', mime = 'audio/webm') {
+  const fd = new FormData();
+  fd.append('file', new Blob([buffer], { type: mime }), filename);
+  const res = await fetch(`${WHISPER_URL}/transcribe`, {
+    method: 'POST', body: fd, signal: AbortSignal.timeout(120000)
+  });
+  if (!res.ok) throw new Error(`whisper ${res.status}`);
+  const j = await res.json();
+  return { text: (j.text || '').trim(), duration: j.duration, device: j.device };
+}
+
+export async function health() {
+  try {
+    const res = await fetch(`${WHISPER_URL}/health`, { signal: AbortSignal.timeout(5000) });
+    return res.ok ? await res.json() : null;
+  } catch { return null; }
+}

package.json

@@ -1,6 +1,6 @@
 {
   "name": "void-server",
-  "version": "2.11.0",
+  "version": "2.12.0",
   "type": "module",
   "private": true,
   "scripts": {

public/components/dross_bubble.js

@@ -22,8 +22,9 @@ export async function renderDrossBubble() {
   const input = el('textarea', { rows: 1, placeholder: 'Ask Dross…' });
   const sendBtn = el('button', { class: 'dross-send', title: 'Send' },
     el('span', { html: '<svg viewBox="0 0 24 24" width="20" height="20" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><path d="M22 2 11 13M22 2l-7 20-4-9-9-4 20-7z"/></svg>' }));
-  const mic = el('button', { class: 'dross-mic', disabled: true, title: 'Voice arrives in Phase 2' },
+  const micLabel = el('span', {}, 'Tap to record');
-    el('span', { html: '<svg viewBox="0 0 24 24" width="22" height="22" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><rect x="9" y="3" width="6" height="11" rx="3"/><path d="M5 11a7 7 0 0 0 14 0"/><line x1="12" y1="18" x2="12" y2="22"/><line x1="8" y1="22" x2="16" y2="22"/></svg>' }), 'Hold to talk');
+  const mic = el('button', { class: 'dross-mic', title: 'Record a voice note' },
+    el('span', { html: '<svg viewBox="0 0 24 24" width="22" height="22" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><rect x="9" y="3" width="6" height="11" rx="3"/><path d="M5 11a7 7 0 0 0 14 0"/><line x1="12" y1="18" x2="12" y2="22"/><line x1="8" y1="22" x2="16" y2="22"/></svg>' }), micLabel);
   const closeBtn = el('button', { class: 'dross-x', title: 'Close' }, '⤬');
   const header = el('div', { class: 'dross-hd' }, drossAvatar(cfg.avatar, 30),
     el('div', { class: 'dross-who' }, 'Dross', el('small', {}, 'always here, regrettably')), closeBtn);
@@ -62,6 +63,47 @@ export async function renderDrossBubble() {
 
   drag(fab, fab, true); drag(header, panel, false);
 
+  // ---- voice: tap mic to record, tap again to stop → transcribe → review-and-send ----
+  let media = null, chunks = [], recording = false;
+  function setMic(label, rec) { micLabel.textContent = label; mic.classList.toggle('rec', !!rec); }
+  async function startRec() {
+    try {
+      const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
+      chunks = [];
+      const opt = (window.MediaRecorder && MediaRecorder.isTypeSupported('audio/webm;codecs=opus'))
+        ? { mimeType: 'audio/webm;codecs=opus' } : {};
+      media = new MediaRecorder(stream, opt);
+      media.ondataavailable = (e) => { if (e.data && e.data.size) chunks.push(e.data); };
+      media.onstop = async () => {
+        stream.getTracks().forEach(t => t.stop());
+        await sendClip(new Blob(chunks, { type: media.mimeType || 'audio/webm' }));
+      };
+      media.start();
+      recording = true; setMic('● Recording… tap to stop', true);
+    } catch {
+      setMic('Mic blocked', false); setTimeout(() => setMic('Tap to record', false), 1800);
+    }
+  }
+  function stopRec() {
+    if (media && recording) { recording = false; setMic('Transcribing…', false); media.stop(); }
+  }
+  async function sendClip(blob) {
+    try {
+      const fd = new FormData(); fd.append('audio', blob, 'clip.webm');
+      const res = await fetch('/api/voice/transcribe', {
+        method: 'POST', headers: { Authorization: 'Bearer ' + (localStorage.getItem('void_token') || '') }, body: fd
+      });
+      if (!res.ok) throw new Error('stt');
+      const { text } = await res.json();
+      setMic('Tap to record', false);
+      if (text) { input.value = input.value ? (input.value + ' ' + text) : text; input.focus(); }
+      // voiceMode 'handsfree'/'action' (Phase 2b+) would branch here.
+    } catch {
+      setMic('Transcribe failed', false); setTimeout(() => setMic('Tap to record', false), 2000);
+    }
+  }
+  mic.addEventListener('click', () => recording ? stopRec() : startRec());
+
   window.addEventListener('dross-settings-changed', async () => {
     try { cfg = { ...cfg, ...(await api.get('/api/dross/settings')) }; } catch { return; }
     applyAccent(fab, cfg.accent); applyAccent(panel, cfg.accent);

public/style.css

@@ -765,6 +765,8 @@ body.drawer-open #scrim { opacity: 1; pointer-events: auto; }
 .dross-btnrow{display:flex;gap:10px}
 .dross-mic{flex:1;height:50px;border-radius:12px;border:1px solid var(--dross-dim);background:var(--dross-soft);color:var(--dross-glow);cursor:pointer;display:flex;align-items:center;justify-content:center;gap:9px;font-family:var(--font-ui);font-size:13px}
 .dross-mic[disabled]{opacity:.5;cursor:not-allowed}
+.dross-mic.rec{background:#3a1010;border-color:var(--accent);color:#fff;animation:dross-rec 1.2s infinite}
+@keyframes dross-rec{0%,100%{box-shadow:0 0 0 0 rgba(255,79,46,.5)}50%{box-shadow:0 0 0 8px rgba(255,79,46,0)}}
 .dross-send{width:64px;height:50px;border-radius:12px;border:1px solid var(--dross-dim);background:linear-gradient(180deg,var(--dross),var(--dross-dim));color:#fff;cursor:pointer;display:grid;place-items:center}
 .dross-collapse{display:flex;align-items:center;justify-content:center;gap:8px;height:34px;cursor:pointer;color:var(--muted);
   font-family:var(--font-ui);font-size:11px;letter-spacing:.12em;text-transform:uppercase;background:#0b0810;border-top:1px solid var(--border)}

tests/routes/voice.test.js

@@ -0,0 +1,24 @@
+import { describe, it, expect, beforeAll } from 'vitest';
+import request from 'supertest';
+import { createApp } from '../../server.js';
+import { resetDb } from '../helpers/db.js';
+import { migrateUp } from '../../lib/db/migrate.js';
+
+let app;
+const owner = { Authorization: 'Bearer test-token' };
+beforeAll(async () => {
+  await resetDb(); await migrateUp();
+  process.env.OWNER_TOKEN = 'test-token';
+  app = createApp();
+});
+
+describe('voice transcribe route', () => {
+  it('401 without a token', async () => {
+    const res = await request(app).post('/api/voice/transcribe');
+    expect(res.status).toBe(401);
+  });
+  it('400 when no audio supplied', async () => {
+    const res = await request(app).post('/api/voice/transcribe').set(owner);
+    expect(res.status).toBe(400);
+  });
+});