@memef1f1y/dsh-plugin-tts 0.6.0

package/tools/gen-ui-preview.mjs

@@ -0,0 +1,174 @@
+#!/usr/bin/env node
+// Generates a standalone UI preview page (tools/ui-preview.html) that loads the
+// REAL lib/client.js in a browser, renders the injected components into actual
+// DOM using a lightweight React-agnostic renderer, and lets you screenshot it
+// for visual verification of the plugin UI (P2-P5).
+//
+//   node tools/gen-ui-preview.mjs [out.html]
+// Then open out.html in a browser (or headless screenshot):
+//   npx playwright screenshot tools/ui-preview.html shot.png  (example)
+//
+// Note: this is a visual-verification aid only, NOT shipped into the plugin.
+import { readFileSync, writeFileSync } from 'node:fs';
+import { fileURLToPath } from 'node:url';
+import path from 'node:path';
+
+const here = path.dirname(fileURLToPath(import.meta.url));
+const clientSrc = readFileSync(path.resolve(here, '..', 'lib', 'client.js'), 'utf8')
+  .replace(/<\/script/gi, '<\\/script');
+
+const out = path.resolve(here, process.argv[2] || 'ui-preview.html');
+
+const html = `<!doctype html>
+<html lang="zh-CN"><head><meta charset="utf-8"><title>dsh-plugin-tts UI preview</title>
+<style>
+  :root{
+    /* minimal dsh web theme tokens so --dsw-* styles resolve */
+    --dsw-alias-label-primary:#eceff3;
+    --dsw-alias-label-secondary:#c3c9d1;
+    --dsw-alias-label-tertiary:#7f8894;
+    --dsw-alias-label-dimmed:#5a6472;
+    --dsw-alias-label-inverse:#0f1115;
+    --dsw-alias-label-error:#e5484d;
+    --dsw-alias-interactive-bg-hover:rgba(255,255,255,.06);
+    --dsw-alias-interactive-bg-primary:#3b82f6;
+    --dsw-alias-bg-layer-2:#161a20;
+    --dsw-alias-bg-layer-3:#1c2128;
+    --dsw-alias-border-secondary:rgba(255,255,255,.08);
+    --dsw-alias-border-l2:#2a313b;
+    --dsw-alias-brand-primary:#3b82f6;
+  }
+  body{background:#0d0f13;color:var(--dsw-alias-label-primary);font-family:system-ui,-apple-system,'PingFang SC','Microsoft YaHei',sans-serif;margin:0;padding:24px}
+  h2{font-size:13px;color:var(--dsw-alias-label-tertiary);font-weight:600;margin:18px 0 8px;text-transform:uppercase;letter-spacing:.5px}
+  .card{background:#14181d;border:1px solid var(--dsw-alias-border-secondary);border-radius:12px;padding:14px;margin-bottom:6px;max-width:640px}
+  .toolrow{display:flex;align-items:center;gap:6px;padding:6px 8px;background:#15191f;border:1px solid var(--dsw-alias-border-secondary);border-radius:10px;margin-bottom:8px}
+  .toolrow::before{content:"▍";color:var(--dsw-alias-label-dimmed);font-size:13px}
+  pre.tip{color:var(--dsw-alias-label-tertiary);font-size:11px;margin:2px 0 0}
+</style></head>
+<body>
+  <h2>1) auto-read pill (off)</h2>
+  <div class="toolrow"><div class="slot-input" data-state="off"></div></div>
+  <h2>2) auto-read pill (on)</h2>
+  <div class="toolrow"><div class="slot-input" data-state="on"></div></div>
+  <h2>3) voice settings panel — Edge provider</h2>
+  <div class="card"><div class="slot-settings" data-provider="edge"></div></div>
+  <h2>4) voice settings panel — RVC provider (onboarding)</h2>
+  <div class="card"><div class="slot-settings" data-provider="rvc"></div></div>
+  <h2>5) selection-read chip (demo)</h2>
+  <div class="card"><button type="button" class="dsh-tts-sel-btn">朗读选中文本</button> <span style="color:var(--dsw-alias-label-tertiary);font-size:11px">← 悬浮在选择旁的“朗读选中”chip（真实交互中由 mouseup 触发）</span></div>
+
+<script>
+/* ---------- minimal React-agnostic shim + hooks + DOM renderer ---------- */
+let activeHooks = null;
+const react = {
+  createElement(type, props, ...children){
+    const flat=[];
+    const push=(x)=>{ if(x==null||x===false||x===true)return; if(Array.isArray(x)){x.forEach(push);return;} flat.push(x); };
+    children.forEach(push);
+    return {type, props: props||{}, children: flat};
+  },
+  Fragment: Symbol('frag'),
+  useState:i=>activeHooks.useState(i),
+  useEffect:(f,d)=>activeHooks.effect(f,d),
+  useRef:i=>activeHooks.ref(i),
+  useMemo:f=>activeHooks.memo(f),
+};
+function makeHooks(){
+  let c=0; const mem=[];
+  return {
+    useState(init){ const i=c++; if(!(i in mem)) mem[i]=typeof init==='function'?init():init; return [mem[i],()=>{}]; },
+    effect(){ c++; /* do not run component effects (no interaction/nav) */ },
+    ref(init){ c++; return {current:init}; },
+    memo(f){ c++; try{return f();}catch(e){return undefined;} },
+  };
+}
+function mount(v){
+  if(v==null||typeof v==='boolean') return null;
+  if(typeof v==='string'||typeof v==='number') return document.createTextNode(String(v));
+  if(v.type===react.Fragment){ const f=document.createDocumentFragment(); (v.children||[]).forEach(c=>{const n=mount(c);if(n)f.appendChild(n);}); return f; }
+  if(typeof v.type==='function'){ activeHooks=makeHooks(); const n=mount(v.type(v.props) ); activeHooks=null; return n; }
+  const el=document.createElement(v.type);
+  const p=v.props||{};
+  for(const k in p){
+    const val=p[k];
+    if(k==='children'||k==='key'||k==='ref') continue;
+    if(typeof val==='function') continue; // handlers ignored for static render
+    if(k==='className'){ el.setAttribute('class', val); }
+    else if(k==='style'&&val&&typeof val==='object'){ for(const sk in val){ try{ el.style[sk]=val[sk]; }catch(e){} } }
+    else if(k==='value'){ try{ el.value=val; }catch(e){ el.setAttribute('value',val); } }
+    else if(k==='disabled'){ if(val) el.setAttribute('disabled',''); }
+    else { try{ el.setAttribute(k, val); }catch(e){} }
+  }
+  (v.children||[]).forEach(c=>{ const n=mount(c); if(n)el.appendChild(n); });
+  return el;
+}
+/* ---------- slots / ctx (mirrors client-load) ---------- */
+const injectedComponents=[];
+const slots={
+  inject(slot,fn){ injectedComponents.push({slot,fn}); },
+  register(spec,component){ return component; },
+};
+const ctx={ get(n){ return n==='slots'?slots:undefined; }, effect(fn){ try{ fn(); }catch(e){} } };
+
+/* ---------- load client bundle once; we'll re-eval per state ---------- */
+const CLIENT = ${JSON.stringify(clientSrc)};
+function evalClient(){
+  injectedComponents.length=0;
+  window.__ModuleLoader__={
+    load({factory}){
+      const require_=name=>{ if(name==='react') return react; throw new Error('unknown require '+name); };
+      const mod=factory(require_);
+      if(!mod||!mod.apply) throw new Error('no apply');
+      mod.apply(ctx);
+    },
+  };
+  const fn=new Function('window','navigator','document','Audio', CLIENT + '\\n;return window.__ModuleLoader__;');
+  fn(window,navigator,document,Audio);
+}
+function seedSettings(obj){
+  if(obj) localStorage.setItem('dsh-tts-settings', JSON.stringify(obj));
+  else localStorage.removeItem('dsh-tts-settings');
+}
+function renderSlot(slotName, container){
+  try{
+    const c=injectedComponents.find(x=>x.slot===slotName);
+    if(!c) throw new Error('no component for '+slotName);
+    const comp=c.fn();
+    if(typeof comp!=='function') throw new Error(slotName+' fn did not return component');
+    activeHooks=makeHooks();
+    const node=comp({ session:null, sessionId:'s1', useSession:s=>s({nodes:[]}), messageId:'m1' });
+    const dom=mount(node);
+    activeHooks=null;
+    if(dom) container.appendChild(dom);
+  }catch(e){
+    logErr(slotName+': '+ (e&&e.stack||e));
+  }
+}
+function logErr(msg){
+  const el=document.createElement('pre');
+  el.style.cssText='color:#e5484d;font-size:11px;white-space:pre-wrap';
+  el.textContent=msg;
+  document.body.appendChild(el);
+}
+
+/* off */
+seedSettings({autoRead:false, provider:'edge-tts', voice:'zh-CN-XiaoxuanNeural'});
+evalClient(); renderSlot('conversation.input.left', document.querySelector('.slot-input[data-state=off]'));
+
+/* on */
+seedSettings({autoRead:true, provider:'edge-tts', voice:'zh-CN-YunyangNeural'});
+evalClient(); renderSlot('conversation.input.left', document.querySelector('.slot-input[data-state=on]'));
+
+/* settings edge */
+seedSettings({autoRead:false, provider:'edge-tts', voice:'zh-CN-XiaoxuanNeural'});
+evalClient(); renderSlot('settings.plugins.tab', document.querySelector('.slot-settings[data-provider=edge]'));
+
+/* settings rvc (shows onboarding) */
+seedSettings({autoRead:false, provider:'rvc', voice:'zh-CN-XiaoxuanNeural', rvc:{model:'C:/rvc/model.pth',index:'C:/rvc/model.index',f0UpKey:-2,indexRate:0.75,protect:0.33}});
+evalClient(); renderSlot('settings.plugins.tab', document.querySelector('.slot-settings[data-provider=rvc]'));
+</script>
+</body></html>
+`;
+
+writeFileSync(out, html, 'utf8');
+console.log('wrote', out);

package/tools/mac-verify.sh

@@ -0,0 +1,158 @@
+#!/bin/bash
+# mac-verify.sh — end-to-end verification of the dsh-plugin-tts RVC server on macOS.
+#
+# Checks, in order:
+#   1. GET  /health                      server up + device/CPU info
+#   2. GET  /files?kind=pth              model discovery
+#   3. POST /load                        load the voice model (+ optional index)
+#   4. POST /convert                     real conversion of synthesized speech
+#   5. output file sanity (RIFF/duration)
+#   6. POST /compact-index               (only when an index is given)
+#
+# Usage: bash mac-verify.sh <model.pth> [index.index] [host:port]
+#   PY env var overrides the python used for JSON/base64 (default python3 —
+#   use the rvc-work venv python:  ~/rvc-work/venv/bin/python)
+#
+# Exit code 0 = all checks passed; otherwise the failing step's exit code.
+set -u
+
+MODEL="${1:-}"
+INDEX="${2:-}"
+HOST="${3:-127.0.0.1:4892}"
+BASE="http://$HOST"
+PY="${PY:-python3}"
+FAIL=0
+
+if [ -z "$MODEL" ]; then
+  echo "usage: bash mac-verify.sh <model.pth> [index.index] [host:port]"
+  exit 2
+fi
+if ! command -v "$PY" >/dev/null 2>&1; then
+  echo "ERROR: python not found: $PY (set PY=.../venv/bin/python)"
+  exit 2
+fi
+echo "== target: $BASE  model: $MODEL  index: ${INDEX:-<none>}  py: $PY =="
+
+step() { echo; echo "== [$1/6] $2 =="; }
+
+step 1 "/health"
+curl -sS --max-time 15 "$BASE/health" | "$PY" -c '
+import json, sys
+d = json.load(sys.stdin)
+print(json.dumps(d, ensure_ascii=False, indent=1))
+assert d.get("ok") is True, "health ok != true"
+assert d.get("device") == "cpu", "expected cpu device on macOS, got %r" % d.get("device")
+assert d.get("cuda_available") is False, "cuda must not be available on macOS"
+' || FAIL=1
+
+step 2 "/files?kind=pth"
+curl -sS --max-time 15 "$BASE/files?kind=pth" | "$PY" -c '
+import json, sys
+d = json.load(sys.stdin)
+files = d.get("files", [])
+print("discovered %d .pth files" % len(files))
+for f in files[:5]:
+    print("  ", f["name"], f["size"])
+' || FAIL=1
+
+step 3 "/load"
+"$PY" - "$BASE" "$MODEL" "$INDEX" <<'PYEOF' || FAIL=1
+import json, sys, urllib.request
+base, model, index = sys.argv[1], sys.argv[2], sys.argv[3]
+payload = {"model": model}
+if index:
+    payload["index"] = index
+req = urllib.request.Request(base + "/load",
+                             data=json.dumps(payload).encode(),
+                             headers={"Content-Type": "application/json"})
+try:
+    with urllib.request.urlopen(req, timeout=300) as r:
+        print(json.dumps(json.load(r), ensure_ascii=False))
+except urllib.error.HTTPError as e:
+    print("load HTTP %s: %s" % (e.code, e.read().decode("utf-8", "ignore")[:600]))
+    sys.exit(1)
+PYEOF
+
+step 4 "/convert (real speech -> converted wav)"
+TMPD="$(mktemp -d)"
+trap 'rm -rf "$TMPD"' EXIT
+if command -v say >/dev/null 2>&1; then
+  say -o "$TMPD/speech.aiff" "你好，这是一段跨平台语音转换测试。" 2>/dev/null \
+    && ffmpeg -y -loglevel error -i "$TMPD/speech.aiff" -ac 1 -ar 44100 "$TMPD/speech.wav"
+fi
+if [ ! -s "$TMPD/speech.wav" ]; then
+  echo "note: say/ffmpeg failed, synthesizing a 3s tone instead"
+  ffmpeg -y -loglevel error -f lavfi -i "sine=frequency=440:duration=3" -ac 1 -ar 44100 "$TMPD/speech.wav"
+fi
+ls -l "$TMPD/speech.wav"
+"$PY" - "$BASE" "$TMPD/speech.wav" "$TMPD/out.wav" <<'PYEOF' || FAIL=1
+import base64, json, sys, time, urllib.request
+base, wav, out = sys.argv[1], sys.argv[2], sys.argv[3]
+data = open(wav, "rb").read()
+payload = {
+    "audio_base64": base64.b64encode(data).decode("ascii"),
+    "params": {"f0_method": "rmvpe", "index_rate": 0.75},
+}
+t0 = time.time()
+req = urllib.request.Request(base + "/convert",
+                             data=json.dumps(payload).encode(),
+                             headers={"Content-Type": "application/json"})
+try:
+    with urllib.request.urlopen(req, timeout=900) as r:
+        resp = json.load(r)
+except urllib.error.HTTPError as e:
+    print("convert HTTP %s: %s" % (e.code, e.read().decode("utf-8", "ignore")[:800]))
+    sys.exit(1)
+wall = time.time() - t0
+out_b64 = resp.get("audio_base64")
+if not out_b64:
+    print("CONVERT FAILED (no audio in response):", json.dumps(resp, ensure_ascii=False)[:600])
+    sys.exit(1)
+open(out, "wb").write(base64.b64decode(out_b64))
+sr = int(resp.get("sample_rate", 0))
+in_sec = len(data) / 2.0 / 44100.0
+print("convert ok: wall=%.1fs  sample_rate=%s  out=%s  (%.2fx realtime)" % (wall, sr, out, wall / max(in_sec, 0.01)))
+PYEOF
+
+step 5 "output sanity (RIFF + duration)"
+# RIFF magic check. NB: macOS `od -An -tx1` pads bytes with *two* spaces
+# ("52  49  46  46") while GNU `od` uses one ("52 49 46 46"), so a literal
+# "52 49 46 46" grep wrongly fails on macOS even for a valid RIFF file.
+# Strip whitespace and compare the raw hex instead (portable across both).
+if command -v xxd >/dev/null 2>&1; then
+  MAGIC="$(head -c 4 "$TMPD/out.wav" | xxd -p | tr -d '[:space:]')"
+else
+  MAGIC="$(head -c 4 "$TMPD/out.wav" | od -An -tx1 | tr -d '[:space:]')"
+fi
+if [ "$MAGIC" != "52494646" ]; then echo "FAIL: not a RIFF file"; FAIL=1; fi
+if command -v ffprobe >/dev/null 2>&1; then
+  DUR="$(ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 "$TMPD/out.wav" 2>/dev/null)"
+  echo "out.wav duration: ${DUR:-unknown}s (input speech was ~3s)"
+fi
+
+if [ -n "$INDEX" ]; then
+  step 6 "/compact-index"
+  "$PY" - "$BASE" "$INDEX" <<'PYEOF' || FAIL=1
+import json, sys, urllib.request
+base, idx = sys.argv[1], sys.argv[2]
+req = urllib.request.Request(base + "/compact-index",
+                             data=json.dumps({"index": idx, "target_vectors": 2000}).encode(),
+                             headers={"Content-Type": "application/json"})
+try:
+    with urllib.request.urlopen(req, timeout=900) as r:
+        print(json.dumps(json.load(r), ensure_ascii=False, indent=1))
+except urllib.error.HTTPError as e:
+    print("compact-index HTTP %s: %s" % (e.code, e.read().decode("utf-8", "ignore")[:600]))
+    sys.exit(1)
+PYEOF
+else
+  step 6 "skipped (/compact-index needs an index path)"
+fi
+
+echo
+if [ "$FAIL" -eq 0 ]; then
+  echo "ALL CHECKS PASSED (macOS)"
+else
+  echo "SOME CHECKS FAILED — see output above"
+fi
+exit "$FAIL"

package/tools/make-pack.mjs

@@ -0,0 +1,155 @@
+#!/usr/bin/env node
+// make-pack.mjs — one-command voice-pack addition for a dsh-plugin-tts pack repo.
+//
+// Scans a folder for a model (.pth) + index files (.index), copies them into
+// <repo>/packs/<id>/, computes size + sha256, and merges a manifest.json entry
+// (schema 2, multi-index). No dependencies beyond Node.
+//
+// Usage:
+//   node tools/make-pack.mjs --id azusa2 --name "Azusa 2" \
+//       --dir D:/incoming/azusa2 --repo . \
+//       [--desc "..."] [--author me] [--license author-owned]
+//       [--base-voice zh-CN-YunyangNeural] [--f0-method rmvpe] [--index-rate 0.75]
+//       [--version 1.0.0]
+//
+// Validate an existing repo (existence + sha256 + size for every pack):
+//   node tools/make-pack.mjs --check --repo .
+import {
+  readFileSync, writeFileSync, existsSync, mkdirSync, copyFileSync,
+  readdirSync, statSync
+} from 'node:fs';
+import { createHash } from 'node:crypto';
+import path from 'node:path';
+
+function parseArgs(argv) {
+  const out = {};
+  for (let i = 0; i < argv.length; i++) {
+    const a = argv[i];
+    if (!a.startsWith('--')) continue;
+    const key = a.slice(2);
+    const next = argv[i + 1];
+    if (next !== undefined && !next.startsWith('--')) { out[key] = next; i++; }
+    else out[key] = true;
+  }
+  return out;
+}
+
+const args = parseArgs(process.argv.slice(2));
+const repo = path.resolve(args.repo || '.');
+const manifestPath = path.join(repo, 'manifest.json');
+
+function sha256(p) {
+  return createHash('sha256').update(readFileSync(p)).digest('hex');
+}
+
+function loadManifest() {
+  try {
+    return JSON.parse(readFileSync(manifestPath, 'utf8'));
+  } catch (e) {
+    return { schema: 2, updated: '', packs: [] };
+  }
+}
+
+function saveManifest(m) {
+  m.schema = 2;
+  m.updated = new Date().toISOString();
+  writeFileSync(manifestPath, JSON.stringify(m, null, 2));
+}
+
+// ---- --check: validate every pack in the manifest ---------------------------
+if (args.check) {
+  const m = loadManifest();
+  let bad = 0;
+  const checkFile = (packId, label, f) => {
+    if (!f || !f.url) return;
+    const fp = path.join(repo, f.url);
+    if (!existsSync(fp)) { console.log(`MISSING  ${packId} ${label}: ${f.url}`); bad++; return; }
+    const s = sha256(fp);
+    if (s !== String(f.sha256).toLowerCase()) { console.log(`SHA-BAD  ${packId} ${label}: ${f.url}`); bad++; return; }
+    if (f.size && statSync(fp).size !== f.size) { console.log(`SIZE-BAD ${packId} ${label}: ${f.url}`); bad++; }
+    console.log(`ok       ${packId} ${label}: ${f.url}`);
+  };
+  for (const p of m.packs) {
+    checkFile(p.id, 'model', p.model);
+    for (const i of (p.indexes || [])) checkFile(p.id, 'index', i);
+    checkFile(p.id, 'index', p.index);
+  }
+  console.log(bad ? `\nFAIL: ${bad} problem(s) in ${m.packs.length} pack(s)` : `\nOK: ${m.packs.length} pack(s) validated`);
+  process.exit(bad ? 1 : 0);
+}
+
+// ---- add a pack ------------------------------------------------------------
+const id = String(args.id || '').trim();
+if (!id) { console.error('--id is required (e.g. --id azusa2)'); process.exit(1); }
+const srcDir = path.resolve(String(args.dir || ''));
+if (!existsSync(srcDir)) { console.error(`--dir not found: ${srcDir}`); process.exit(1); }
+
+const files = readdirSync(srcDir);
+const modelSrc = files.find(f => f.toLowerCase().endsWith('.pth'));
+const indexSrcs = files.filter(f => f.toLowerCase().endsWith('.index')).sort();
+if (!modelSrc) { console.error(`no .pth model found in ${srcDir}`); process.exit(1); }
+if (!indexSrcs.length) console.log('note: no .index files found -> index-free pack');
+
+const stem = id.replace(/[^\w.-]/g, '_');
+const packDir = path.join(repo, 'packs', stem);
+mkdirSync(packDir, { recursive: true });
+
+// model
+const modelName = stem + '.pth';
+const modelDst = path.join(packDir, modelName);
+copyFileSync(path.join(srcDir, modelSrc), modelDst);
+const model = {
+  url: `packs/${stem}/${modelName}`,
+  size: statSync(modelDst).size,
+  sha256: sha256(modelDst)
+};
+
+// indexes (each .index file becomes a selectable variant)
+const indexes = [];
+for (const src of indexSrcs) {
+  const dst = path.join(packDir, src);
+  copyFileSync(path.join(srcDir, src), dst);
+  const size = statSync(dst).size;
+  const base = path.basename(src, path.extname(src)).replace(/[^\w.-]/g, '_').slice(0, 24) || 'index';
+  indexes.push({
+    id: base,
+    name: `${base}（${(size / 1048576).toFixed(0)}MB）`,
+    url: `packs/${stem}/${src}`,
+    size,
+    sha256: sha256(dst)
+  });
+}
+
+const entry = {
+  id,
+  name: String(args.name || id),
+  description: String(args.desc || ''),
+  version: String(args.version || '1.0.0'),
+  author: String(args.author || ''),
+  license: String(args.license || 'unknown'),
+  baseVoice: String(args['base-voice'] || 'zh-CN-YunyangNeural'),
+  f0Method: String(args['f0-method'] || 'rmvpe'),
+  indexRate: Number(args['index-rate'] ?? 0.75),
+  model,
+  indexes
+};
+if (!indexes.length) delete entry.indexes;
+
+const m = loadManifest();
+if (!Array.isArray(m.packs)) m.packs = [];
+const i = m.packs.findIndex(p => p && p.id === id);
+if (i >= 0) m.packs[i] = entry;
+else m.packs.push(entry);
+saveManifest(m);
+
+console.log(`\nadded pack "${entry.name}" (${id}):`);
+console.log(`  model  ${model.url}  ${(model.size / 1048576).toFixed(1)}MB`);
+for (const idx of entry.indexes || []) {
+  console.log(`  index  ${idx.url}  ${(idx.size / 1048576).toFixed(1)}MB  (variant: ${idx.id})`);
+}
+console.log('\nnext steps:');
+console.log(`  git -C "${repo}" add packs/${stem} manifest.json`);
+console.log(`  git -C "${repo}" commit -m "add voice pack ${id}"`);
+console.log(`  git -C "${repo}" push origin main`);
+console.log('\nverify:');
+console.log(`  node tools/make-pack.mjs --check --repo "${repo}"`);