dsh-pet-live2d 3.1.3 → 3.2.0

package/lib/index.js

@@ -30,6 +30,7 @@ import { fileURLToPath } from 'node:url'
 import { createDisplayLayer } from './display.js'
 import { desktopHint, desktopSupported, resolveDesktopBinary } from './desktop.js'
 import { readSettings, writeSettings } from './settings.js'
+import { SOUND_PHASES, MAX_SOUND_BODY, decodeSound, readSound, resetSound, writeSound } from './sound-files.js'
 
 export const name = 'live2d-pet'
 
@@ -547,6 +548,8 @@ export function scanPet(dir, id) {
     // 台词：宠物自己的声音（问候 / 点击 / 摸头 / 摸尾巴 / 转晕 / 归位 / 每个相位）。
     // 用户能在设置里逐条改，改过的存浏览器；这里是**默认值**。
     lines: typeof block.lines === 'object' && block.lines !== null ? block.lines : {},
+    // null 表示旧宠物没有声明提示音，浏览器半区才使用内置兜底。
+    sounds: typeof block.sounds === 'object' && block.sounds !== null && !Array.isArray(block.sounds) ? block.sounds : null,
     // 互动反应候选（标签：动作的中文名或表情名）。摸头 / 摸尾巴 / 转晕各一组。
     patReactions: readReactionList(block.patReactions),
     tailReactions: readReactionList(block.tailReactions),
@@ -604,6 +607,8 @@ const BUNDLED_PET_HASHES = {
     '18840cd90fe70aa68632c45f77b4af1254595b5f963d261d2cbf34f6fbcaf579',
     // 1.0.1（2.3.0 ~ 2.3.2 随包的那份）：内容与上面那条相同，只是换了插件版本号
     '7c6cdb9c9f3d92636c388bffb3229c439cf01a65fe6a8a88499a8e4071f7884a',
+    // 1.1.0（会话提示音进入随包宠物之前）
+    '31eb2a210d3d99d2f2d53462b8a074ae22209579f7b8fcaa41dd5516e0a861b3',
   ],
 }
 
@@ -796,6 +801,7 @@ function sendJson(response, status, payload) {
   const body = Buffer.from(JSON.stringify(payload), 'utf8')
   response.writeHead(status, {
     'cache-control': 'no-store',
+    'x-content-type-options': 'nosniff',
     'content-type': 'application/json; charset=utf-8',
     'content-length': String(body.byteLength),
   })
@@ -901,6 +907,7 @@ function catalogRoute() {
           tailParts: pet.tailParts,
           partNames: pet.partNames,
           lines: pet.lines,
+          sounds: pet.sounds,
           patReactions: pet.patReactions,
           tailReactions: pet.tailReactions,
           spinReactions: pet.spinReactions,
@@ -1464,12 +1471,90 @@ function settingsRoute(home) {
   }
 }
 
+// 回环地址仍可能被 DNS 重绑定命中：写入接口只认服务器实际监听端口的本机 Host。
+function trustedSoundHost(request) {
+  const port = request.socket?.localPort
+  const host = request.headers.host?.toLowerCase()
+  return Number.isInteger(port) && [
+    `127.0.0.1:${port}`, `localhost:${port}`, `[::1]:${port}`,
+  ].includes(host)
+}
+
+// 音频接口的 pet id 必须来自当前目录；音频文件本身永不进入宠物目录。
+function soundRoute(home) {
+  return {
+    kind: 'prefix',
+    path: API + '/sound',
+    handler: async (request, response) => {
+      const fail = (status, error) => sendJson(response, status, { ok: false, error })
+      if (!loopbackOnly(request)) return fail(403, 'forbidden')
+      let parts
+      try {
+        parts = segmentsAfter(new URL(request.url ?? '/', 'http://pet.local').pathname, API + '/sound')?.map(decodeURIComponent)
+      } catch { return fail(400, 'invalid-path') }
+      if (!parts || (parts.length !== 1 && parts.length !== 2)
+        || !/^[A-Za-z0-9_-]+$/.test(parts[0])
+        || (parts.length === 2 && !SOUND_PHASES.includes(parts[1]))) return fail(404, 'not-found')
+      const [id, phase] = parts
+      if (!buildCatalog().some((pet) => pet.id === id)) return fail(404, 'not-found')
+      if (request.method === 'POST') {
+        if (phase === undefined) return fail(405, 'method-not-allowed')
+        if (!trustedSoundHost(request)) return fail(403, 'untrusted-host')
+        // 浏览器的跨站请求不能更改本地文件；代理端用自己的 Host 转发。
+        const origin = request.headers.origin
+        if (typeof origin === 'string' && origin !== `http://${request.headers.host}` && origin !== `https://${request.headers.host}`) return fail(403, 'cross-origin')
+        if (!/^application\/json(?:\s*;|$)/i.test(request.headers['content-type'] ?? '')) return fail(415, 'json-required')
+        let size = 0
+        const chunks = []
+        try {
+          for await (const chunk of request) {
+            size += chunk.length
+            if (size > MAX_SOUND_BODY) return fail(413, 'body-too-large')
+            chunks.push(chunk)
+          }
+        } catch { return fail(400, 'bad-body') }
+        let parsed
+        try { parsed = JSON.parse(Buffer.concat(chunks).toString('utf8')) } catch { return fail(400, 'bad-body') }
+        if (!parsed || typeof parsed !== 'object' || Array.isArray(parsed)) return fail(400, 'bad-body')
+        try {
+          if (parsed.action === 'reset' && Object.keys(parsed).length === 1) resetSound(home, id, phase)
+          else if (Object.keys(parsed).length === 1 && Object.hasOwn(parsed, 'base64')) {
+            const bytes = decodeSound(parsed.base64)
+            if (!bytes) return fail(400, 'invalid-audio')
+            writeSound(home, id, phase, bytes)
+          } else return fail(400, 'bad-body')
+        } catch { return fail(500, 'write-failed') }
+        return sendJson(response, 200, { ok: true })
+      }
+      if (request.method !== 'GET' && request.method !== 'HEAD') return fail(405, 'method-not-allowed')
+      if (phase === undefined) {
+        const sounds = {}
+        for (const key of SOUND_PHASES) {
+          const sound = readSound(home, id, key)
+          if (sound) sounds[key] = {
+            url: `${API}/sound/${encodeURIComponent(id)}/${key}?token=${sound.token}`,
+            mime: sound.mime, bytes: sound.bytes.length, token: sound.token,
+          }
+        }
+        return sendJson(response, 200, { ok: true, sounds })
+      }
+      const sound = readSound(home, id, phase)
+      if (!sound) return fail(404, 'not-found')
+      response.writeHead(200, {
+        'cache-control': 'no-store', 'x-content-type-options': 'nosniff',
+        'content-type': sound.mime, 'content-length': String(sound.bytes.length),
+      })
+      response.end(request.method === 'HEAD' ? undefined : sound.bytes)
+    },
+  }
+}
+
 /** The complete route table this plugin owns. */
 export function buildRoutes(hub, display, home) {
   const routes = [catalogRoute(), assetRoute(), runtimeRoute(), eventsRoute(hub)]
   if (display !== undefined) routes.push(layerRoute(display))
   // 共享设置：不依赖显示层（测试里也可能只给 home）。
-  if (typeof home === 'string' && home !== '') routes.push(settingsRoute(home))
+  if (typeof home === 'string' && home !== '') routes.push(settingsRoute(home), soundRoute(home))
   return routes
 }
 

package/lib/sound-files.js

@@ -0,0 +1,231 @@
+// 用户上传的相位音频只落在独立目录；请求路径和文件名均由白名单生成。
+import { createHash, randomBytes } from 'node:crypto'
+import { closeSync, existsSync, lstatSync, mkdirSync, openSync, readFileSync, renameSync, unlinkSync, writeFileSync } from 'node:fs'
+import { join } from 'node:path'
+
+export const SOUND_PHASES = Object.freeze(['thinking', 'tool', 'waiting', 'asking', 'helper', 'queued', 'done', 'failed'])
+export const MAX_AUDIO = 1024 * 1024
+export const MAX_SOUND_BODY = 2 * 1024 * 1024
+
+// ---- 容器判据：**与桌面端 Rust 宿主逐条一致** -------------------------------
+//
+// `dsh-live2d-pet-desktop/src-tauri/src/host/http.rs` 的 `valid_wav` / `valid_ogg` /
+// `valid_mp3` / `ogg_checksum` 是同一套规则的另一个实现。**改一边必须同时改另一边**，
+// 而且两边都要跑 `tools/sound-vectors.json` 的合同向量（`tools/browser-test/test-sound-upload.mjs`
+// 与 Rust 单测 `sound_vectors_contract` 都读它）—— 判据分叉的症状是"网页端传得进、
+// 桌面端读不出来/相位的已上传消失"，而且两边都不报错。
+//
+// 只认魔数是不够的：那样一个 4 字节的 "OggS"、一个乱改过长度的 WAV 都会写进
+// `$DSH_HOME/pet-sounds/`，桌面端却读不出来。
+const u16 = (bytes, at) => bytes[at] | (bytes[at + 1] << 8)
+const u32 = (bytes, at) => (bytes[at] | (bytes[at + 1] << 8) | (bytes[at + 2] << 16) | (bytes[at + 3] << 24)) >>> 0
+const at4 = (bytes, from, to) => bytes.toString('latin1', from, to)
+
+function validWav(bytes) {
+  if (bytes.length < 12 || at4(bytes, 0, 4) !== 'RIFF' || at4(bytes, 8, 12) !== 'WAVE') return false
+  // RIFF 头里的长度字段：流式录音这类文件常写 0 或 0xFFFFFFFF（"长度未知/一直写到尾"），
+  // 浏览器照样能播，所以只对这三个值放行，其余仍要求精确匹配（畸形头照旧拒绝）。
+  const declared = u32(bytes, 4)
+  if (declared !== 0 && declared !== 0xffffffff && declared !== bytes.length - 8) return false
+  let offset = 12
+  let blockAlign
+  let dataSize
+  while (offset < bytes.length) {
+    if (offset + 8 > bytes.length) return false
+    const id = at4(bytes, offset, offset + 4)
+    const size = u32(bytes, offset + 4)
+    const end = offset + 8 + size
+    if (end > bytes.length) return false // 块越界（对应 Rust 的 get() 失败）
+    if (id === 'fmt ') {
+      if (blockAlign !== undefined || size < 16) return false
+      const format = u16(bytes, offset + 8)
+      const channels = u16(bytes, offset + 10)
+      const rate = u32(bytes, offset + 12)
+      const byteRate = u32(bytes, offset + 16)
+      const align = u16(bytes, offset + 20)
+      const bits = u16(bytes, offset + 22)
+      if ((format !== 1 && format !== 3) || channels === 0 || rate === 0 || bits === 0 || bits % 8 !== 0) return false
+      if (format === 3 && bits !== 32 && bits !== 64) return false
+      if (align !== channels * (bits / 8)) return false
+      if (byteRate !== rate * align) return false // rate ≤ 2^32、align ≤ 2^16：JS 双精度够用
+      blockAlign = align
+    } else if (id === 'data') {
+      if (dataSize !== undefined) return false
+      dataSize = size
+    }
+    offset = end + (size % 2) // 块按偶数字节对齐
+    if (offset > bytes.length) return false
+  }
+  if (blockAlign === undefined || dataSize === undefined) return false
+  return dataSize > 0 && dataSize % blockAlign === 0
+}
+
+/** Ogg 的 CRC32（多项式 0x04c11db7，高位先行，校验时把 22..26 这 4 字节当 0）。 */
+function oggChecksum(page) {
+  let crc = 0
+  for (let index = 0; index < page.length; index += 1) {
+    const byte = index >= 22 && index < 26 ? 0 : page[index]
+    crc = (crc ^ (byte << 24)) >>> 0
+    for (let bit = 0; bit < 8; bit += 1) {
+      crc = ((crc << 1) ^ ((crc & 0x80000000) !== 0 ? 0x04c11db7 : 0)) >>> 0
+    }
+  }
+  return crc
+}
+
+function validOgg(bytes) {
+  if (bytes.length < 27) return false
+  if (at4(bytes, 0, 4) !== 'OggS' || bytes[4] !== 0 || bytes[5] !== 2) return false
+  if (u32(bytes, 18) !== 0) return false // 页序号必须是 0：只认第一页
+  const segmentCount = bytes[26]
+  const segmentsAt = 27
+  if (segmentsAt + segmentCount > bytes.length) return false
+  const segments = bytes.subarray(segmentsAt, segmentsAt + segmentCount)
+  let payloadSize = 0
+  for (const size of segments) payloadSize += size
+  const start = segmentsAt + segmentCount
+  if (start + payloadSize > bytes.length) return false
+  const page = bytes.subarray(0, start + payloadSize)
+  if (oggChecksum(page) !== u32(bytes, 22)) return false
+  let packetEnd = -1
+  for (let index = 0; index < segments.length; index += 1) {
+    if (segments[index] < 255) { packetEnd = index; break }
+  }
+  if (packetEnd < 0) return false
+  let packetSize = 0
+  for (let index = 0; index <= packetEnd; index += 1) packetSize += segments[index]
+  const packet = page.subarray(start, start + packetSize)
+  if (packet.length >= 8 && at4(packet, 0, 8) === 'OpusHead') {
+    if (packet.length < 19 || packet[8] !== 1 || packet[9] === 0) return false
+    if (packet[18] === 0) return packet[9] <= 2 && packet.length === 19
+    return packet[18] !== 255 && packet.length >= 21 + packet[9]
+  }
+  if (packet.length >= 7 && packet[0] === 0x01 && at4(packet, 1, 7) === 'vorbis') {
+    return packet.length === 30
+      && packet[7] === 0 && packet[8] === 0 && packet[9] === 0 && packet[10] === 0
+      && packet[11] !== 0
+      && !(packet[12] === 0 && packet[13] === 0 && packet[14] === 0 && packet[15] === 0)
+      && (packet[28] & 15) >= 6 && (packet[28] >> 4) >= (packet[28] & 15) && (packet[28] >> 4) <= 13
+      && packet[29] === 1
+  }
+  return false
+}
+
+const MP3_BITRATES = {
+  mpeg1L1: [0, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448, 0],
+  mpeg1L2: [0, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384, 0],
+  mpeg1L3: [0, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 0],
+  mpeg2L1: [0, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256, 0],
+  mpeg2L23: [0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160, 0],
+}
+const mp3Bitrates = (version, layer) => {
+  if (version === 3) return layer === 3 ? MP3_BITRATES.mpeg1L1 : layer === 2 ? MP3_BITRATES.mpeg1L2 : MP3_BITRATES.mpeg1L3
+  return layer === 3 ? MP3_BITRATES.mpeg2L1 : MP3_BITRATES.mpeg2L23
+}
+
+function validMp3(bytes) {
+  let offset = 0
+  if (at4(bytes, 0, 3) === 'ID3') {
+    if (bytes.length < 10) return false
+    const major = bytes[3]
+    const flags = bytes[5]
+    if (major < 2 || major > 4 || bytes[4] === 0xff) return false
+    for (let index = 6; index < 10; index += 1) {
+      if ((bytes[index] & 0x80) !== 0) return false // 尺寸字段是 synchsafe 的
+    }
+    const reserved = major === 4 ? 0x0f : major === 3 ? 0x1f : 0x3f
+    if ((flags & reserved) !== 0) return false
+    let tagSize = 0
+    for (let index = 6; index < 10; index += 1) tagSize = tagSize * 128 + bytes[index]
+    offset = 10 + tagSize + (major === 4 && (flags & 0x10) !== 0 ? 10 : 0)
+  }
+  if (offset + 4 > bytes.length) return false
+  const b1 = bytes[offset + 1]
+  const b2 = bytes[offset + 2]
+  const b3 = bytes[offset + 3]
+  if (bytes[offset] !== 0xff || (b1 & 0xe0) !== 0xe0) return false
+  const version = (b1 >> 3) & 3
+  const layer = (b1 >> 1) & 3
+  const bitrateIndex = (b2 >> 4) & 0x0f
+  const rateIndex = (b2 >> 2) & 3
+  if (version === 1 || layer === 0 || bitrateIndex === 0 || bitrateIndex === 15 || rateIndex === 3) return false
+  const bitrate = mp3Bitrates(version, layer)[bitrateIndex] * 1000
+  const rate = Math.floor([44100, 48000, 32000][rateIndex] / (version === 3 ? 1 : version === 2 ? 2 : 4))
+  const padding = (b2 >> 1) & 1
+  const frameSize = layer === 3
+    ? Math.floor(12 * bitrate / rate + padding) * 4
+    : Math.floor((layer === 1 && version !== 3 ? 72 : 144) * bitrate / rate) + padding
+  return (b3 & 3) !== 2 && bytes.length - offset >= frameSize
+}
+
+export function soundMime(bytes) {
+  if (validWav(bytes)) return 'audio/wav'
+  if (validOgg(bytes)) return 'audio/ogg'
+  if (validMp3(bytes)) return 'audio/mpeg'
+  return undefined
+}
+
+function soundDir(home, create) {
+  if (create) mkdirSync(home, { recursive: true })
+  const dir = join(home, 'pet-sounds')
+  if (create && !existsSync(dir)) mkdirSync(dir, { mode: 0o700 })
+  const stat = lstatSync(dir)
+  if (!stat.isDirectory() || stat.isSymbolicLink()) throw new Error('unsafe-path')
+  return dir
+}
+
+function soundPath(home, id, phase, create = false) {
+  if (!/^[A-Za-z0-9_-]+$/.test(id) || !SOUND_PHASES.includes(phase)) throw new Error('invalid-path')
+  return join(soundDir(home, create), `${id}--${phase}.audio`)
+}
+
+export function readSound(home, id, phase) {
+  try {
+    const path = soundPath(home, id, phase)
+    const stat = lstatSync(path)
+    if (!stat.isFile() || stat.isSymbolicLink() || stat.size > MAX_AUDIO) return undefined
+    const bytes = readFileSync(path)
+    const mime = soundMime(bytes)
+    if (!mime || bytes.length > MAX_AUDIO) return undefined
+    return { bytes, mime, token: createHash('sha256').update(bytes).digest('hex') }
+  } catch { return undefined }
+}
+
+export function writeSound(home, id, phase, bytes) {
+  const path = soundPath(home, id, phase, true)
+  if (existsSync(path) && (lstatSync(path).isSymbolicLink() || !lstatSync(path).isFile())) throw new Error('unsafe-path')
+  const temp = join(soundDir(home, false), `.${randomBytes(16).toString('hex')}.tmp`)
+  const fd = openSync(temp, 'wx', 0o600)
+  try {
+    writeFileSync(fd, bytes)
+    closeSync(fd)
+    renameSync(temp, path)
+  } catch (error) {
+    try { closeSync(fd) } catch { /* 已关闭 */ }
+    try { unlinkSync(temp) } catch { /* 已移走 */ }
+    throw error
+  }
+}
+
+export function resetSound(home, id, phase) {
+  let path
+  try { path = soundPath(home, id, phase) }
+  catch (error) {
+    if (error.code === 'ENOENT') return
+    throw error
+  }
+  try {
+    const stat = lstatSync(path)
+    if (!stat.isFile() || stat.isSymbolicLink()) throw new Error('unsafe-path')
+    unlinkSync(path)
+  } catch (error) {
+    if (error.code !== 'ENOENT') throw error
+  }
+}
+
+export function decodeSound(value) {
+  if (typeof value !== 'string' || value.length === 0 || value.length > Math.ceil(MAX_AUDIO / 3) * 4 + 4 || !/^(?:[A-Za-z0-9+/]{4})*(?:[A-Za-z0-9+/]{2}==|[A-Za-z0-9+/]{3}=)?$/.test(value)) return undefined
+  const bytes = Buffer.from(value, 'base64')
+  if (bytes.length === 0 || bytes.length > MAX_AUDIO || bytes.toString('base64') !== value || !soundMime(bytes)) return undefined
+  return bytes
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-pet-live2d",
-  "version": "3.1.3",
+  "version": "3.2.0",
   "description": "Live2D 桌宠插件：为 DSH Web GUI 挂一只可拖动、会跟随鼠标、能换动作与表情的 Live2D 宠物。",
   "type": "module",
   "main": "lib/index.js",
@@ -49,8 +49,8 @@
     }
   },
   "optionalDependencies": {
-    "dsh-pet-live2d-desktop-darwin-arm64": "3.1.3",
-    "dsh-pet-live2d-desktop-win32-x64": "3.1.3"
+    "dsh-pet-live2d-desktop-darwin-arm64": "3.2.0",
+    "dsh-pet-live2d-desktop-win32-x64": "3.2.0"
   },
   "devDependencies": {
     "esbuild": "^0.25.10",

package/pets/ds-whale-girl/README.md

@@ -58,6 +58,8 @@
 | pet.json 字段 | 变成设置页里的什么 |
 |---|---|
 | `live2d.looksByPhase` | 每个相位的默认池子：一个槽位一个选择 = 那个槽位池子里"权重 1 的一条" |
+| `live2d.lines.phase` | 每个相位的默认气泡台词 |
+| `live2d.sounds` | 每个相位的提示音符 `[频率 Hz, 起始秒数]`；这只宠物的完成、失败、待批准、待回答各有两音 |
 | `live2d.motions` | 每个相位的默认动作（`done` 吹泡泡糖、`failed` 鲸鱼喷水就是这么来的）|
 | `expressionSlots[].options[].pairs` | 该选项的「同时」关系（选了它就一起点亮哪个槽位） |
 | `expressionSlots[].options[].requires` | 该选项的「前提」关系（必须先处于哪个状态才播得出来） |

package/pets/ds-whale-girl/pet.json

@@ -3,7 +3,7 @@
   "id": "ds-whale-girl",
   "displayName": "DS鲸鱼娘",
   "description": "氵六青的 DS 鲸鱼娘 Live2D 桌宠：8 组动作 + 44 个表情/道具，点她会有反应。",
-  "version": "1.1.0",
+  "version": "1.1.1",
   "author": "模型：上善无形 / ZipZipPipe / 氵六青 · 插件：A8Chann",
   "license": "CC BY-NC-SA 4.0 (attribution, NON-COMMERCIAL, share-alike). Character: original OC by 上善无形 (https://www.pixiv.net/users/62155430); DeepSeek maid redesign by ZipZipPipe (https://www.pixiv.net/users/18604994); this Live2D model by 氵六青 (https://space.bilibili.com/11272072), who authorised redistribution here — which does NOT lift the base NC/SA terms. Commercial use requires permission from each rights holder. See NOTICE.md.",
   "homepage": "https://space.bilibili.com/11272072",
@@ -661,6 +661,12 @@
         "failed": "呜…出错了"
       }
     },
+    "sounds": {
+      "done": [[660, 0], [880, 0.13]],
+      "failed": [[370, 0], [277, 0.16]],
+      "waiting": [[520, 0], [660, 0.16]],
+      "asking": [[520, 0], [780, 0.16]]
+    },
     "patReactions": [
       "重锤出击",
       "问号",