feat: 增加 SILK 语音转 WAV 功能

新增 silk-wasm 依赖，实现语音附件检测与 SILK/AMR 格式转 WAV，并在网关中集成语音转换处理流程。

package-lock.json

@@ -11,6 +11,7 @@
         "ws"
       ],
       "dependencies": {
+        "silk-wasm": "^3.7.1",
         "ws": "^8.18.0"
       },
       "bin": {
@@ -8937,6 +8938,15 @@
         "url": "https://github.com/sponsors/isaacs"
       }
     },
+    "node_modules/silk-wasm": {
+      "version": "3.7.1",
+      "resolved": "https://registry.npmjs.org/silk-wasm/-/silk-wasm-3.7.1.tgz",
+      "integrity": "sha512-mXPwLRtZxrYV3TZx41jMAeKc80wvmyrcXIcs8HctFxK15Ahz2OJQENYhNgEPeCEOdI6Mbx1NxQsqxzwc3DKerw==",
+      "license": "MIT",
+      "engines": {
+        "node": ">=16.11.0"
+      }
+    },
     "node_modules/simple-git": {
       "version": "3.30.0",
       "resolved": "https://registry.npmjs.org/simple-git/-/simple-git-3.30.0.tgz",

package.json

@@ -19,13 +19,19 @@
     "moltbot.plugin.json"
   ],
   "clawdbot": {
-    "extensions": ["./index.ts"]
+    "extensions": [
+      "./index.ts"
+    ]
   },
   "moltbot": {
-    "extensions": ["./index.ts"]
+    "extensions": [
+      "./index.ts"
+    ]
   },
   "openclaw": {
-    "extensions": ["./index.ts"]
+    "extensions": [
+      "./index.ts"
+    ]
   },
   "scripts": {
     "build": "tsc || true",
@@ -33,9 +39,11 @@
     "prepack": "npm install --omit=dev"
   },
   "dependencies": {
+    "silk-wasm": "^3.7.1",
     "ws": "^8.18.0"
   },
   "bundledDependencies": [
+    "silk-wasm",
     "ws"
   ],
   "devDependencies": {
@@ -48,5 +56,9 @@
     "moltbot": "*",
     "openclaw": "*"
   },
-  "homepage": "https://github.com/sliverp/qqbot"
+  "homepage": "https://github.com/sliverp/qqbot",
+  "bundleDependencies": [
+    "silk-wasm",
+    "ws"
+  ]
 }

src/gateway.ts

@@ -9,6 +9,7 @@ import { getQQBotRuntime } from "./runtime.js";
 import { startImageServer, isImageServerRunning, downloadFile, type ImageServerConfig } from "./image-server.js";
 import { getImageSize, formatQQBotMarkdownImage, hasQQBotImageSize, DEFAULT_IMAGE_SIZE } from "./utils/image-size.js";
 import { parseQQBotPayload, encodePayloadForCron, isCronReminderPayload, isMediaPayload, type CronReminderPayload, type MediaPayload } from "./utils/payload.js";
+import { convertSilkToWav, isVoiceAttachment, formatDuration } from "./utils/audio-convert.js";
 
 // QQ Bot intents - 按权限级别分组
 const INTENTS = {
@@ -539,9 +540,9 @@ openclaw cron add \\
         const downloadDir = path.join(process.env.HOME || "/home/ubuntu", "clawd", "downloads");
         
         if (event.attachments?.length) {
-          // ============ 接收图片的自然语言描述生成 ============
-          // 根据需求 4：将图片信息转换为自然语言描述，便于 AI 理解
+          // ============ 接收附件描述生成（图片 / 语音 / 其他） ============
           const imageDescriptions: string[] = [];
+          const voiceDescriptions: string[] = [];
           const otherAttachments: string[] = [];
           
           for (const att of event.attachments) {
@@ -563,6 +564,46 @@ openclaw cron add \\
 - 发送时间：${timestamp}
 
 请根据图片内容进行回复。`);
+              } else if (isVoiceAttachment(att)) {
+                // ============ 语音消息处理：SILK → WAV ============
+                log?.info(`[qqbot:${account.accountId}] Voice attachment detected: ${att.filename}, converting SILK to WAV...`);
+                try {
+                  const result = await convertSilkToWav(localPath, downloadDir);
+                  if (result) {
+                    const durationStr = formatDuration(result.duration);
+                    log?.info(`[qqbot:${account.accountId}] Voice converted: ${result.wavPath} (duration: ${durationStr})`);
+                    
+                    const timestamp = new Date().toLocaleString("zh-CN", { timeZone: "Asia/Shanghai" });
+                    voiceDescriptions.push(`
+用户发送了一条语音消息：
+- 语音文件：${result.wavPath}
+- 语音时长：${durationStr}
+- 原始文件：${localPath}
+- 消息ID：${event.messageId}
+- 发送时间：${timestamp}
+
+请使用语音文件进行语音转文字处理后回复用户。`);
+                  } else {
+                    // SILK 解码失败，保留原始文件
+                    log?.info(`[qqbot:${account.accountId}] Voice file is not SILK format, keeping original: ${localPath}`);
+                    voiceDescriptions.push(`
+用户发送了一条语音消息（非SILK格式，无法转换）：
+- 语音文件：${localPath}
+- 原始格式：${att.filename || "unknown"}
+- 消息ID：${event.messageId}
+
+请告知用户该语音格式暂不支持解析。`);
+                  }
+                } catch (convertErr) {
+                  log?.error(`[qqbot:${account.accountId}] Voice conversion failed: ${convertErr}`);
+                  voiceDescriptions.push(`
+用户发送了一条语音消息（转换失败）：
+- 原始文件：${localPath}
+- 错误信息：${convertErr}
+- 消息ID：${event.messageId}
+
+请告知用户语音处理出现问题。`);
+                }
               } else {
                 otherAttachments.push(`[附件: ${localPath}]`);
               }
@@ -591,10 +632,13 @@ openclaw cron add \\
             }
           }
           
-          // 组合附件信息：先图片描述，后其他附件
+          // 组合附件信息：先图片描述，后语音描述，后其他附件
           if (imageDescriptions.length > 0) {
             attachmentInfo += "\n" + imageDescriptions.join("\n");
           }
+          if (voiceDescriptions.length > 0) {
+            attachmentInfo += "\n" + voiceDescriptions.join("\n");
+          }
           if (otherAttachments.length > 0) {
             attachmentInfo += "\n" + otherAttachments.join("\n");
           }

src/utils/audio-convert.ts

@@ -0,0 +1,138 @@
+import * as fs from "node:fs";
+import * as path from "node:path";
+import { decode, isSilk } from "silk-wasm";
+
+/**
+ * 检查文件是否为 SILK 格式（QQ/微信语音常用格式）
+ * QQ 语音文件通常以 .amr 扩展名保存，但实际编码可能是 SILK v3
+ * SILK 文件头部标识: 0x02 "#!SILK_V3"
+ */
+function isSilkFile(filePath: string): boolean {
+  try {
+    const buf = fs.readFileSync(filePath);
+    return isSilk(new Uint8Array(buf.buffer, buf.byteOffset, buf.byteLength));
+  } catch {
+    return false;
+  }
+}
+
+/**
+ * 将 PCM (s16le) 数据封装为 WAV 文件格式
+ * WAV = 44 字节 RIFF 头 + PCM 原始数据
+ */
+function pcmToWav(pcmData: Uint8Array, sampleRate: number, channels: number = 1, bitsPerSample: number = 16): Buffer {
+  const byteRate = sampleRate * channels * (bitsPerSample / 8);
+  const blockAlign = channels * (bitsPerSample / 8);
+  const dataSize = pcmData.length;
+  const headerSize = 44;
+  const fileSize = headerSize + dataSize;
+
+  const buffer = Buffer.alloc(fileSize);
+
+  // RIFF header
+  buffer.write("RIFF", 0);
+  buffer.writeUInt32LE(fileSize - 8, 4);
+  buffer.write("WAVE", 8);
+
+  // fmt sub-chunk
+  buffer.write("fmt ", 12);
+  buffer.writeUInt32LE(16, 16);         // sub-chunk size
+  buffer.writeUInt16LE(1, 20);          // PCM format
+  buffer.writeUInt16LE(channels, 22);
+  buffer.writeUInt32LE(sampleRate, 24);
+  buffer.writeUInt32LE(byteRate, 28);
+  buffer.writeUInt16LE(blockAlign, 32);
+  buffer.writeUInt16LE(bitsPerSample, 34);
+
+  // data sub-chunk
+  buffer.write("data", 36);
+  buffer.writeUInt32LE(dataSize, 40);
+  Buffer.from(pcmData.buffer, pcmData.byteOffset, pcmData.byteLength).copy(buffer, headerSize);
+
+  return buffer;
+}
+
+/**
+ * 去除 QQ 语音文件的 AMR 头（如果存在）
+ * QQ 的 .amr 文件可能在 SILK 数据前有 "#!AMR\n" 头（6 字节）
+ * 需要去除后才能被 silk-wasm 正确解码
+ */
+function stripAmrHeader(buf: Buffer): Buffer {
+  const AMR_HEADER = Buffer.from("#!AMR\n");
+  if (buf.length > 6 && buf.subarray(0, 6).equals(AMR_HEADER)) {
+    return buf.subarray(6);
+  }
+  return buf;
+}
+
+/**
+ * 将 SILK/AMR 语音文件转换为 WAV 格式
+ *
+ * @param inputPath 输入文件路径（.amr / .silk / .slk）
+ * @param outputDir 输出目录（默认与输入文件同目录）
+ * @returns 转换后的 WAV 文件路径，失败返回 null
+ */
+export async function convertSilkToWav(
+  inputPath: string,
+  outputDir?: string,
+): Promise<{ wavPath: string; duration: number } | null> {
+  if (!fs.existsSync(inputPath)) {
+    return null;
+  }
+
+  const fileBuf = fs.readFileSync(inputPath);
+
+  // 去除可能的 AMR 头
+  const strippedBuf = stripAmrHeader(fileBuf);
+
+  // 转为 Uint8Array 以兼容 silk-wasm 类型要求
+  const rawData = new Uint8Array(strippedBuf.buffer, strippedBuf.byteOffset, strippedBuf.byteLength);
+
+  // 验证是否为 SILK 格式
+  if (!isSilk(rawData)) {
+    return null;
+  }
+
+  // SILK 解码为 PCM (s16le)
+  // QQ 语音通常采样率为 24000Hz
+  const sampleRate = 24000;
+  const result = await decode(rawData, sampleRate);
+
+  // PCM → WAV
+  const wavBuffer = pcmToWav(result.data, sampleRate);
+
+  // 写入 WAV 文件
+  const dir = outputDir || path.dirname(inputPath);
+  if (!fs.existsSync(dir)) {
+    fs.mkdirSync(dir, { recursive: true });
+  }
+  const baseName = path.basename(inputPath, path.extname(inputPath));
+  const wavPath = path.join(dir, `${baseName}.wav`);
+  fs.writeFileSync(wavPath, wavBuffer);
+
+  return { wavPath, duration: result.duration };
+}
+
+/**
+ * 判断是否为语音附件（根据 content_type 或文件扩展名）
+ */
+export function isVoiceAttachment(att: { content_type?: string; filename?: string }): boolean {
+  if (att.content_type === "voice" || att.content_type?.startsWith("audio/")) {
+    return true;
+  }
+  const ext = att.filename ? path.extname(att.filename).toLowerCase() : "";
+  return [".amr", ".silk", ".slk"].includes(ext);
+}
+
+/**
+ * 格式化语音时长为可读字符串
+ */
+export function formatDuration(durationMs: number): string {
+  const seconds = Math.round(durationMs / 1000);
+  if (seconds < 60) {
+    return `${seconds}秒`;
+  }
+  const minutes = Math.floor(seconds / 60);
+  const remainSeconds = seconds % 60;
+  return remainSeconds > 0 ? `${minutes}分${remainSeconds}秒` : `${minutes}分钟`;
+}