使用 qwen 进行语音识别，加入 tag

Co-authored-by: Copilot <copilot@github.com>
2026-04-29 20:10:08 +08:00
parent 040273586b
commit 206a561ee7
26 changed files with 3720 additions and 443 deletions
@@ -0,0 +1,135 @@
+#!/usr/bin/env node
+const fs = require("node:fs");
+const Module = require("node:module");
+const path = require("node:path");
+
+const rootDir = path.resolve(__dirname, "..");
+
+function loadEnvFile(filePath, override = false) {
+  if (!fs.existsSync(filePath)) return;
+
+  const lines = fs.readFileSync(filePath, "utf8").split(/\r?\n/);
+  for (const line of lines) {
+    const trimmed = line.trim();
+    if (!trimmed || trimmed.startsWith("#")) continue;
+
+    const normalized = trimmed.startsWith("export ")
+      ? trimmed.slice("export ".length).trim()
+      : trimmed;
+    const match = normalized.match(/^([A-Za-z_][A-Za-z0-9_]*)\s*=\s*(.*)$/);
+    if (!match) continue;
+
+    const key = match[1];
+    let value = match[2].trim();
+    if (
+      (value.startsWith('"') && value.endsWith('"')) ||
+      (value.startsWith("'") && value.endsWith("'"))
+    ) {
+      value = value.slice(1, -1);
+    }
+    value = value.replace(/\\n/g, "\n");
+
+    if (override || process.env[key] === undefined) {
+      process.env[key] = value;
+    }
+  }
+}
+
+function registerTypeScriptRuntime() {
+  const ts = require("typescript");
+  const originalResolveFilename = Module._resolveFilename;
+
+  Module._resolveFilename = function resolveFilename(
+    request,
+    parent,
+    isMain,
+    options,
+  ) {
+    if (request.startsWith("@/")) {
+      return originalResolveFilename.call(
+        this,
+        path.join(rootDir, request.slice(2)),
+        parent,
+        isMain,
+        options,
+      );
+    }
+    return originalResolveFilename.call(this, request, parent, isMain, options);
+  };
+
+  const compile = (module, filename) => {
+    const source = fs.readFileSync(filename, "utf8");
+    const output = ts.transpileModule(source, {
+      fileName: filename,
+      compilerOptions: {
+        module: ts.ModuleKind.CommonJS,
+        target: ts.ScriptTarget.ES2022,
+        esModuleInterop: true,
+        resolveJsonModule: true,
+        jsx: ts.JsxEmit.ReactJSX,
+        moduleResolution: ts.ModuleResolutionKind.Node10,
+      },
+    }).outputText;
+    module._compile(output, filename);
+  };
+
+  require.extensions[".ts"] = compile;
+  require.extensions[".tsx"] = compile;
+  require.extensions[".md"] = (module, filename) => {
+    module.exports = fs.readFileSync(filename, "utf8");
+  };
+}
+
+function getAwemeId() {
+  const args = process.argv.slice(2);
+  const fromFlag = args.find((arg) => arg.startsWith("--aweme-id="));
+  if (fromFlag) return fromFlag.slice("--aweme-id=".length).trim();
+
+  const flagIndex = args.findIndex(
+    (arg) => arg === "--aweme-id" || arg === "-a",
+  );
+  if (flagIndex >= 0) return (args[flagIndex + 1] || "").trim();
+
+  return (args.find((arg) => !arg.startsWith("-")) || "").trim();
+}
+
+function usage() {
+  console.log("用法: npm run asr -- <aweme_id>");
+  console.log("或:   node scripts/asr-cli.cjs --aweme-id <aweme_id>");
+}
+
+function logStep(message) {
+  const now = new Date().toISOString();
+  console.log(`[asr ${now}] ${message}`);
+}
+
+async function main() {
+  loadEnvFile(path.join(rootDir, ".env"));
+  loadEnvFile(path.join(rootDir, ".env.local"), true);
+  registerTypeScriptRuntime();
+
+  const awemeId = getAwemeId();
+  if (!awemeId) {
+    usage();
+    process.exitCode = 1;
+    return;
+  }
+
+  logStep(`开始执行 aweme_id=${awemeId}`);
+
+  const { transcriptAweme } = require(path.join(rootDir, "app/api/asr"));
+  const { prisma } = require(path.join(rootDir, "lib/prisma"));
+
+  try {
+    const result = await transcriptAweme(awemeId, { log: logStep });
+    logStep("最终结果:");
+    console.log(JSON.stringify(result, null, 2));
+  } finally {
+    await prisma.$disconnect();
+  }
+}
+
+main().catch((error) => {
+  console.error("[asr] 执行失败:", error);
+  process.exitCode = 1;
+});