@biunow/agent-biuart 0.0.0-stage → 0.1.0

package/studio/to-english-prompt.mjs

@@ -0,0 +1,149 @@
+/** 把用户看到的中文提示词收成交给大模型的英文提示词。画风名保持原样，方便对上模板。 */
+
+const REPLACEMENTS = [
+  [/there is no word of mouth/gi, 'no voiceover'],
+  [/no word of mouth/gi, 'no voiceover'],
+  [/word of mouth/gi, 'voiceover'],
+  [/no spoken words?/gi, 'no voiceover'],
+  [/without spoken words?/gi, 'no voiceover'],
+  [/no oral (?:broadcast|narration)/gi, 'no voiceover'],
+  [/without oral (?:broadcast|narration)/gi, 'with no voiceover'],
+  [/oral broadcast/gi, 'voiceover'],
+  [/there was no voiceover/gi, 'no voiceover'],
+  [/horizontal screen/gi, 'landscape frame'],
+  [/vertical screen/gi, 'portrait frame'],
+];
+
+const SOURCE_FLAGS = [
+  [/没有口播|不要口播|无口播|没有旁白|不要旁白|无旁白|不依赖旁白|不要另起旁白|全程靠字幕/, 'No voiceover.'],
+  [/口播用你上传的录音|只用上传的录音/, 'Use the uploaded recording as the voiceover.'],
+  [/横屏/, 'Landscape frame.'],
+  [/竖屏/, 'Portrait frame.'],
+  [/中英字幕/, 'On-screen subtitles in Chinese and English.'],
+  [/必须的附件/, 'The required attachment is part of this film.'],
+];
+
+export function filmLanguage(text) {
+  return /[\u4e00-\u9fff]/.test(String(text || '')) ? 'Chinese' : 'English';
+}
+
+export function audienceLine(text) {
+  const language = filmLanguage(text);
+  return `On-screen text and narration language: ${language}. Write the on-screen title and every spoken or written sentence in ${language}.`;
+}
+
+function enforceFlags(source, english) {
+  let out = polishEnglish(english);
+  for (const [pattern, sentence] of SOURCE_FLAGS) {
+    if (!pattern.test(source)) continue;
+    const key = sentence.replace(/\.$/, '').toLowerCase();
+    if (!out.toLowerCase().includes(key)) out = polishEnglish(`${out} ${sentence}`);
+  }
+  return out;
+}
+
+export async function modelTitle(title, source, englishBrief, fetchImpl = globalThis.fetch) {
+  const raw = String(title || '').trim();
+  const brief = String(englishBrief || '').trim();
+  if (raw && !/[\u4e00-\u9fff]/.test(raw)) return raw.slice(0, 120);
+  if ((!raw || String(source || '').trim().startsWith(raw)) && brief) {
+    const subject = brief.match(/Subject:\s*"([^"]+)"/);
+    if (subject && subject[1] && !/[\u4e00-\u9fff]/.test(subject[1])) return subject[1].replace(/[.。\s]+$/g, '').slice(0, 120);
+    const line = brief.replace(/^Use the [\s\S]+? style\.\s*/i, '').split('\n')[0].trim();
+    if (line && !/[\u4e00-\u9fff]/.test(line)) return line.slice(0, 120);
+  }
+  if (!raw) return 'Untitled';
+  const translated = await toEnglishPrompt(raw, fetchImpl);
+  return (translated || 'Untitled').slice(0, 120);
+}
+
+export function polishEnglish(text) {
+  let out = String(text || '');
+  for (const [pattern, to] of REPLACEMENTS) out = out.replace(pattern, to);
+  return out.replace(/[ \t]+/g, ' ').replace(/\s*\n\s*/g, '\n').trim();
+}
+
+export function parseMyMemory(payload) {
+  const data = typeof payload === 'string' ? JSON.parse(payload) : payload;
+  if (!data || Number(data.responseStatus) >= 400) return '';
+  const text = data.responseData && data.responseData.translatedText;
+  if (!text || /MYMEMORY WARNING|QUERY LENGTH LIMIT/i.test(text)) return '';
+  return String(text)
+    .replace(/"/g, '"')
+    .replace(/'|'/g, "'")
+    .replace(/&/g, '&')
+    .replace(/&lt;/g, '<')
+    .replace(/&gt;/g, '>')
+    .trim();
+}
+
+function chunks(text) {
+  const parts = String(text || '').split(/(?<=[。！？!?\n])/);
+  const out = [];
+  let buf = '';
+  const push = (value) => {
+    const piece = value.trim();
+    if (piece) out.push(piece);
+  };
+  for (const part of parts) {
+    const next = buf + part;
+    if (Buffer.byteLength(next, 'utf8') > 420 && buf) {
+      push(buf);
+      buf = part;
+    } else {
+      buf = next;
+    }
+  }
+  push(buf);
+  return out.flatMap((piece) => {
+    if (Buffer.byteLength(piece, 'utf8') <= 420) return [piece];
+    const chars = [...piece];
+    const slices = [];
+    let current = '';
+    for (const char of chars) {
+      if (Buffer.byteLength(current + char, 'utf8') > 420) {
+        slices.push(current);
+        current = char;
+      } else current += char;
+    }
+    if (current) slices.push(current);
+    return slices;
+  });
+}
+
+async function translatePiece(text, fetchImpl) {
+  const pieces = [];
+  for (const piece of chunks(text)) {
+    if (!/[\u4e00-\u9fff]/.test(piece)) {
+      pieces.push(piece);
+      continue;
+    }
+    const url = `https://api.mymemory.translated.net/get?q=${encodeURIComponent(piece)}&langpair=zh-CN|en`;
+    const res = await fetchImpl(url, { signal: AbortSignal.timeout(12000) });
+    if (!res.ok) throw new Error('英文提示词没有生成');
+    const english = polishEnglish(parseMyMemory(await res.json()));
+    if (!english || /[\u4e00-\u9fff]/.test(english)) throw new Error('英文提示词没有生成');
+    pieces.push(english);
+  }
+  const joined = polishEnglish(pieces.join(' '));
+  if (!joined) throw new Error('英文提示词没有生成');
+  return joined;
+}
+
+export async function toEnglishPrompt(text, fetchImpl = globalThis.fetch) {
+  const source = String(text || '').trim();
+  if (!source) return '';
+  if (!/[\u4e00-\u9fff]/.test(source)) return source;
+  const framed = source.match(/^用([\s\S]{1,40}?)讲「([^」]+)」\s*[。.]?\s*([\s\S]*)$/);
+  if (framed) {
+    const topicEn = (await translatePiece(framed[2], fetchImpl)).replace(/[.。\s]+$/g, '');
+    const tailEn = framed[3].trim() ? await translatePiece(framed[3], fetchImpl) : '';
+    return enforceFlags(source, `Use the ${framed[1].trim()} style. Subject: "${topicEn}".${tailEn ? ` ${tailEn}` : ''}`);
+  }
+  const ref = source.match(/^选\s*(ref\d+)\s*[。.]?\s*([\s\S]*)$/i);
+  if (ref) {
+    const body = ref[2].trim() ? await translatePiece(ref[2], fetchImpl) : '';
+    return enforceFlags(source, `Lock ${ref[1]}. ${body}`);
+  }
+  return enforceFlags(source, await translatePiece(source, fetchImpl));
+}

package/studio/to-english-prompt.test.mjs

@@ -0,0 +1,72 @@
+import assert from 'node:assert/strict';
+import test from 'node:test';
+import { audienceLine, filmLanguage, modelTitle, parseMyMemory, polishEnglish, toEnglishPrompt } from './to-english-prompt.mjs';
+
+test('已经是英文时不请求翻译', async () => {
+  let called = false;
+  const got = await toEnglishPrompt('Keep the sentences short.', async () => {
+    called = true;
+    return { ok: true, json: async () => ({}) };
+  });
+  assert.equal(got, 'Keep the sentences short.');
+  assert.equal(called, false);
+});
+
+test('模板句保留画风名，主题和后半句译成英文', async () => {
+  const fetchImpl = async (url) => {
+    const q = decodeURIComponent(url.split('q=')[1].split('&')[0]);
+    const translated = q.includes('古代人')
+      ? 'why everyone in antiquity wanted to be emperor'
+      : 'State the conclusion first in the firelight, then one memorable detail. No oral broadcast.';
+    return { ok: true, json: async () => ({ responseStatus: 200, responseData: { translatedText: translated } }) };
+  };
+  const got = await toEnglishPrompt(
+    '用灯火史话讲「古代人都想当皇帝」。火光里先给结论，再给一个能记住的细节，没有口播。',
+    fetchImpl,
+  );
+  assert.equal(got.includes('灯火史话'), true);
+  assert.equal(got.includes('why everyone in antiquity wanted to be emperor'), true);
+  assert.equal(got.includes('no voiceover'), true);
+  assert.equal(/[\u4e00-\u9fff]/.test(got.replace('灯火史话', '')), false);
+});
+
+test('口播被收成 voiceover', () => {
+  assert.equal(polishEnglish('There was no oral broadcast.'), 'no voiceover.');
+  assert.equal(polishEnglish('There is no word of mouth.'), 'no voiceover.');
+});
+
+test('原文里的没有口播，译文漏掉时也会补上', async () => {
+  const fetchImpl = async () => ({
+    ok: true,
+    json: async () => ({ responseStatus: 200, responseData: { translatedText: 'A short film about bees.' } }),
+  });
+  const got = await toEnglishPrompt('讲蜜蜂。没有口播。', fetchImpl);
+  assert.equal(got.includes('No voiceover.'), true);
+});
+
+test('片名从英文主题来，不再把中文原句交给模型', async () => {
+  let called = false;
+  const source = '用灯火史话讲「古代人都想当皇帝」。没有口播。';
+  const title = await modelTitle(
+    source.slice(0, 40),
+    source,
+    'Use the 灯火史话 style. Subject: "why everyone wanted to be emperor". No voiceover.',
+    async () => {
+      called = true;
+      return { ok: true, json: async () => ({}) };
+    },
+  );
+  assert.equal(title, 'why everyone wanted to be emperor');
+  assert.equal(called, false);
+});
+
+test('解析翻译结果', () => {
+  assert.equal(parseMyMemory({ responseStatus: 200, responseData: { translatedText: 'Hello & welcome' } }), 'Hello & welcome');
+  assert.equal(parseMyMemory({ responseStatus: 200, responseData: { translatedText: 'MYMEMORY WARNING: too long' } }), '');
+});
+
+test('成片语言跟着用户原文', () => {
+  assert.equal(filmLanguage('古代人都想当皇帝'), 'Chinese');
+  assert.equal(filmLanguage('airplane mode'), 'English');
+  assert.equal(audienceLine('古代人').includes('in Chinese'), true);
+});

package/README.md

@@ -1,3 +0,0 @@
-# Temporary Holding Version
-
-This version is a temporary placeholder for this package. An operational version to replace this has been submitted for review and is awaiting a staged release.