fetch-meta.js 8.3 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275
  1. const models = require('../../models');
  2. const fetch = require('node-fetch');
  3. let apiKey = require('process').env.ARK_API_KEY;
  4. const url = "https://ark.cn-beijing.volces.com/api/v3/chat/completions";
  5. let headers = {
  6. 'Authorization': `Bearer ${apiKey}`,
  7. 'Content-Type': 'application/json'
  8. }
  9. /**
  10. * 豆包视觉API
  11. * 从豆包获取图片的标题和文案描述
  12. * @param {*} imageUrl
  13. */
  14. async function fetchMetaByImageFromDoubao(imageUrl) {
  15. let data = {
  16. "model": "ep-20250206115552-7qg5c", // Doubao-1.5...ion-pro-32k 当前最新,贵,响应慢,效果好
  17. // "model": "ep-20250204231910-4phb8", // Doubao-vision-lite-32k 便宜点,相应速度快
  18. "messages": [
  19. {
  20. "role": "user", "content": [
  21. {
  22. "type": "text",
  23. "text": "根据图片生成标题(title, 60个字符以内)、简要描述(description, 用于网站SEO的meta description,160个字符以内,尽量包含Coloring Page关键字)、文案描述(copy, 200字左右,尽量包含Coloring Page 关键字)。 以json格式输出,支持语言中文(zh)、英语(en)、西班牙语(es)、葡萄牙语(pt)、日语(ja),形如: { title: {zh:'', en:'', es: '', pt: '', ja: ''}, description: {zh:'', en:'', es: '', pt: '', ja: ''}, copy:{zh:'', en:'', es: '', pt: '', ja: ''} }"
  24. },
  25. {
  26. "type": "image_url",
  27. "image_url": {
  28. "url": `${imageUrl}`
  29. }
  30. }
  31. ]
  32. }
  33. ]
  34. }
  35. console.log(data);
  36. const jsonData = JSON.stringify(data);
  37. const response = await fetch(url, { method: 'POST', headers, body: jsonData });
  38. let responseJson = await response.json();
  39. console.log(responseJson);
  40. return responseJson.choices[0].message.content;
  41. }
  42. /**
  43. * 豆包对话API
  44. * @param {*} text
  45. */
  46. async function fetchMetaByTxtFromDoubao(text) {
  47. let data = {
  48. "model": "ep-20250206115552-7qg5c", // Doubao-1.5...ion-pro-32k 当前最新,贵,响应慢,效果好
  49. // "model": "ep-20250204231910-4phb8", // Doubao-vision-lite-32k 便宜点,相应速度快
  50. "messages": [
  51. { "role": "user", "content": text }
  52. ]
  53. }
  54. console.log(data);
  55. const jsonData = JSON.stringify(data);
  56. const response = await fetch(url, { method: 'POST', headers, body: jsonData });
  57. let responseJson = await response.json();
  58. console.log(responseJson);
  59. return responseJson.choices[0].message.content;
  60. }
  61. async function runAlbumMeta() {
  62. let done = 0;
  63. let duration = 0;
  64. let hour, minute, second;
  65. let start = Date.now();
  66. // 筛选出所有album
  67. let query = { pid: 'art', enabled: true, $or: [{ seoTitle: { $exists: false } }, { seoTitle: null }] }
  68. let docs = await models.ArtAlbum
  69. .find(query)
  70. .sort({ order: 'asc' })
  71. .populate('title')
  72. .select('tag title slogon');
  73. let total = docs.length;
  74. console.log('total:', total);
  75. if (total <= 0) return;
  76. for (let doc of docs) {
  77. console.log(`process album ${doc._id}`);
  78. console.time(doc._id);
  79. try {
  80. let text = `帮我生成填色页专辑《${doc.title.zh}》的SEO title(60个字符以内) 和 description(160个字符以内), 以json格式输出,支持语言中文(zh)、英语(en)、西班牙语(es)、葡萄牙语(pt)、日语(ja),形如: { title: {zh:'', en:'', es: '', pt: '', ja: ''}, description: {zh:'', en:'', es: '', pt: '', ja: ''}`;
  81. let metaInfo = await fetchMetaByTxtFromDoubao(text);
  82. console.log(metaInfo);
  83. let metaInfoJson = JSON.parse(metaInfo);
  84. let titleJson = metaInfoJson.title;
  85. let descJson = metaInfoJson.description;
  86. let title = JSON.stringify(titleJson);
  87. let desc = JSON.stringify(descJson);
  88. doc.seoTitle = title;
  89. doc.seoDescription = desc;
  90. await doc.save();
  91. } catch (e) {
  92. console.error(e.message);
  93. }
  94. console.timeEnd(doc._id);
  95. done++;
  96. duration = (Date.now() - start) / 1000;
  97. hour = (Math.floor(duration / 60 / 60)).toString().padStart(2, '0');
  98. minute = (Math.floor(duration / 60) % 60).toString().padStart(2, '0');
  99. second = (Math.floor(duration) % 60).toString().padStart(2, '0');
  100. console.log('progress: ' + Math.floor((100 * done / total)) + '% used time: ' + hour + ':' + minute + ':' + second);
  101. }
  102. }
  103. async function runVideoStoryMeta() {
  104. let done = 0;
  105. let duration = 0;
  106. let hour, minute, second;
  107. let start = Date.now();
  108. // 筛选出所有album
  109. let query = { enabled: true, $or: [{ seoTitle: { $exists: false } }, { seoTitle: null }] }
  110. let docs = await models.ArtVideoStory
  111. .find(query)
  112. .sort({ order: 'asc' })
  113. .select('name');
  114. let total = docs.length;
  115. console.log('total:', total);
  116. if (total <= 0) return;
  117. for (let doc of docs) {
  118. console.log(`process video-story ${doc._id}`);
  119. console.time(doc._id);
  120. try {
  121. let text = `帮我生成填色页视频故事《${doc.name}》的SEO title(60个字符以内) 和 description(160个字符以内), 以json格式输出,支持语言中文(zh)、英语(en)、西班牙语(es)、葡萄牙语(pt)、日语(ja),形如: { title: {zh:'', en:'', es: '', pt: '', ja: ''}, description: {zh:'', en:'', es: '', pt: '', ja: ''}`;
  122. let metaInfo = await fetchMetaByTxtFromDoubao(text);
  123. console.log(metaInfo);
  124. let metaInfoJson = JSON.parse(metaInfo);
  125. let titleJson = metaInfoJson.title;
  126. let descJson = metaInfoJson.description;
  127. let title = JSON.stringify(titleJson);
  128. let desc = JSON.stringify(descJson);
  129. doc.seoTitle = title;
  130. doc.seoDescription = desc;
  131. await doc.save();
  132. } catch (e) {
  133. console.error(e.message);
  134. }
  135. console.timeEnd(doc._id);
  136. done++;
  137. duration = (Date.now() - start) / 1000;
  138. hour = (Math.floor(duration / 60 / 60)).toString().padStart(2, '0');
  139. minute = (Math.floor(duration / 60) % 60).toString().padStart(2, '0');
  140. second = (Math.floor(duration) % 60).toString().padStart(2, '0');
  141. console.log('progress: ' + Math.floor((100 * done / total)) + '% used time: ' + hour + ':' + minute + ':' + second);
  142. }
  143. }
  144. async function runArtMeta() {
  145. let done = 0;
  146. let duration = 0;
  147. let hour, minute, second;
  148. let start = Date.now();
  149. // 筛选出所有已经ready并且还没有title的图
  150. let query = { status: { $gte: 7000 }, open: true, $or: [{ copy: { $exists: false } }, { copy: null }] };
  151. let docs = await models.Art.find(query).sort({ publishTime: 'desc' }); // 内存有限,每次跑1000个
  152. let total = docs.length;
  153. console.log('total:', total);
  154. if (total <= 0) return;
  155. for (let doc of docs) {
  156. console.log(`process ${doc._id}`);
  157. let thumbUrl = `http://color.jccytech.cn/thumbs/v2/work/320/${doc._id}.png`;
  158. if (doc.hasSpecial) {
  159. thumbUrl = `http://color.jccytech.cn/thumbs/v2/special/320/${doc._id}.png`;
  160. }
  161. console.time(doc._id);
  162. try {
  163. let metaInfo = await fetchMetaByImageFromDoubao(thumbUrl);
  164. console.log(metaInfo);
  165. let metaInfoJson = JSON.parse(metaInfo);
  166. let titleJson = metaInfoJson.title;
  167. let descJson = metaInfoJson.description;
  168. let copyJson = metaInfoJson.copy;
  169. let title = JSON.stringify(titleJson);
  170. let desc = JSON.stringify(descJson);
  171. let copy = JSON.stringify(copyJson);
  172. doc.title = title;
  173. doc.desc = desc;
  174. doc.copy = copy;
  175. await doc.save();
  176. } catch (e) {
  177. console.error(e.message);
  178. }
  179. console.timeEnd(doc._id);
  180. done++;
  181. duration = (Date.now() - start) / 1000;
  182. hour = (Math.floor(duration / 60 / 60)).toString().padStart(2, '0');
  183. minute = (Math.floor(duration / 60) % 60).toString().padStart(2, '0');
  184. second = (Math.floor(duration) % 60).toString().padStart(2, '0');
  185. console.log('progress: ' + Math.floor((100 * done / total)) + '% used time: ' + hour + ':' + minute + ':' + second);
  186. }
  187. runArtMeta();
  188. }
  189. async function run() {
  190. await runAlbumMeta();
  191. await runVideoStoryMeta();
  192. await runArtMeta();
  193. }
  194. async function test() {
  195. let metaInfo = await fetchMetaByImageFromDoubao("https://color.jccytech.cn/thumbs/v2/work/640/67a254ec4f9d65537938e5c5.png");
  196. console.log(metaInfo);
  197. let metaInfoJson = JSON.parse(metaInfo);
  198. let titleJson = metaInfoJson.title;
  199. let descJson = metaInfoJson.copy;
  200. let title = JSON.stringify(titleJson);
  201. let desc = JSON.stringify(descJson);
  202. console.log(title);
  203. console.log(desc);
  204. }
  205. module.exports = { run }
  206. if (require.main == module) {
  207. run();
  208. }