Prompt as Code | GPT Image 2 / 2.5 提示词与案例库,530+ 个案例、20+ 套工业级模板与可复用 Skills,新增 2.5 同提示词对比专区,附完整提示词与生成记录,持续更新。
JavaScript agentsai-image-generationchatgptdsh-plugingpt-image-2image-promptsprompt-as-codeprompt-engineeringskillsworkflow-automation
import { mkdirSync, readFileSync, writeFileSync } from 'node:fs';
import { dirname, join } from 'node:path';
import { fileURLToPath } from 'node:url';
const root = dirname(dirname(fileURLToPath(import.meta.url)));
const docsDir = join(root, 'docs');
const outFile = join(root, 'data', 'cases.json');
const styleLibraryFile = join(root, 'data', 'style-library.json');
const styleLibrary = JSON.parse(readFileSync(styleLibraryFile, 'utf8'));
const galleryFiles = [
{ file: 'gallery-part-1.md', part: 1 },
{ file: 'gallery-part-2.md', part: 2 }
];
const categoryLabels = Object.fromEntries(
styleLibrary.categories.map((category) => [category.anchor, category.value])
);
const featuredIds = new Set([
1, 2, 6, 17, 166, 310, 330, 334, 338, 341, 344, 346, 350, 353, 354, 359, 360,
361, 362, 365, 370, 373, 375, 376, 377, 378
]);
function cleanText(value = '') {
return value
.replace(/\\_/g, '_')
.replace(/\r/g, '')
.replace(/[ \t]+\n/g, '\n')
.trim();
}
function stripMarkdown(value = '') {
return cleanText(value)
.replace(/\*\*/g, '')
.replace(/`/g, '')
.replace(/\[(.*?)\]\((.*?)\)/g, '$1')
.trim();
}
function parseCategoryMap() {
const text = readFileSync(join(docsDir, 'gallery.md'), 'utf8');
const map = new Map();
const sections = text.split(/<a name="(cat-[^"]+)"><\/a>/g);
for (let i = 1; i < sections.length; i += 2) {
const categoryId = sections[i];
const body = sections[i + 1] || '';
const category = categoryLabels[categoryId] || 'Other Use Cases';
for (const match of body.matchAll(/#case-(\d+)\)/g)) {
map.set(Number(match[1]), category);
}
}
return map;
}
function extractPrompt(block) {
const normalized = block.replace(/\r/g, '');
const match = normalized.match(/\*\*提示词:\*\*[\s\S]*?```(?:text)?\n([\s\S]*?)```/);
return cleanText(match?.[1] || '');
}
function extractSource(block) {
const line = block.match(/\*\*来源:\*\*\s*([^\n]+)/)?.[1] || '';
const link = line.match(/\[([^\]]+)\]\(([^)]+)\)/);
if (link) {
return {
label: stripMarkdown(link[1]),
url: link[2]
};
}
return {
label: stripMarkdown(line) || 'Community',
url: ''
};
}
function inferCategory(caseItem) {
if (caseItem.category) return caseItem.category;
const text = `${caseItem.title} ${caseItem.prompt}`.toLowerCase();
const rules = [
['UI & Interfaces', ['ui', 'app', 'interface', 'dashboard', 'screenshot', '网页', '界面', '截图']],
['Charts & Infographics', ['infographic', 'diagram', 'chart', 'atlas', '图谱', '信息图', '图解']],
['Posters & Typography', ['poster', 'cover', 'typography', '海报', '封面', '字体']],
['Products & E-commerce', ['product', 'packaging', 'e-commerce', '商品', '电商', '包装']],
['Brand & Logos', ['logo', 'brand', 'identity', '品牌', '标志']],
['Architecture & Spaces', ['architecture', 'interior', 'map', '建筑', '室内', '地图']],
['Photography & Realism', ['photo', 'portrait', 'camera', 'realistic', '写真', '摄影', '写实']],
['Illustration & Art', ['illustration', 'painting', 'watercolor', '插画', '艺术', '水墨']],
['Characters & People', ['character', 'pose', 'avatar', '角色', '人物', '头像']],
['Scenes & Storytelling', ['storyboard', 'scene', 'narrative', '场景', '叙事', '分镜']],
['History & Classical Themes', ['history', 'dynasty', 'classical', '历史', '古风', '唐朝', '宋']],
['Documents & Publishing', ['document', 'manual', 'prescription', '文档', '手册', '处方']]
];
return rules.find(([, keys]) => keys.some((key) => text.includes(key)))?.[0] || 'Other Use Cases';
}
function inferTags(caseItem) {
const text = `${caseItem.title} ${caseItem.prompt}`.toLowerCase();
const styleOrder = ['UI', 'Infographic', 'Poster', 'Realistic', 'Illustration', 'Product', 'Brand', 'Character', 'Classical', '3D'];
const sceneOrder = ['Tech', 'Commerce', 'Education', 'Social', 'Fashion', 'Food', 'Travel', 'Story', 'History', 'Creative'];
const styleByValue = new Map(styleLibrary.styles.map((style) => [style.value, style]));
const sceneByValue = new Map(styleLibrary.scenes.map((scene) => [scene.value, scene]));
const styleRules = styleOrder.map((value) => [
value,
(styleByValue.get(value)?.keywords || []).map((key) => key.toLowerCase())
]);
const sceneRules = sceneOrder.map((value) => [
value,
(sceneByValue.get(value)?.keywords || []).map((key) => key.toLowerCase())
]);
const pick = (rules, fallback) => {
const tags = rules
.filter(([, keys]) => keys.some((key) => text.includes(key)))
.map(([label]) => label);
return tags.length ? tags.slice(0, 3) : [fallback];
};
return {
styles: pick(styleRules, caseItem.category.split(' & ')[0].replace('Posters', 'Poster')),
scenes: pick(sceneRules, 'Creative')
};
}
function parseCases() {
const categoryMap = parseCategoryMap();
const cases = [];
for (const { file, part } of galleryFiles) {
const text = readFileSync(join(docsDir, file), 'utf8');
const chunks = text.split(/<a name="case-(\d+)"><\/a>/g);
for (let i = 1; i < chunks.length; i += 2) {
const id = Number(chunks[i]);
const block = chunks[i + 1] || '';
const title = stripMarkdown(block.match(/###\s*例\s*\d+:([^\n]+)/)?.[1] || `Case ${id}`);
const imageMatch = block.match(/!\[([^\]]*)\]\(([^)]+)\)/);
const prompt = extractPrompt(block);
const source = extractSource(block);
const category = inferCategory({
title,
prompt,
category: categoryMap.get(id)
});
const image = imageMatch?.[2]
? imageMatch[2].replace('../data/', '/')
: `/images/case${id}.jpg`;
const tags = inferTags({ title, prompt, category });
cases.push({
id,
title,
image,
imageAlt: stripMarkdown(imageMatch?.[1] || title),
sourceLabel: source.label,
sourceUrl: source.url,
prompt,
promptPreview: prompt.replace(/\n+/g, ' ').slice(0, 220),
category,
styles: tags.styles,
scenes: tags.scenes,
featured: featuredIds.has(id),
githubUrl: `https://github.com/freestylefly/awesome-gpt-image-2/blob/main/docs/gallery-part-${part}.md#case-${id}`
});
}
}
return cases.sort((a, b) => b.id - a.id);
}
const cases = parseCases();
const categories = [...new Set(cases.map((item) => item.category))].sort();
const styles = [...new Set(cases.flatMap((item) => item.styles))].sort();
const scenes = [...new Set(cases.flatMap((item) => item.scenes))].sort();
const payload = {
repository: 'https://github.com/freestylefly/awesome-gpt-image-2',
totalCases: cases.length,
categories,
styles,
scenes,
cases
};
mkdirSync(dirname(outFile), { recursive: true });
writeFileSync(outFile, `${JSON.stringify(payload, null, 2)}\n`);
console.log(`Generated ${cases.length} cases at ${outFile}`);