Spot it. Save it. Free, local-first Chrome extension for images, videos and prompts. Explore 2,000+ curated cases and connect your library to AI Agents via MCP. 提示词导演
JavaScript ai-artai-videochrome-extensionclaude-codecodexhiggsfieldjimenglocal-firstmcpmidjourneymoodboardprompt-engineeringprompt-libraryprompt-managerreference-libraryseedancevisual-bookmarksweb-clipper
// Keep images in the original DOCX. Markdown stores stable document-local
// references so indexing and Agent reads never include base64 image payloads.
export async function convertDocx(blob, { includeImages = false } = {}) {
await import("./vendor/document-ingestion/mammoth.browser.js");
const mammoth = globalThis.mammoth;
if (!mammoth?.convertToHtml) throw new Error("DOCX 解析组件未加载");
const images = new Map();
let imageIndex = 0;
const result = await mammoth.convertToHtml({ arrayBuffer: await blob.arrayBuffer() }, {
externalFileAccess: false,
styleMap: ["p[style-name='Title'] => h1:fresh", "p[style-name='Subtitle'] => p:fresh"],
convertImage: mammoth.images.imgElement(async image => {
const reference = `document-image:${imageIndex++}`;
if (includeImages && /^image\/(?:png|jpeg|gif|webp|avif)$/u.test(image.contentType)) {
images.set(reference, `data:${image.contentType};base64,${await image.read("base64")}`);
}
return { src: reference };
})
});
const warnings = result.messages.map(item => /^Unrecognised paragraph style:/u.test(item.message)
? "部分段落样式未完全还原,文字已保留,请对照原文档查看排版"
: `文档解析提示:${item.message}`);
return { html: result.value, images, warnings: [...new Set(warnings)] };
}
export function docxImageLoader(blob) {
let pending;
return async reference => {
pending ??= convertDocx(blob, { includeImages: true });
const image = (await pending).images.get(reference);
if (!image) throw new Error("此内嵌图片暂不能预览,请下载原文档查看");
return image;
};
}