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
import { sha256Blob } from './blob-digest.js';
const stable = value => Array.isArray(value) ? value.map(stable) : value && typeof value === 'object'
? Object.fromEntries(Object.keys(value).sort().filter(k => value[k] !== undefined).map(k => [k, stable(value[k])])) : value;
export const duplicateJson = value => JSON.stringify(stable(value));
const digest = value => sha256Blob(new Blob([duplicateJson(value)]));
const LOCAL = new Set(['id', 'caseInstanceId', 'importBatchId', 'importSource', 'copyOrigin', 'savedAt', 'libraryAddedAt', 'libraryUpdatedAt', 'schemaVersion', 'textRevision', 'contentRole', 'contentTypeName']);
const ASSET_LOCAL = new Set(['id', 'assetPath', 'relativePath', 'storageMode', 'capturedAt', 'linkStatus', 'sourceLastModified', 'reviewStatus']);
const MEANING = {
title: 'title', text: 'text', mediaPrompts: 'prompts', articleDocument: 'document',
customLabels: 'labels', metadataLabels: 'labels', classification: 'labels', facetAssignments: 'labels',
timeNotes: 'notes', videoAnalyses: 'analysis', visualSetAnalyses: 'analysis', analysisBreakdown: 'analysis',
analysisCandidates: 'analysis', analysisMeta: 'analysis', url: 'source', sourceFacts: 'source', sourcePages: 'source',
primaryMediaId: 'media'
};
export function mediaIdentity(asset) {
return /^[a-f0-9]{64}$/iu.test(asset.contentHash || '') && Number.isSafeInteger(asset.byteSize) && asset.byteSize >= 0
&& asset.storageMode !== 'reference' ? `${asset.byteSize}:${asset.contentHash.toLowerCase()}` : '';
}
// A case is a duplicate only when its full image/video originals match.
// Posters and documents are supporting material, never matching evidence.
export const duplicateVisualAssets = entry => (entry.mediaAssets || []).filter(a => ['image', 'video'].includes(a.kind) && a.usage !== 'poster');
export function duplicateMediaSignature(entry) {
const assets = duplicateVisualAssets(entry);
if (!assets.length || assets.some(a => !mediaIdentity(a) || a.byteSize === 0)) return '';
return duplicateJson(assets.map(a => `${a.kind}:${mediaIdentity(a)}`).sort());
}
export function duplicateProjects(state) {
const projects = new Map();
for (const c of state.organizerState?.collections || []) for (const id of c.entryIds || []) {
if (!projects.has(id)) projects.set(id, []);
projects.get(id).push(c.id);
}
return projects;
}
export function duplicateContent(entry) {
const replacements = new Map();
(entry.mediaAssets || []).forEach((a, i) => replacements.set(a.id, `$media:${i}:${mediaIdentity(a) || 'unknown'}`));
// Canonicalize media references only; literal IDs in user text remain content.
const referenceKey = key => /^(?:asset|visual|primaryMedia|posterAsset|coverVisual|derivedFromAsset|frameAsset)Ids?$/u.test(key);
const rewrite = (value, key = '') => typeof value === 'string' ? referenceKey(key) ? replacements.get(value) || value : value
: Array.isArray(value) ? value.map(v => rewrite(v, key)) : value && typeof value === 'object'
? Object.fromEntries(Object.entries(value).map(([k, v]) => [k, rewrite(v, k)])) : value;
const fields = {};
for (const [key, value] of Object.entries(entry)) {
if (LOCAL.has(key) || key === 'mediaAssets') continue;
const section = MEANING[key] || 'other';
(fields[section] ||= {})[key] = rewrite(value, key);
}
fields.media ||= {};
fields.media.assets = (entry.mediaAssets || []).map(a => Object.fromEntries(Object.entries(a)
.filter(([key]) => !ASSET_LOCAL.has(key)).map(([key, value]) => [key, rewrite(value, key)])));
// Source observation time and classifier implementation do not alter the saved content.
if (fields.source?.sourceFacts) { fields.source.sourceFacts = { ...fields.source.sourceFacts }; delete fields.source.sourceFacts.capturedAt; }
if (fields.labels?.classification) {
fields.labels.classification = { ...fields.labels.classification };
delete fields.labels.classification.reason; delete fields.labels.classification.classifierVersion;
}
return fields;
}
export function compareDuplicateCases(a, b, projects = new Map()) {
const left = duplicateContent(a), right = duplicateContent(b);
const differences = [...new Set([...Object.keys(left), ...Object.keys(right)])]
.filter(key => duplicateJson(left[key]) !== duplicateJson(right[key]));
const leftMedia = duplicateMediaSignature(a), rightMedia = duplicateMediaSignature(b);
const mediaExact = Boolean(leftMedia && rightMedia && leftMedia === rightMedia);
const complete = Boolean(leftMedia && rightMedia) && [...(a.mediaAssets || []), ...(b.mediaAssets || [])].every(asset => Boolean(mediaIdentity(asset)));
const sameProjects = duplicateJson([...(projects.get(a.id) || [])].sort()) === duplicateJson([...(projects.get(b.id) || [])].sort());
const exact = mediaExact && complete && differences.length === 0;
return { exact, mediaExact, complete, sameProjects, differences,
category: exact ? sameProjects ? 'exact' : 'projects' : 'different' };
}
export async function duplicateRevision(entry, projects = []) {
const { contentRole: _role, contentTypeName: _name, ...stored } = entry;
return digest({ entry: stored, projects: [...projects].sort() });
}
// Yield after every case: callers decide scheduling and can pause without restarting.
export async function* scanDuplicateCases(state, { cache = new WeakMap() } = {}) {
const projects = duplicateProjects(state), profiles = new Map(), buckets = new Map();
let checked = 0, unknown = 0;
for (const entry of state.entries || []) {
let profile = cache.get(entry);
if (!profile) {
const assets = duplicateVisualAssets(entry);
profile = { signature: duplicateMediaSignature(entry),
unknown: assets.filter(a => !mediaIdentity(a) || a.byteSize === 0).length };
cache.set(entry, profile);
}
profiles.set(entry.id, profile); unknown += profile.unknown;
if (profile.signature) {
if (!buckets.has(profile.signature)) buckets.set(profile.signature, []);
buckets.get(profile.signature).push(entry.id);
}
yield { checked: ++checked, total: state.entries.length, unknown };
}
const mediaGroups = [...buckets.values()].filter(ids => ids.length > 1)
.map(ids => ({ id: [...ids].sort().join('\n'), caseIds: ids }));
return { mediaGroups, duplicateCount: mediaGroups.length, checked, unknown, profiles, projects };
}
export function chooseDuplicateKeeper(entries, rule, projectId, projects = new Map(), previous) {
if (rule === 'project') {
const found = entries.filter(e => (projects.get(e.id) || []).includes(projectId));
return { id: found.length === 1 ? found[0].id : previous || entries[0]?.id, uncertain: found.length !== 1 };
}
const field = rule === 'edited' ? 'libraryUpdatedAt' : 'libraryAddedAt';
const dated = entries.map(e => ({ id: e.id, time: Date.parse(e[field]) }));
if (dated.some(e => !Number.isFinite(e.time))) return { id: previous || entries[0]?.id, uncertain: true };
dated.sort((a, b) => rule === 'oldest' ? a.time - b.time : b.time - a.time);
const tied = dated.filter(e => e.time === dated[0].time);
return { id: tied.find(e => e.id === previous)?.id || dated[0].id, uncertain: tied.length > 1 };
}
export function duplicateOrigin(entries) {
if (entries.some(e => e.copyOrigin && entries.some(other => other.id === e.copyOrigin.caseId))) return 'copy';
const batches = new Set(entries.map(e => e.importBatchId).filter(Boolean));
if (batches.size > 1) return 'imports';
if (entries.some(e => e.agentProvenance?.references?.some(r => entries.some(other => other.id === r.caseId)))) return 'reference';
return 'unknown';
}
// Partition exact-media matches by project unless the user explicitly includes cross-project copies.
// These disjoint groups make bulk selection independent of the visible page and keep/delete conflicts impossible.
export function bulkDuplicateGroups(result, entries, acrossProjects = false) {
const groups = [];
for (const mediaGroup of result.mediaGroups) {
const locations = new Map();
for (const id of mediaGroup.caseIds) {
const key = acrossProjects ? '*' : duplicateJson([...(result.projects.get(id) || [])].sort());
if (!locations.has(key)) locations.set(key, []);
locations.get(key).push(id);
}
for (const ids of locations.values()) if (ids.length > 1) {
groups.push({ id: [...ids].sort().join('\n'), caseIds: ids });
}
}
return groups;
}