Expand seller store functionality

2026-09-07 17:51:11 +08:00
parent 4c0f3a4a79
commit ceb644a3f9
35 changed files with 32883 additions and 364 deletions
@@ -2,13 +2,11 @@ import type { CollectedProductPayload } from '@atlas/contracts';

 const apiOrigin = (import.meta.env.ATLAS_API_ORIGIN as string | undefined) ?? 'http://localhost:8200';
 const webOrigin = (import.meta.env.ATLAS_WEB_ORIGIN as string | undefined) ?? 'http://localhost:8100';
-const clientId = (import.meta.env.ATLAS_EXTENSION_CLIENT_ID as string | undefined) ?? 'atlas-collector';
-type Tokens = { access_token: string; expires_in: number; refresh_token: string; refresh_token_expires_in: number; token_type: 'Bearer' };
 type Session = { authenticated: true; workspace: { id: string; available: boolean }; can_collect: boolean };
-const storageKey = 'atlas.extension.tokens.v1';
+const storageKey = 'atlas.extension.web-token.v1';

-async function loadTokens() { return (await browser.storage.local.get(storageKey))[storageKey] as Tokens | undefined; }
-async function saveTokens(tokens: Tokens) { await browser.storage.local.set({ [storageKey]: tokens }); }
+async function loadToken() { return (await browser.storage.local.get(storageKey))[storageKey] as string | undefined; }
+async function saveToken(token: string) { await browser.storage.local.set({ [storageKey]: token }); }
 async function clearTokens() { await browser.storage.local.remove(storageKey); }
 async function request(path: string, options: RequestInit = {}, token?: string) {
    const response = await fetch(`${apiOrigin}/api${path}`, { ...options, headers: { Accept: 'application/json', ...(options.headers ?? {}), ...(token ? { Authorization: `Bearer ${token}` } : {}) } });
@@ -16,24 +14,70 @@ async function request(path: string, options: RequestInit = {}, token?: string)
    return response.status === 204 ? undefined : response.json() as Promise<{ data: unknown }>;
 }
 async function activeAccessToken() {
-    let tokens = await loadTokens(); if (!tokens) return null;
-    try { await request('/extension/session', {}, tokens.access_token); return tokens.access_token; }
-    catch (error) {
-        if ((error as { status?: number }).status !== 401) throw error;
-        try { const refreshed = await request('/extension/auth/refresh', { method: 'POST', headers: { 'Content-Type': 'application/json' }, body: JSON.stringify({ refresh_token: tokens.refresh_token }) }) as { data: Tokens }; tokens = refreshed.data; await saveTokens(tokens); return tokens.access_token; }
-        catch { await clearTokens(); return null; }
-    }
+    const cached = await loadToken();
+    const usableCached = await validateAccessToken(cached);
+    if (usableCached) return usableCached;
+
+    // 用户可能先在 Atlas Web 登录、再打开来源商品页。此时插件还没有自己的
+    // 缓存，仍应直接从已打开的 Atlas 标签页同步，而不是错误地展示“登录”。
+    const synchronized = await syncAccessTokenFromWebTabs();
+    return validateAccessToken(synchronized);
 }
 export async function getSession(): Promise<Session | null> { const token = await activeAccessToken(); if (!token) return null; return (await request('/extension/session', {}, token) as { data: Session }).data; }
 export async function login() {
-    const verifier = crypto.getRandomValues(new Uint8Array(64)); const verifierText = base64url(verifier); const challenge = base64url(new Uint8Array(await crypto.subtle.digest('SHA-256', new TextEncoder().encode(verifierText))));
-    const state = base64url(crypto.getRandomValues(new Uint8Array(32))); const redirectUri = browser.identity.getRedirectURL('callback');
-    const authorize = new URL(`${apiOrigin}/api/extension/authorize`); authorize.search = new URLSearchParams({ client_id: clientId, redirect_uri: redirectUri, state, code_challenge: challenge, code_challenge_method: 'S256' }).toString();
-    const callback = await browser.identity.launchWebAuthFlow({ url: authorize.toString(), interactive: true });
-    if (!callback) throw new Error('未完成 Atlas ERP 登录'); const result = new URL(callback); if (result.searchParams.get('state') !== state || !result.searchParams.get('code')) throw new Error('扩展授权状态校验失败');
-    const exchanged = await request('/extension/auth/exchange', { method: 'POST', headers: { 'Content-Type': 'application/json' }, body: JSON.stringify({ client_id: clientId, code: result.searchParams.get('code'), redirect_uri: redirectUri, code_verifier: verifierText }) }) as { data: Tokens };
-    await saveTokens(exchanged.data); return getSession();
+    const currentSession = await getSession();
+    if (currentSession) return currentSession;
+
+    const tab = await browser.tabs.create({ url: webOrigin, active: true });
+    if (tab.id === undefined) throw new Error('无法打开 Atlas ERP 登录页');
+    const deadline = Date.now() + 3 * 60 * 1_000;
+    while (Date.now() < deadline) {
+        const token = await readWebAccessToken(tab.id);
+        if (token) {
+            await saveToken(token);
+            const session = await getSession();
+            if (session) return session;
+            await clearTokens();
+        }
+        await delay(2_000);
+    }
+    throw new Error('登录超时，请在 Atlas ERP 网页完成登录后重试');
 }
 export async function submit(payload: CollectedProductPayload) { const token = await activeAccessToken(); if (!token) throw new Error('请先登录 Atlas ERP'); return request('/extension/materials', { method: 'POST', headers: { 'Content-Type': 'application/json', 'Idempotency-Key': crypto.randomUUID() }, body: JSON.stringify(payload) }, token) as Promise<{ data: { editor_entry_path: string; product_id: string; collection_item: { id: string }; reused: boolean } }>; }
 export async function openErp(path = '/') { await browser.tabs.create({ url: new URL(path, webOrigin).toString() }); }
-function base64url(bytes: Uint8Array) { let value = ''; for (const byte of bytes) value += String.fromCharCode(byte); return btoa(value).replace(/\+/g, '-').replace(/\//g, '_').replace(/=+$/, ''); }
+async function readWebAccessToken(tabId: number) {
+    try {
+        const result = await browser.scripting.executeScript({ target: { tabId }, func: () => localStorage.getItem('atlas.extension.web-access') });
+        const raw = result[0]?.result;
+        if (typeof raw !== 'string') return null;
+        const parsed = JSON.parse(raw) as { accessToken?: unknown };
+        return typeof parsed.accessToken === 'string' && parsed.accessToken.length > 0 ? parsed.accessToken : null;
+    } catch { return null; }
+}
+async function syncAccessTokenFromWebTabs() {
+    const tabs = await browser.tabs.query({ url: `${webOrigin}/*` });
+    // tabs.query 返回的顺序未承诺。倒序优先尝试最近的标签页（Chrome 通常最后返回它）。
+    for (const tab of tabs.reverse()) {
+        if (tab.id === undefined) continue;
+        const token = await readWebAccessToken(tab.id);
+        if (token) {
+            await saveToken(token);
+            return token;
+        }
+    }
+    return null;
+}
+async function validateAccessToken(token: string | undefined | null) {
+    if (!token) return null;
+    try {
+        await request('/extension/session', {}, token);
+        return token;
+    } catch (error) {
+        if ((error as { status?: number }).status === 401) {
+            await clearTokens();
+            return null;
+        }
+        throw error;
+    }
+}
+function delay(milliseconds: number) { return new Promise<void>((resolve) => setTimeout(resolve, milliseconds)); }
@@ -0,0 +1,336 @@
+import type { CollectedProductPayload } from '@atlas/contracts';
+
+type SupportedMediaPlatform = 'ozon' | '1688' | 'taobao' | 'tmall';
+type DomMediaPlatform = Exclude<SupportedMediaPlatform, 'ozon'>;
+type Asset = CollectedProductPayload['assets'][number];
+
+type ImageGroup = {
+    role: Asset['role'];
+    relation: Asset['relation'];
+    selectors: string[];
+    sourceProps?: Array<
+        | 'data-lazyload-src'
+        | 'data-src'
+        | 'srcset'
+        | 'currentSrc'
+        | 'src'
+        | 'backgroundImage'
+    >;
+    excludeWithin?: string[];
+    minWidth?: number;
+    minHeight?: number;
+};
+
+/**
+ * These profiles are ported from the proven collector extension.  The crucial
+ * invariant is that images always come from a product widget, never from a
+ * page-wide `document.images` scan (which also includes logos, ads and
+ * recommendation cards).
+ */
+const DOM_IMAGE_PROFILES: Record<DomMediaPlatform, ImageGroup[]> = {
+    '1688': [
+        {
+            role: 'source_main',
+            relation: 'inherited_product_default',
+            selectors: [
+                '[data-module="od_picture_gallery"] img',
+                '.module-od-picture-gallery img',
+                '#recyclerview .detail-gallery-turn-wrapper .detail-gallery-img',
+                '#screen .od-gallery-turn-item-wrapper .od-gallery-img',
+                '#content .od-scroller-item .v-image-cover',
+                '#content .od-picture-gallery-list .v-image-cover',
+                '#dt-tab img',
+                '.detail-gallery-turn img.detail-gallery-img',
+                '.img-list-wrapper img.od-gallery-img',
+            ],
+            minWidth: 200,
+            minHeight: 200,
+        },
+        {
+            role: 'source_sku',
+            relation: 'axis_value',
+            selectors: [
+                '[data-module="od_sku_selection"] img',
+                '.module-od-sku-selection img',
+                '.pc-sku-wrapper .prop-item-inner-wrapper',
+                '.sku-item-wrapper',
+                '.specification-cell',
+                '.sku-filter-button',
+                '.expand-view-item',
+                '.feature-item img',
+            ],
+            sourceProps: ['backgroundImage'],
+            minWidth: 20,
+            minHeight: 20,
+        },
+        {
+            role: 'source_detail',
+            relation: 'product_candidate',
+            selectors: [
+                '#detail img',
+                '.de-description-detail img',
+                '#detailContentContainer img',
+                '.html-description img',
+            ],
+            minWidth: 300,
+            minHeight: 100,
+        },
+    ],
+    taobao: [
+        {
+            role: 'source_main',
+            relation: 'inherited_product_default',
+            selectors: [
+                '[class*="picGallery--"] img',
+                '#picGalleryEle img',
+                '[class*="thumbnailPic--"]',
+            ],
+        },
+        {
+            role: 'source_sku',
+            relation: 'axis_value',
+            selectors: [
+                '[class*="valueItem--"]',
+                '[class*="valueItemImgWrap--"]',
+            ],
+            minWidth: 20,
+            minHeight: 20,
+        },
+        taobaoDetailProfile(),
+    ],
+    tmall: [
+        {
+            role: 'source_main',
+            relation: 'inherited_product_default',
+            selectors: [
+                '[class*="picGallery--"] img',
+                '#picGalleryEle img',
+                '[class*="thumbnailPic--"]',
+            ],
+        },
+        {
+            role: 'source_sku',
+            relation: 'axis_value',
+            selectors: [
+                '[class*="valueItem--"]',
+                '[class*="valueItemImgWrap--"]',
+            ],
+            minWidth: 20,
+            minHeight: 20,
+        },
+        taobaoDetailProfile(),
+    ],
+};
+
+function taobaoDetailProfile(): ImageGroup {
+    return {
+        role: 'source_detail',
+        relation: 'product_candidate',
+        selectors: [
+            '[class*="tabDetailWrap--"] img',
+            '[class*="detailInfo--"] img',
+        ],
+        // The old collector verified that review photos and recommendation
+        // cards can be large enough to pass a size-only filter.
+        excludeWithin: [
+            '[class*="Comment--"]',
+            '[class*="comments--"]',
+            '[class*="userInfo--"]',
+            '[class*="rate"]',
+            '[class*="RecommendInfo--"]',
+            '[data-spm="recommends"]',
+            '[class*="recommend--"]',
+            '[class*="cardPic--"]',
+        ],
+        minWidth: 300,
+        minHeight: 100,
+    };
+}
+
+export function collectPlatformDomAssets(
+    platform: DomMediaPlatform,
+    root: Document,
+    pageUrl: string,
+): Asset[] {
+    const assets: Asset[] = [];
+    const seen = new Set<string>();
+    for (const group of DOM_IMAGE_PROFILES[platform]) {
+        for (const selector of group.selectors) {
+            for (const element of Array.from(
+                root.querySelectorAll<HTMLElement>(selector),
+            )) {
+                if (
+                    group.excludeWithin?.some((excluded) =>
+                        element.closest(excluded),
+                    )
+                )
+                    continue;
+                if (!passesDimensions(element, group)) continue;
+                const url = imageUrlFromElement(
+                    element,
+                    group.sourceProps,
+                    pageUrl,
+                    platform,
+                );
+                if (!url) continue;
+                // Keep the old collector's group-level de-duplication: an
+                // aspect cover may legitimately also be a gallery image.
+                const key = `${group.role}:${dedupeImageUrl(url, platform)}`;
+                if (seen.has(key)) continue;
+                seen.add(key);
+                assets.push({
+                    url,
+                    role: group.role,
+                    relation: group.relation,
+                    sortOrder: assets.length,
+                    evidenceRef: `${platform}:dom:${group.role}`,
+                });
+            }
+        }
+    }
+    return assets;
+}
+
+/** Restores the original CDN URL before payload storage and de-duplication. */
+export function normaliseImageUrl(
+    raw: string,
+    pageUrl: string,
+    platform: SupportedMediaPlatform,
+): string | undefined {
+    let url: URL;
+    try {
+        url = new URL(raw, pageUrl);
+    } catch {
+        return undefined;
+    }
+    if (url.protocol !== 'https:' && url.protocol !== 'http:') return undefined;
+
+    if (platform === 'ozon') {
+        // Ozon thumbnails use /wc50/, /wc100/, /c600/, etc.  The same media
+        // without this segment is its original CDN resource.
+        url.pathname = url.pathname
+            .replace(/\/(?:wc|c)\d+\//i, '/')
+            .replace(/\/{2,}/g, '/');
+    } else {
+        // Alibaba image URLs commonly append a transform after the file
+        // extension (e.g. image.jpg_400x400.jpg).  Keep the original image.
+        const original = url.pathname.match(
+            /^(.+?\.(?:jpe?g|png|gif|bmp|heic|webp|avif))_/i,
+        );
+        if (original) url.pathname = original[1];
+    }
+    for (const parameter of ['width', 'height', 'size', 'quality', 'w', 'h']) {
+        url.searchParams.delete(parameter);
+    }
+    return url.toString();
+}
+
+export function dedupeImageUrl(url: string, platform: SupportedMediaPlatform) {
+    const normalised = normaliseImageUrl(url, url, platform) ?? url;
+    try {
+        const parsed = new URL(normalised);
+        parsed.search = '';
+        parsed.hash = '';
+        return parsed.toString();
+    } catch {
+        return normalised;
+    }
+}
+
+function imageUrlFromElement(
+    element: HTMLElement,
+    sourceProps: ImageGroup['sourceProps'] = [
+        'data-lazyload-src',
+        'data-src',
+        'srcset',
+        'currentSrc',
+        'src',
+        'backgroundImage',
+    ],
+    pageUrl: string,
+    platform: SupportedMediaPlatform,
+) {
+    const nestedTarget = element.matches('img, source')
+        ? null
+        : element.querySelector<HTMLElement>('img, source');
+    const imageTarget = element.matches('img, source') ? element : nestedTarget;
+    for (const prop of sourceProps) {
+        // 1688 uses background images on the selected SKU option container.
+        // For every other property, prefer the actual nested image/source.
+        const target = prop === 'backgroundImage' ? element : imageTarget;
+        if (!target) continue;
+        let raw = '';
+        if (prop === 'srcset')
+            raw = bestSrcset(target.getAttribute('srcset') ?? '');
+        else if (prop === 'currentSrc')
+            raw = (target as HTMLImageElement).currentSrc ?? '';
+        else if (prop === 'backgroundImage') {
+            const style =
+                target.ownerDocument.defaultView?.getComputedStyle(target)
+                    .backgroundImage ?? target.style.backgroundImage;
+            raw = style.match(/url\(["']?(.*?)["']?\)/i)?.[1] ?? '';
+            // Some 1688 SKU selectors directly match an <img> despite the
+            // group being background-image-first.
+            if (!raw && target.tagName === 'IMG') {
+                raw =
+                    (target as HTMLImageElement).currentSrc ||
+                    target.getAttribute('src') ||
+                    '';
+            }
+            if (!raw && nestedTarget && nestedTarget !== target) {
+                const nestedStyle =
+                    nestedTarget.ownerDocument.defaultView?.getComputedStyle(
+                        nestedTarget,
+                    ).backgroundImage ?? nestedTarget.style.backgroundImage;
+                raw =
+                    nestedStyle.match(/url\(["']?(.*?)["']?\)/i)?.[1] ??
+                    ((nestedTarget as HTMLImageElement).currentSrc ||
+                        nestedTarget.getAttribute('src') ||
+                        '');
+            }
+        } else raw = target.getAttribute(prop) ?? '';
+        const url = raw && normaliseImageUrl(raw, pageUrl, platform);
+        if (url) return url;
+    }
+    return undefined;
+}
+
+function passesDimensions(element: HTMLElement, group: ImageGroup) {
+    if (!group.minWidth && !group.minHeight) return true;
+    const image = element.matches('img')
+        ? (element as HTMLImageElement)
+        : element.querySelector<HTMLImageElement>('img');
+    const measured = image ?? element;
+    const width = image?.naturalWidth || measured.offsetWidth || 0;
+    const height = image?.naturalHeight || measured.offsetHeight || 0;
+    // Browser snapshots and unrendered lazy images often report 0.  As in the
+    // old collector, only reject when a real measured size proves it is tiny.
+    return !(
+        width > 0 &&
+        height > 0 &&
+        (width < (group.minWidth ?? 0) || height < (group.minHeight ?? 0))
+    );
+}
+
+function bestSrcset(srcset: string) {
+    let best = '';
+    let bestSize = -1;
+    for (const part of srcset
+        .split(',')
+        .map((value) => value.trim())
+        .filter(Boolean)) {
+        const [url, descriptor = ''] = part.split(/\s+/, 2);
+        const width = descriptor.match(/^(\d+)w$/)?.[1];
+        const density = descriptor.match(/^(\d+(?:\.\d+)?)x$/)?.[1];
+        const size = width
+            ? Number(width)
+            : density
+              ? Number(density) * 1_000
+              : 0;
+        if (size >= bestSize) {
+            best = url;
+            bestSize = size;
+        }
+    }
+    return best;
+}
@@ -1,32 +1,62 @@
 import type { CollectedProductPayload } from '@atlas/contracts';
-import { collectOzonProduct, isOzonProductPage } from './ozon-parser.js';
+import {
+    collectOzonProduct,
+    isOzonProductPage,
+    type OzonPageJsonFetcher,
+} from './ozon-parser.js';
+import { collectPlatformDomAssets } from './product-media.js';

 type SourcePlatform = '1688' | 'taobao' | 'tmall' | 'ozon';
 type Json = Record<string, unknown>;
 type Axis = CollectedProductPayload['variantAxes'][number];
 type Variant = CollectedProductPayload['variants'][number];

-export function supportedProductPlatform(url = location.href): SourcePlatform | null {
+export function supportedProductPlatform(
+    url = location.href,
+): SourcePlatform | null {
    if (isOzonProductPage(url)) return 'ozon';
-    if (/^https:\/\/detail\.1688\.com\/offer\/\d+\.html/.test(url)) return '1688';
+    if (/^https:\/\/detail\.1688\.com\/offer\/\d+\.html/.test(url))
+        return '1688';
    if (/^https:\/\/item\.taobao\.com\/item\.htm/.test(url)) return 'taobao';
    if (/^https:\/\/detail\.tmall\.com\/item\.htm/.test(url)) return 'tmall';
    return null;
 }

-export function collectCurrentProduct(root: Document = document, pageUrl = location.href): CollectedProductPayload {
+export async function collectCurrentProduct(
+    root: Document = document,
+    pageUrl = location.href,
+    fetchOzonPageJson?: OzonPageJsonFetcher,
+): Promise<CollectedProductPayload> {
    const platform = supportedProductPlatform(pageUrl);
-    if (platform === 'ozon') return collectOzonProduct(root, pageUrl);
+    if (platform === 'ozon')
+        return collectOzonProduct(root, pageUrl, fetchOzonPageJson);
    if (!platform) throw new Error('当前页面不是支持的商品详情页');
-    return platform === '1688' ? collect1688(root, pageUrl) : collectTaobao(root, pageUrl, platform);
+    return platform === '1688'
+        ? collect1688(root, pageUrl)
+        : collectTaobao(root, pageUrl, platform);
 }

 function collect1688(root: Document, pageUrl: string): CollectedProductPayload {
-    const state = pageState(root); const skuMap = findFirst(state, 'skuMapOriginal'); const skuProps = findFirst(state, 'skuProps');
-    const axes = parse1688Axes(skuProps); const evidence: CollectedProductPayload['evidence'] = [{ ref: 'page:url', kind: 'page', path: 'location.href', raw_value: pageUrl }];
+    const state = pageState(root);
+    const skuMap = findFirst(state, 'skuMapOriginal');
+    const skuProps = findFirst(state, 'skuProps');
+    const axes = parse1688Axes(skuProps);
+    const evidence: CollectedProductPayload['evidence'] = [
+        {
+            ref: 'page:url',
+            kind: 'page',
+            path: 'location.href',
+            raw_value: pageUrl,
+        },
+    ];
    const variants: Variant[] = [];
    if (isRecord(skuMap)) {
-        evidence.push({ ref: '1688:sku-map', kind: 'state', path: 'skuMapOriginal', raw_value: `组合记录 ${Object.keys(skuMap).length} 条` });
+        evidence.push({
+            ref: '1688:sku-map',
+            kind: 'state',
+            path: 'skuMapOriginal',
+            raw_value: `组合记录 ${Object.keys(skuMap).length} 条`,
+        });
        for (const [combination, row] of Object.entries(skuMap)) {
            if (!isRecord(row)) continue;
            const sourceSku = stringAt(row, ['skuId', 'sku_id', 'skuCode']);
@@ -34,50 +64,391 @@ function collect1688(root: Document, pageUrl: string): CollectedProductPayload {
            // record. Never expand it with the Cartesian product of skuProps.
            const sourceVariantKey = sourceSku ?? `1688:${combination}`;
            if (!combination || combination === 'undefined') continue;
-            variants.push({ sourceVariantKey, ...(sourceSku ? { sourceSku } : {}), displayName: selectionsForCombination(combination, axes).map((item) => item.valueName).join(' / ') || `1688 SKU ${sourceSku ?? combination}`, selections: selectionsForCombination(combination, axes), ...(moneyAt(row, ['discountPrice', 'price', 'salePrice']) ? { sourcePrice: { amount: moneyAt(row, ['discountPrice', 'price', 'salePrice'])!, currencyCode: 'CNY' } } : {}), confidence: 'verified', evidenceRefs: ['1688:sku-map'] });
+            variants.push({
+                sourceVariantKey,
+                ...(sourceSku ? { sourceSku } : {}),
+                displayName:
+                    selectionsForCombination(combination, axes)
+                        .map((item) => item.valueName)
+                        .join(' / ') || `1688 SKU ${sourceSku ?? combination}`,
+                selections: selectionsForCombination(combination, axes),
+                ...(moneyAt(row, ['discountPrice', 'price', 'salePrice'])
+                    ? {
+                          sourcePrice: {
+                              amount: moneyAt(row, [
+                                  'discountPrice',
+                                  'price',
+                                  'salePrice',
+                              ])!,
+                              currencyCode: 'CNY',
+                          },
+                      }
+                    : {}),
+                confidence: 'verified',
+                evidenceRefs: ['1688:sku-map'],
+            });
        }
    }
-    return basePayload('1688', offerId(pageUrl), root, pageUrl, axes, uniqueVariants(variants), evidence, '1688-state-1.0.0');
+    return basePayload(
+        '1688',
+        offerId(pageUrl),
+        root,
+        pageUrl,
+        axes,
+        uniqueVariants(variants),
+        evidence,
+        '1688-state-1.0.0',
+    );
 }

-function collectTaobao(root: Document, pageUrl: string, platform: 'taobao' | 'tmall'): CollectedProductPayload {
-    const state = pageState(root); const skuBase = findFirst(state, 'skuBase'); const skuCore = findFirst(state, 'skuCore');
-    const axes = parseTaobaoAxes(isRecord(skuBase) ? skuBase.props : undefined); const evidence: CollectedProductPayload['evidence'] = [{ ref: 'page:url', kind: 'page', path: 'location.href', raw_value: pageUrl }];
+function collectTaobao(
+    root: Document,
+    pageUrl: string,
+    platform: 'taobao' | 'tmall',
+): CollectedProductPayload {
+    const state = pageState(root);
+    const skuBase = findFirst(state, 'skuBase');
+    const skuCore = findFirst(state, 'skuCore');
+    const axes = parseTaobaoAxes(isRecord(skuBase) ? skuBase.props : undefined);
+    const evidence: CollectedProductPayload['evidence'] = [
+        {
+            ref: 'page:url',
+            kind: 'page',
+            path: 'location.href',
+            raw_value: pageUrl,
+        },
+    ];
    const variants: Variant[] = [];
-    const skus = isRecord(skuBase) && Array.isArray(skuBase.skus) ? skuBase.skus : [];
-    const sku2info = isRecord(skuCore) && isRecord(skuCore.sku2info) ? skuCore.sku2info : {};
-    if (skus.length) evidence.push({ ref: 'taobao:sku-base', kind: 'state', path: 'skuBase.skus', raw_value: `组合记录 ${skus.length} 条` });
+    const skus =
+        isRecord(skuBase) && Array.isArray(skuBase.skus) ? skuBase.skus : [];
+    const sku2info =
+        isRecord(skuCore) && isRecord(skuCore.sku2info) ? skuCore.sku2info : {};
+    if (skus.length)
+        evidence.push({
+            ref: 'taobao:sku-base',
+            kind: 'state',
+            path: 'skuBase.skus',
+            raw_value: `组合记录 ${skus.length} 条`,
+        });
    for (const sku of skus) {
        if (!isRecord(sku)) continue;
-        const skuId = stringAt(sku, ['skuId', 'sku_id']); const propPath = stringAt(sku, ['propPath', 'prop_path']);
+        const skuId = stringAt(sku, ['skuId', 'sku_id']);
+        const propPath = stringAt(sku, ['propPath', 'prop_path']);
        // skuId + propPath is the required combination proof. Props alone only
        // become axes and cause a single default SKU on the server.
        if (!skuId || !propPath) continue;
        const info = isRecord(sku2info[skuId]) ? sku2info[skuId] : {};
        const selections = selectionsForPropPath(propPath, axes);
-        variants.push({ sourceVariantKey: skuId, sourceSku: skuId, displayName: selections.map((item) => item.valueName).join(' / ') || `SKU ${skuId}`, selections, ...(moneyAt(info, ['price', 'promotionPrice', 'discountPrice']) ? { sourcePrice: { amount: moneyAt(info, ['price', 'promotionPrice', 'discountPrice'])!, currencyCode: 'CNY' } } : {}), confidence: 'verified', evidenceRefs: ['taobao:sku-base'] });
+        variants.push({
+            sourceVariantKey: skuId,
+            sourceSku: skuId,
+            displayName:
+                selections.map((item) => item.valueName).join(' / ') ||
+                `SKU ${skuId}`,
+            selections,
+            ...(moneyAt(info, ['price', 'promotionPrice', 'discountPrice'])
+                ? {
+                      sourcePrice: {
+                          amount: moneyAt(info, [
+                              'price',
+                              'promotionPrice',
+                              'discountPrice',
+                          ])!,
+                          currencyCode: 'CNY',
+                      },
+                  }
+                : {}),
+            confidence: 'verified',
+            evidenceRefs: ['taobao:sku-base'],
+        });
    }
-    return basePayload(platform, itemId(pageUrl), root, pageUrl, axes, uniqueVariants(variants), evidence, `${platform}-state-1.0.0`);
+    return basePayload(
+        platform,
+        itemId(pageUrl),
+        root,
+        pageUrl,
+        axes,
+        uniqueVariants(variants),
+        evidence,
+        `${platform}-state-1.0.0`,
+    );
 }

-function basePayload(platform: Exclude<SourcePlatform, 'ozon'>, itemKey: string, root: Document, pageUrl: string, axes: Axis[], variants: Variant[], evidence: CollectedProductPayload['evidence'], parserVersion: string): CollectedProductPayload {
-    const canonical = root.querySelector<HTMLLinkElement>('link[rel="canonical"]')?.href ?? pageUrl.split('#')[0]; const title = root.querySelector('h1')?.textContent?.replace(/\s+/g, ' ').trim() || root.querySelector<HTMLMetaElement>('meta[property="og:title"]')?.content?.trim();
-    const assets = Array.from(root.images).map((image) => image.currentSrc || image.src).filter((url) => /^https:\/\//.test(url)).filter((url, index, all) => all.indexOf(url) === index).slice(0, 40).map((url, sortOrder) => ({ url, role: 'source_main' as const, relation: 'product_candidate' as const, sortOrder, evidenceRef: 'page:gallery' }));
-    return { source: { platform, itemKey, canonicalUrl: canonical, collectedAt: new Date().toISOString(), parserVersion }, common: { ...(title ? { title: { value: title, raw_value: title, confidence: 'verified' as const } } : {}), attributes: [], categories: [] }, variantAxes: axes, variants, assets, evidence };
+function basePayload(
+    platform: Exclude<SourcePlatform, 'ozon'>,
+    itemKey: string,
+    root: Document,
+    pageUrl: string,
+    axes: Axis[],
+    variants: Variant[],
+    evidence: CollectedProductPayload['evidence'],
+    parserVersion: string,
+): CollectedProductPayload {
+    const canonical =
+        root.querySelector<HTMLLinkElement>('link[rel="canonical"]')?.href ??
+        pageUrl.split('#')[0];
+    const title =
+        root.querySelector('h1')?.textContent?.replace(/\s+/g, ' ').trim() ||
+        root
+            .querySelector<HTMLMetaElement>('meta[property="og:title"]')
+            ?.content?.trim();
+    const assets = collectPlatformDomAssets(platform, root, canonical);
+    evidence.push({
+        ref: `${platform}:dom:image-profile`,
+        kind: 'dom',
+        path: 'platform-scoped gallery / sku / detail selectors',
+        raw_value: `采集 ${assets.length} 张受限商品区域图片`,
+    });
+    return {
+        source: {
+            platform,
+            itemKey,
+            canonicalUrl: canonical,
+            collectedAt: new Date().toISOString(),
+            parserVersion,
+        },
+        common: {
+            ...(title
+                ? {
+                      title: {
+                          value: title,
+                          raw_value: title,
+                          confidence: 'verified' as const,
+                      },
+                  }
+                : {}),
+            attributes: [],
+            categories: [],
+        },
+        variantAxes: axes,
+        variants,
+        assets,
+        evidence,
+    };
 }

-function pageState(root: Document) { const all: unknown[] = []; for (const script of Array.from(root.scripts)) { const text = script.textContent?.trim() ?? ''; if (text.length < 2 || text.length > 5_000_000) continue; if (script.type === 'application/ld+json') { try { all.push(JSON.parse(text)); } catch { /* ignored evidence */ } } for (const marker of ['skuMapOriginal', 'skuBase', 'skuCore', 'skuProps']) { const object = objectAfterMarker(text, marker); if (object) all.push(object); } } return all; }
-function objectAfterMarker(text: string, marker: string): unknown | null { const position = text.indexOf(marker); if (position < 0) return null; const start = text.slice(position).search(/[\[{]/); if (start < 0) return null; const first = position + start; const open = text[first]; const close = open === '{' ? '}' : ']'; let depth = 0; let quote = ''; let escaped = false; for (let index = first; index < text.length && index - first < 2_000_000; index += 1) { const char = text[index]; if (quote) { if (escaped) escaped = false; else if (char === '\\') escaped = true; else if (char === quote) quote = ''; continue; } if (char === '"' || char === "'") { quote = char; continue; } if (char === open) depth += 1; if (char === close && --depth === 0) { try { return JSON.parse(text.slice(first, index + 1)); } catch { return null; } } } return null; }
-function findFirst(states: unknown[], key: string): unknown { for (const state of states) { const found = findKey(state, key); if (found !== undefined) return found; } return undefined; }
-function findKey(value: unknown, key: string, seen = new Set<unknown>()): unknown { if (!value || typeof value !== 'object' || seen.has(value)) return undefined; seen.add(value); if (isRecord(value) && key in value) return value[key]; for (const child of Array.isArray(value) ? value : Object.values(value)) { const found = findKey(child, key, seen); if (found !== undefined) return found; } return undefined; }
-function parse1688Axes(value: unknown): Axis[] { const entries = Array.isArray(value) ? value : isRecord(value) && Array.isArray(value.skuProps) ? value.skuProps : []; return entries.flatMap((entry, axisIndex) => { if (!isRecord(entry)) return []; const axisKey = stringAt(entry, ['prop', 'propId', 'id']) ?? `axis:${axisIndex}`; const name = stringAt(entry, ['name', 'propName']) ?? axisKey; const values = (Array.isArray(entry.value) ? entry.value : Array.isArray(entry.values) ? entry.values : []).flatMap((item, valueIndex) => isRecord(item) ? [{ key: stringAt(item, ['valueId', 'vid', 'id']) ?? `${axisKey}:${valueIndex}`, name: stringAt(item, ['name', 'valueName']) ?? `${valueIndex + 1}`, ...(urlAt(item, ['imageUrl', 'image', 'imageUrlLarge']) ? { image_url: urlAt(item, ['imageUrl', 'image', 'imageUrlLarge'])! } : {}) }] : []); return values.length ? [{ key: axisKey, name, values }] : []; }); }
-function parseTaobaoAxes(value: unknown): Axis[] { return (Array.isArray(value) ? value : []).flatMap((entry, index) => { if (!isRecord(entry)) return []; const key = stringAt(entry, ['pid', 'propId']) ?? `axis:${index}`; const name = stringAt(entry, ['name', 'propName']) ?? key; const values = (Array.isArray(entry.values) ? entry.values : []).flatMap((item, valueIndex) => isRecord(item) ? [{ key: stringAt(item, ['vid', 'valueId']) ?? `${key}:${valueIndex}`, name: stringAt(item, ['name', 'valueName']) ?? `${valueIndex + 1}`, ...(urlAt(item, ['image', 'imageUrl']) ? { image_url: urlAt(item, ['image', 'imageUrl'])! } : {}) }] : []); return values.length ? [{ key, name, values }] : []; }); }
-function selectionsForPropPath(path: string, axes: Axis[]) { return path.split(/[;,]/).flatMap((pair) => { const [axisKey, valueKey] = pair.split(':'); const axis = axes.find((item) => item.key === axisKey); const value = axis?.values.find((item) => item.key === valueKey); return axis && value ? [{ axisKey, valueKey, valueName: value.name }] : []; }); }
-function selectionsForCombination(combination: string, axes: Axis[]) { const ids = combination.split(/[;,&]/).map((value) => value.trim()).filter(Boolean); return axes.flatMap((axis) => { const value = axis.values.find((item) => ids.includes(item.key)); return value ? [{ axisKey: axis.key, valueKey: value.key, valueName: value.name }] : []; }); }
-function stringAt(value: Json, keys: string[]) { for (const key of keys) if (typeof value[key] === 'string' || typeof value[key] === 'number') return String(value[key]); return undefined; }
-function moneyAt(value: Json, keys: string[]) { const raw = stringAt(value, keys)?.replace(/,/g, '.'); return raw && /^\d+(\.\d{1,4})?$/.test(raw) ? raw : undefined; }
-function urlAt(value: Json, keys: string[]) { const raw = stringAt(value, keys); return raw && /^https:\/\//.test(raw) ? raw : undefined; }
-function uniqueVariants(variants: Variant[]) { const seen = new Set<string>(); return variants.filter((variant) => !seen.has(variant.sourceVariantKey) && !!seen.add(variant.sourceVariantKey)); }
-function offerId(url: string) { return url.match(/\/offer\/(\d+)\.html/)?.[1] ?? `1688:${new URL(url).pathname}`; }
-function itemId(url: string) { return new URL(url).searchParams.get('id') ?? new URL(url).pathname; }
-function isRecord(value: unknown): value is Json { return !!value && typeof value === 'object' && !Array.isArray(value); }
+function pageState(root: Document) {
+    const all: unknown[] = [];
+    for (const script of Array.from(root.scripts)) {
+        const text = script.textContent?.trim() ?? '';
+        if (text.length < 2 || text.length > 5_000_000) continue;
+        if (script.type === 'application/ld+json') {
+            try {
+                all.push(JSON.parse(text));
+            } catch {
+                /* ignored evidence */
+            }
+        }
+        for (const marker of [
+            'skuMapOriginal',
+            'skuBase',
+            'skuCore',
+            'skuProps',
+        ]) {
+            const object = objectAfterMarker(text, marker);
+            if (object) all.push(object);
+        }
+    }
+    return all;
+}
+function objectAfterMarker(text: string, marker: string): unknown | null {
+    const position = text.indexOf(marker);
+    if (position < 0) return null;
+    const start = text.slice(position).search(/[\[{]/);
+    if (start < 0) return null;
+    const first = position + start;
+    const open = text[first];
+    const close = open === '{' ? '}' : ']';
+    let depth = 0;
+    let quote = '';
+    let escaped = false;
+    for (
+        let index = first;
+        index < text.length && index - first < 2_000_000;
+        index += 1
+    ) {
+        const char = text[index];
+        if (quote) {
+            if (escaped) escaped = false;
+            else if (char === '\\') escaped = true;
+            else if (char === quote) quote = '';
+            continue;
+        }
+        if (char === '"' || char === "'") {
+            quote = char;
+            continue;
+        }
+        if (char === open) depth += 1;
+        if (char === close && --depth === 0) {
+            try {
+                return JSON.parse(text.slice(first, index + 1));
+            } catch {
+                return null;
+            }
+        }
+    }
+    return null;
+}
+function findFirst(states: unknown[], key: string): unknown {
+    for (const state of states) {
+        const found = findKey(state, key);
+        if (found !== undefined) return found;
+    }
+    return undefined;
+}
+function findKey(
+    value: unknown,
+    key: string,
+    seen = new Set<unknown>(),
+): unknown {
+    if (!value || typeof value !== 'object' || seen.has(value))
+        return undefined;
+    seen.add(value);
+    if (isRecord(value) && key in value) return value[key];
+    for (const child of Array.isArray(value) ? value : Object.values(value)) {
+        const found = findKey(child, key, seen);
+        if (found !== undefined) return found;
+    }
+    return undefined;
+}
+function parse1688Axes(value: unknown): Axis[] {
+    const entries = Array.isArray(value)
+        ? value
+        : isRecord(value) && Array.isArray(value.skuProps)
+          ? value.skuProps
+          : [];
+    return entries.flatMap((entry, axisIndex) => {
+        if (!isRecord(entry)) return [];
+        const axisKey =
+            stringAt(entry, ['prop', 'propId', 'id']) ?? `axis:${axisIndex}`;
+        const name = stringAt(entry, ['name', 'propName']) ?? axisKey;
+        const values = (
+            Array.isArray(entry.value)
+                ? entry.value
+                : Array.isArray(entry.values)
+                  ? entry.values
+                  : []
+        ).flatMap((item, valueIndex) =>
+            isRecord(item)
+                ? [
+                      {
+                          key:
+                              stringAt(item, ['valueId', 'vid', 'id']) ??
+                              `${axisKey}:${valueIndex}`,
+                          name:
+                              stringAt(item, ['name', 'valueName']) ??
+                              `${valueIndex + 1}`,
+                          ...(urlAt(item, [
+                              'imageUrl',
+                              'image',
+                              'imageUrlLarge',
+                          ])
+                              ? {
+                                    image_url: urlAt(item, [
+                                        'imageUrl',
+                                        'image',
+                                        'imageUrlLarge',
+                                    ])!,
+                                }
+                              : {}),
+                      },
+                  ]
+                : [],
+        );
+        return values.length ? [{ key: axisKey, name, values }] : [];
+    });
+}
+function parseTaobaoAxes(value: unknown): Axis[] {
+    return (Array.isArray(value) ? value : []).flatMap((entry, index) => {
+        if (!isRecord(entry)) return [];
+        const key = stringAt(entry, ['pid', 'propId']) ?? `axis:${index}`;
+        const name = stringAt(entry, ['name', 'propName']) ?? key;
+        const values = (
+            Array.isArray(entry.values) ? entry.values : []
+        ).flatMap((item, valueIndex) =>
+            isRecord(item)
+                ? [
+                      {
+                          key:
+                              stringAt(item, ['vid', 'valueId']) ??
+                              `${key}:${valueIndex}`,
+                          name:
+                              stringAt(item, ['name', 'valueName']) ??
+                              `${valueIndex + 1}`,
+                          ...(urlAt(item, ['image', 'imageUrl'])
+                              ? {
+                                    image_url: urlAt(item, [
+                                        'image',
+                                        'imageUrl',
+                                    ])!,
+                                }
+                              : {}),
+                      },
+                  ]
+                : [],
+        );
+        return values.length ? [{ key, name, values }] : [];
+    });
+}
+function selectionsForPropPath(path: string, axes: Axis[]) {
+    return path.split(/[;,]/).flatMap((pair) => {
+        const [axisKey, valueKey] = pair.split(':');
+        const axis = axes.find((item) => item.key === axisKey);
+        const value = axis?.values.find((item) => item.key === valueKey);
+        return axis && value
+            ? [{ axisKey, valueKey, valueName: value.name }]
+            : [];
+    });
+}
+function selectionsForCombination(combination: string, axes: Axis[]) {
+    const ids = combination
+        .split(/[;,&]/)
+        .map((value) => value.trim())
+        .filter(Boolean);
+    return axes.flatMap((axis) => {
+        const value = axis.values.find((item) => ids.includes(item.key));
+        return value
+            ? [
+                  {
+                      axisKey: axis.key,
+                      valueKey: value.key,
+                      valueName: value.name,
+                  },
+              ]
+            : [];
+    });
+}
+function stringAt(value: Json, keys: string[]) {
+    for (const key of keys)
+        if (typeof value[key] === 'string' || typeof value[key] === 'number')
+            return String(value[key]);
+    return undefined;
+}
+function moneyAt(value: Json, keys: string[]) {
+    const raw = stringAt(value, keys)?.replace(/,/g, '.');
+    return raw && /^\d+(\.\d{1,4})?$/.test(raw) ? raw : undefined;
+}
+function urlAt(value: Json, keys: string[]) {
+    const raw = stringAt(value, keys);
+    return raw && /^https:\/\//.test(raw) ? raw : undefined;
+}
+function uniqueVariants(variants: Variant[]) {
+    const seen = new Set<string>();
+    return variants.filter(
+        (variant) =>
+            !seen.has(variant.sourceVariantKey) &&
+            !!seen.add(variant.sourceVariantKey),
+    );
+}
+function offerId(url: string) {
+    return (
+        url.match(/\/offer\/(\d+)\.html/)?.[1] ??
+        `1688:${new URL(url).pathname}`
+    );
+}
+function itemId(url: string) {
+    return new URL(url).searchParams.get('id') ?? new URL(url).pathname;
+}
+function isRecord(value: unknown): value is Json {
+    return !!value && typeof value === 'object' && !Array.isArray(value);
+}