Update seller store experience

2026-09-08 16:19:38 +08:00
parent ceb644a3f9
commit 319887726e
25 changed files with 31417 additions and 127 deletions
@@ -61,6 +61,7 @@ export async function collectOzonProduct(
    const localStates = readLocalWidgetStates(documentRoot);
    let attributes = extractOzonCharacteristics(localStates);
    let description = extractOzonDescription(localStates);
+    const jsonLdDescription = extractOzonJsonLdDescription(documentRoot);
    const heading = firstState(localStates, 'webProductHeading');
    const priceState = firstState(localStates, 'webPrice');
    const title =
@@ -101,6 +102,14 @@ export async function collectOzonProduct(
        'ozon:gallery:page',
        canonical,
    );
+    addGalleryVideos(
+        localStates,
+        currentSku,
+        assets,
+        assetKeys,
+        'ozon:gallery:page',
+        canonical,
+    );
    if (attributes.length) {
        evidence.push({
            ref: 'ozon:characteristics:page',
@@ -109,6 +118,14 @@ export async function collectOzonProduct(
            raw_value: `发现 ${attributes.length} 项商品特征`,
        });
    }
+    if (jsonLdDescription) {
+        evidence.push({
+            ref: 'ozon:json-ld:product.description',
+            kind: 'json_ld',
+            path: 'script[type="application/ld+json"] Product.description',
+            raw_value: `JSON-LD 商品描述 ${jsonLdDescription.value.length} 个字符`,
+        });
+    }

    const categories = collectCategories(documentRoot, canonical);
    if (categories.length) {
@@ -162,12 +179,22 @@ export async function collectOzonProduct(
                canonical,
            );
            description = description ?? extractOzonDescription(detailStates);
+            const detailAttributes = extractOzonCharacteristics(detailStates);
+            attributes = mergeOzonCharacteristics(attributes, detailAttributes);
            evidence.push({
                ref: 'ozon:detail:page',
                kind: 'state',
                path: 'webDescription|webRichContent|webFeatures image fields',
                raw_value: `发现 ${added} 张商品详情图`,
            });
+            if (detailAttributes.length) {
+                evidence.push({
+                    ref: 'ozon:characteristics:detail',
+                    kind: 'state',
+                    path: 'webShortCharacteristics|webDetailedCharacteristics|webCharacteristics.characteristics[]',
+                    raw_value: `详情段补全 ${detailAttributes.length} 项商品特征`,
+                });
+            }
        } catch (error) {
            evidence.push(
                fetchFailureEvidence('ozon:detail:page', currentSku, error),
@@ -230,6 +257,14 @@ export async function collectOzonProduct(
                    `ozon:gallery:${result.sku}`,
                    canonical,
                );
+                addGalleryVideos(
+                    result.states,
+                    result.sku,
+                    assets,
+                    assetKeys,
+                    `ozon:gallery:${result.sku}`,
+                    canonical,
+                );
                if (result.sku === currentSku) {
                    const detailedAttributes = extractOzonCharacteristics(
                        result.states,
@@ -270,6 +305,11 @@ export async function collectOzonProduct(
        });
    }

+    // Ozon's rich widget is preferred because it is the dedicated description
+    // segment. JSON-LD is the proven old-collector fallback for pages where
+    // Ozon only places the seller description in the initial document.
+    description = description ?? jsonLdDescription;
+
    addAspectCoverAssets(axes, assets, assetKeys, canonical);

    const variants = [...candidates.values()]
@@ -301,7 +341,7 @@ export async function collectOzonProduct(
            itemKey: currentSku,
            canonicalUrl: canonical,
            collectedAt: new Date().toISOString(),
-            parserVersion: 'ozon-public-2.2.0',
+            parserVersion: 'ozon-public-2.3.0',
        },
        common: {
            ...(title
@@ -452,6 +492,7 @@ function addAspectCoverAssets(
            assetKeys.add(key);
            assets.push({
                url,
+                mediaType: 'image',
                role: 'source_sku',
                relation: 'axis_value',
                sortOrder: assets.length,
@@ -491,6 +532,7 @@ function addDetailAssets(
        assetKeys.add(key);
        assets.push({
            url,
+            mediaType: 'image',
            role: 'source_detail',
            relation: 'product_candidate',
            sortOrder: added++,
@@ -621,6 +663,55 @@ export function extractOzonDescription(
    return undefined;
 }

+/** Reads only schema.org Product.description as a fallback, never arbitrary
+ * JSON-LD page copy. This mirrors the mature collector's JSON-LD path. */
+export function extractOzonJsonLdDescription(
+    root: Document,
+): OzonDescription | undefined {
+    for (const script of Array.from(root.scripts)) {
+        if (script.type !== 'application/ld+json' || !script.textContent?.trim())
+            continue;
+        try {
+            const product = findJsonLdProduct(JSON.parse(script.textContent));
+            const raw = product && typeof product.description === 'string'
+                ? product.description.trim()
+                : '';
+            if (!raw) continue;
+            return {
+                value: raw.replace(/\s+\n/g, '\n').replace(/\n{3,}/g, '\n\n').slice(0, 100_000),
+                rawValue: raw.slice(0, 120_000),
+            };
+        } catch {
+            // A malformed analytics JSON-LD block is not product evidence.
+        }
+    }
+    return undefined;
+}
+
+function findJsonLdProduct(
+    value: unknown,
+    visited = new Set<unknown>(),
+): JsonRecord | undefined {
+    if (!value || typeof value !== 'object' || visited.has(value)) return undefined;
+    visited.add(value);
+    if (Array.isArray(value)) {
+        for (const child of value) {
+            const product = findJsonLdProduct(child, visited);
+            if (product) return product;
+        }
+        return undefined;
+    }
+    const record = value as JsonRecord;
+    const type = record['@type'];
+    if (type === 'Product' || (Array.isArray(type) && type.includes('Product')))
+        return record;
+    if (Array.isArray(record['@graph'])) {
+        const product = findJsonLdProduct(record['@graph'], visited);
+        if (product) return product;
+    }
+    return undefined;
+}
+
 function findRichAnnotationJson(
    value: unknown,
    visited = new Set<unknown>(),
@@ -716,6 +807,7 @@ function addGalleryAssets(
        assetKeys.add(key);
        assets.push({
            url,
+            mediaType: 'image',
            // A gallery is product photography for this real SKU.  It is not
            // a SKU-option thumbnail merely because it is the second image.
            role: 'source_main',
@@ -727,6 +819,103 @@ function addGalleryAssets(
    }
 }

+/** Adds only Ozon's named product-gallery videos. Page-wide <video> nodes can
+ * belong to reviews or recommendations and are intentionally never considered. */
+function addGalleryVideos(
+    states: JsonRecord[],
+    sku: string,
+    assets: Asset[],
+    assetKeys: Set<string>,
+    evidenceRef: string,
+    pageUrl: string,
+) {
+    const galleries = states.filter((state) => {
+        const widget = String(state.__atlasWidgetName ?? '').replace(/^state-/, '');
+        return /^web(?:Gallery|Video)-/.test(widget);
+    });
+    for (const gallery of galleries) {
+        for (const video of extractOzonGalleryVideos(gallery, pageUrl)) {
+            if (assets.length >= 500) return;
+            const key = `video:${sku}:${video.url}`;
+            if (assetKeys.has(key)) continue;
+            assetKeys.add(key);
+            assets.push({
+                url: video.url,
+                mediaType: 'video',
+                ...(video.posterUrl ? { posterUrl: video.posterUrl } : {}),
+                role: 'source_main',
+                variantSourceKey: sku,
+                relation: 'variant_direct',
+                sortOrder: assets.length,
+                evidenceRef,
+            });
+        }
+    }
+}
+
+/** Pure helper for the strictly named Ozon gallery/video state. HLS, blobs,
+ * and non-HTTPS URLs stay out of the collection contract in this first phase. */
+export function extractOzonGalleryVideos(gallery: JsonRecord, pageUrl: string) {
+    const videos: Array<{ url: string; posterUrl?: string }> = [];
+    const seen = new Set<string>();
+    // Ozon changes whether the direct file lives in videos[], video, or a
+    // nested gallery payload. The old collector walked this *one allowlisted
+    // widget* for direct video extensions; retaining that scope is critical.
+    for (const { value, parent } of nestedVideoValues(gallery)) {
+        const url = directVideoUrl(
+            typeof value === 'string'
+                ? value
+                : stringAt(value, ['url', 'videoUrl', 'src']),
+            pageUrl,
+        );
+        if (!url) continue;
+        const posterUrl = ozonImageUrl(
+            (typeof value === 'string'
+                ? undefined
+                : stringAt(value, ['coverImage', 'coverUrl', 'poster', 'previewImage']))
+                ?? stringAt(parent, ['coverImage', 'coverUrl', 'poster', 'previewImage']),
+            pageUrl,
+        );
+        if (seen.has(url)) {
+            const existing = videos.find((video) => video.url === url);
+            if (existing && !existing.posterUrl && posterUrl)
+                existing.posterUrl = posterUrl;
+            continue;
+        }
+        seen.add(url);
+        videos.push({ url, ...(posterUrl ? { posterUrl } : {}) });
+    }
+    return videos;
+}
+
+function nestedVideoValues(root: JsonRecord) {
+    const values: Array<{ value: string | JsonRecord; parent?: JsonRecord }> = [];
+    const pending: Array<{ value: unknown; parent?: JsonRecord }> = [{ value: root }];
+    const seen = new Set<object>();
+    while (pending.length) {
+        const current = pending.pop()!;
+        if (typeof current.value === 'string') {
+            if (/\.(mp4|webm)(?:$|[?#])/i.test(current.value))
+                values.push({ value: current.value, parent: current.parent });
+            continue;
+        }
+        if (!current.value || typeof current.value !== 'object') continue;
+        if (seen.has(current.value)) continue;
+        seen.add(current.value);
+        if (Array.isArray(current.value)) {
+            for (const child of current.value)
+                pending.push({ value: child, parent: current.parent });
+            continue;
+        }
+        const record = current.value as JsonRecord;
+        if (stringAt(record, ['url', 'videoUrl', 'src']))
+            values.push({ value: record, parent: current.parent });
+        for (const child of Object.values(record))
+            pending.push({ value: child, parent: record });
+    }
+    return values;
+}
+
 /** Pure helper used by fixture tests and the structured Ozon gallery reader. */
 export function extractOzonGalleryImageUrls(
    gallery: JsonRecord,
@@ -777,7 +966,7 @@ function widgetStatesFromPageJson(value: unknown): JsonRecord[] {
    const states: JsonRecord[] = [];
    for (const [widgetName, raw] of Object.entries(value.widgetStates)) {
        if (
-            !/^web(Aspects|AspectsModal|ProductHeading|Price|Gallery|Description|RichContent|Features|ShortCharacteristics|DetailedCharacteristics|Characteristics)-/.test(
+            !/^web(Aspects|AspectsModal|ProductHeading|Price|Gallery|Video|Description|RichContent|Features|ShortCharacteristics|DetailedCharacteristics|Characteristics)-/.test(
                widgetName,
            )
        )
@@ -997,3 +1186,15 @@ function ozonImageUrl(value: string | undefined, pageUrl: string) {
    const url = normaliseImageUrl(value, pageUrl, 'ozon');
    return url?.startsWith('https://') ? url : undefined;
 }
+
+function directVideoUrl(value: string | undefined, pageUrl: string) {
+    if (!value) return undefined;
+    try {
+        const url = new URL(value, pageUrl);
+        if (url.protocol !== 'https:' || /\.m3u8(?:$|[?#])/i.test(url.pathname))
+            return undefined;
+        return /\.(mp4|webm)(?:$|[?#])/i.test(url.pathname) ? url.href : undefined;
+    } catch {
+        return undefined;
+    }
+}