Linkfox Geekbi Temu Market Research

Temu Category & Keyword Demand Research

This skill researches Temu category opportunities and buyer-demand keywords from public marketplace data. Use only the endpoints needed for the user's request, and read references/api.md for the full request and response contract.

The site-list and category-list endpoints are helper capabilities. A helper-only request must not auto-select this skill unless the user explicitly invokes it or the skill is already active.

Core Concepts

• regionId selects the marketplace. Resolve it from a non-null positive sites[].regionId; never substitute siteId. The gateway default is regionId: 211 (United States).
• dsr is a supply-demand signal, not a standalone go/no-go score. Cross-check demand scale, recent growth, item/shop supply, and price.
• Search responses are paginated items[]. Defaults are page: 1 and size: 20; size is 1–200 and page * size must not exceed 10,000. order has no gateway default; when supplied, it must be asc or desc (case-insensitive).
• Range filters are limited to dsr, totalSold, avgPrice, totalSales, itemCount, and mallCount, each with Min / Max. Never make a minimum greater than its maximum; all except dsr must be non-negative.
• Use site/category helpers only when the user has not supplied a valid ID. Stop if no unambiguous ID is returned.
• Do not launch a paid search merely to answer a site-list or category-tree request.

Endpoint: /geekbi/temu/categorySearch

Request

• Category research uses keyword to compare market capacity and competition. Use the separate category-list helper with parentCatId when a trusted category tree must be traversed; categorySearch itself does not accept catLevel or parentCatId.
• Use keyword for a Chinese or English category name. catLevel and parentCatId are response/helper concepts, not categorySearch request fields.
• Filter or rank broad demand with totalSold; use dsr for candidate discovery. Treat monthSoldRate and monthSalesRate as response fields for recent-growth analysis, not as range-filter request fields.
• itemCount and mallCount can be filtered with Min/Max fields. Period *ItemCount and *MallCount values are response-only changes and can be negative.

# Compare US pet-related categories by cumulative demand
python scripts/geekbi_temu_category_search.py '{"regionId":211,"keyword":"宠物","totalSoldMin":1000,"sort":"totalSold","order":"desc","page":1,"size":20}'

Response

Endpoint: /geekbi/temu/keywordSearch

Request

• Keyword research uses keyword for a specific phrase or catIds for category-led discovery. Combine them only when the user explicitly wants a phrase within a category.
• Resolve catIds from valid categories[].catId values or a verified category-search result. Prefer level-1 or level-2 IDs for broad discovery.
• Search a supplied phrase with keyword without silently translating or replacing it.
• For category-led discovery, use catIds from the category helper. Multiple IDs match any supplied category.
• Rank hot terms with totalSold and blue-ocean candidates with dsr. Analyze recent demand from returned monthSoldRate or monthSalesRate; these period metrics are not range-filter request fields. Use firstOnSaleTimeMin/Max for market recency.
• firstOnSaleTimeMin/Max use ISO-8601 timestamps and describe the earliest associated product listing time.

# Discover dress-related demand terms within a verified category
python scripts/geekbi_temu_keyword_search.py '{"regionId":211,"keyword":"dress","catIds":[27011],"totalSoldMin":1000,"sort":"totalSold","order":"desc","page":1,"size":20}'

Response

Endpoint: /geekbi/temu/siteList

Request

# Resolve a non-US marketplace; use regionId, not siteId
python scripts/geekbi_temu_site_list.py '{}'

Response

Endpoint: /geekbi/temu/categoryList

Request

# Resolve a trusted category path when only a category name is known
python scripts/geekbi_temu_category_list.py '{}'
python scripts/geekbi_temu_category_list.py '{"parentCatId":27011}'

Response

Fields can be absent or null. Preserve runtime JSON types and raw rate values; do not replace missing values with zero.

调用方式

• API 端点：POST /geekbi/temu/categorySearch、POST /geekbi/temu/keywordSearch、POST /geekbi/temu/siteList 或 POST /geekbi/temu/categoryList（完整参数/响应/错误码见 references/api.md）
• Python 脚本：python scripts/geekbi_temu_category_search.py '<JSON 参数>' [--inline]、python scripts/geekbi_temu_keyword_search.py '<JSON 参数>' [--inline]、python scripts/geekbi_temu_site_list.py '{}' [--inline] 或 python scripts/geekbi_temu_category_list.py '<JSON 参数>' [--inline]
• 成本约束：类目搜索和关键词搜索各消耗 15 算力/次，站点列表和品类列表不消耗算力；同一端点同一参数组合使用 24h 本地缓存，同一会话默认只调用一次。--inline 不绕过缓存；--no-cache 会跳过缓存读写并强制真实请求，付费端点可能再次扣算力，免费辅助端点仅强制刷新。失败、空结果或瞬时错误不得自动重试，也不得自动换关键词、翻页、改站点或筛选条件连续试探；需要继续付费检索时先向用户说明会产生额外消耗。

输出策略（脚本默认行为）：

• 始终将完整响应写入 <cwd>/linkfox/<YYYY-MM-DD>/<session>/data/linkfox-geekbi-temu-market-research-<timestamp>.json；<cwd> 为脚本执行时的当前工作目录，<session> 取自环境变量 SESSION_ID，未设置时自动生成。当前目录不可写时直接报错，不回退到用户目录或系统临时目录。
• 响应体 ≤ 8 KB：落盘后把完整 JSON 打印到 stdout
• 响应体 > 8 KB：落盘后 stdout 只输出摘要（顶层状态字段、常见业务计数、忽略 columns 后最大业务列表的长度 + 前 3 条样本）
• 加 --inline 强制全量打印到 stdout（同样落盘）

读数据建议：先看摘要判断是否足够；需要具体字段时优先用 jq或ConvertFrom-Json 从保存的 json 文件按需抽取，避免整份 JSON 进入上下文。