不可变版本 / Immutable version 2026-10-25.2. 本副本不加载分析或应用脚本 / No analytics or app scripts.
适用于搜索 API、网页提取、Playground 与 MCP 接口。
2026-10-25.2 · 2026-10-25
本附件适用于 Laintas Search 的相应功能，与平台服务条款、隐私政策、数据用途说明及收费与退款政策共同适用。仅在本附件明确的产品事项上补充或优先于总条款；强制性法律及双方有效的个别协商约定优先。以购买或使用时明确提供且依法有效接受的版本为准，网页更新不自动改变已有订单。
搜索提供标题、URL、摘要或网页提取结果，可能受索引时间、页面变化、权限、地区和上游状态影响，不保证覆盖所有页面或内容真实完整。结果中的链接和摘录不授予原站内容、数据库、图片或付费资料的再发布权；请遵守原站许可、访问限制和适用法律，不得绕过登录、付费墙或其他权限。
请求可预留额度，成功后按当前接口的单位计量，失败按该接口的预留退回和去重机制处理。搜索和网页提取可能采用不同单位，不应把供应商内部重试次数直接当作用户调用次数。客户端新提交的有效请求可能单独计费；幂等或去重保障只在接口支持的字段和窗口内成立。空结果、部分结果、缓存和超时应按接口文档与实际交付状态核查，错误计费可请求更正。
Search 控制台请求日志不保存原始查询正文，记录状态、数量、耗时、用量等元数据；这不表示查询未被处理或所有上游、网络和用户自己的应用都不保存内容。查询与所需 URL 或页面内容会传给搜索路径，必要时使用 Serper 等已配置的后备服务；访问目标站点本身也会产生连接。不要在查询、URL 或 API key 标签中放入不必要的敏感信息。
保护 API key 并为泄露密钥撤销授权；合法请求、接口配额、并发和速率限制同时适用。平台可为稳定性和防滥用进行相称限流，不保证无限并发或永久免费试用。REST 与 MCP 是同一服务的不同入口，不能用于重复领取或规避额度。涉及第三方用户数据时，调用者仍须履行对其用户的告知、授权和保护义务。
搜索与网页提取分别按当前接口文档和购买确认的单位计量，不因使用同一个 key 就采用同一种折算。空搜索结果可能是有效检索结果，不当然等于接口失败；网页被禁止访问、正文未提取和部分 URL 成功则应按实际接口交付与已公布计费规则核对。超时或网络错误应区分客户端没有收到响应和服务端实际未完成，不能仅凭 HTTP 状态码就断言全部免费或全部计费。供应商内部重试不得在无相应交付或授权依据时重复转成多笔用户费用。
仅在接口支持且正确使用约定请求标识时，才能主张相应去重保障。当前 Search 支持的 Idempotency-Key 去重记录技术窗口为七天；窗口外、缺失标识、改变请求内容或自行生成新标识的调用，不保证与旧请求合并。网络中断后应先查可得用量及请求状态再重发；重试可能返回既有结果、失败或形成新有效请求，应按记录核实。请求元数据的保存不意味着可以恢复未保存的查询正文；费用争议请提供时间、key 标识而非完整密钥、请求号及接口。
网页内容可因 robots、登录、地区、付费和原站许可而不可用，接口不承诺绕过限制。缓存或摘要可能过时，涉及价格、法律、健康和其他时效信息应核对原站。将结果提供给自己用户时，您仍需履行对其用户的数据告知、授权、来源说明及原站权利义务，不可把本接口访问权解释为原站整库复制或商业再发行许可。抓取目标、查询和页面内容可能暴露业务意图，请避免在 URL 或查询中附带无关的访问令牌与个人秘密。
For search APIs, page extraction, the Playground and MCP endpoints.
2026-10-25.2 · 2026-10-25
This annex applies to the relevant Laintas Search functions together with the Platform Terms, Privacy Policy, Data Use Notice and Billing and Refund Policy. It supplements or overrides general terms only on its stated product matters. Mandatory law and valid individually negotiated agreements take priority. The version expressly provided and validly accepted at purchase or use applies; a web update does not automatically change existing orders.
Search returns titles, URLs, snippets or extracted pages and depends on indexing, changes, access, region and upstream state. It does not promise complete coverage or accurate content. Links and excerpts do not license republication of site content, databases, images or paid material. Respect source licences, access restrictions and law; do not bypass login, paywalls or other permissions.
A request may reserve allowance and settle in the endpoint’s units on success, with failures following its reservation-refund and deduplication mechanism. Search and fetch can have different units; internal supplier attempts are not automatically user calls. New valid client requests can be separately charged; idempotency only applies within the supported fields and window. Empty or partial results, cache and timeouts are assessed against endpoint documentation and delivery, with billing corrections available.
Search console request logs omit raw query bodies and record status, counts, latency and usage metadata. This does not mean queries are unprocessed or that every upstream, network or user application retains nothing. Queries and required URLs or page content reach the search path, including configured fallback such as Serper when needed; target sites also receive connections. Do not put unnecessary sensitive information in queries, URLs or key labels.
Protect and revoke leaked API keys. Request validity, quotas, concurrency and rate limits apply together. Proportionate throttling may protect stability and prevent abuse; unlimited concurrency and permanent trials are not promised. REST and MCP are different interfaces to the same service and cannot evade limits or duplicate grants. Callers processing their users’ data retain notice, authority and protection duties to those users.
Search and extraction use their respective documented and confirmed units even with the same key. Empty search results can be valid retrieval, not necessarily endpoint failure. Blocked pages, missing text and partially successful URLs follow actual delivery and disclosed endpoint pricing. Timeouts distinguish an undelivered client response from server non-completion; an HTTP status alone establishes neither universal free use nor universal charging. Internal supplier retries must not become repeated user charges without corresponding delivery or authority.
Deduplication applies only where an endpoint supports it and the agreed identifier is used correctly. Supported Search Idempotency-Key records currently have a seven-day technical window. Outside it, missing IDs, changed bodies or newly generated IDs do not guarantee merging with an earlier request. After a network loss, check available usage and request state before resubmission; retries may return prior results, fail or form new valid requests. Metadata retention does not make unstored queries recoverable. Supply time, a key identifier rather than its secret, request ID and endpoint for billing review.
Robots rules, login, geography, paywalls and licences can limit content access; the API does not promise circumvention. Cached text or summaries may be outdated, especially for prices, law, health and other timely facts. Serving your users still requires notices, authority, attribution and source rights; API access is not a licence to copy or commercially redistribute an entire source database. Targets, queries and pages can disclose business intentions. Avoid unrelated access tokens and personal secrets in URLs and queries.