hosting

POST /v1/sites/{siteId}/chat-agent/sources

Tell the agent what to read, and read it now.

所有 hosting 端点

所有开发者文档

身份验证

请将 API 密钥作为 bearer 令牌发送。此端点在规范中未指明具体的权限,因此请为您的密钥赋予所需的最小权限,并通过检查响应来确认,而不是盲目假设。

此端点不需要组织 ID。您的密钥已用于识别其所属的组织,且响应范围也仅限于该组织。

免费试用

将尖括号中的内容替换为您自己的值,并将键占位符替换为您仪表板中的一个键。

curl -X POST https://api.zinndigital.com/v1/sites/{siteId}/chat-agent/sources \
  -H "Authorization: Bearer zdk_live_…" \
  -H "Content-Type: application/json" \
  -d '{ "kind": <string<hosted_files, crawl, manual>> }'

已登录?您仪表板中的 API 控制台会自动填入您真实的组织 ID 和您自己的密钥,并针对实时 API 运行请求,以便您查看实际的响应。 在 API 控制台中打开此端点

详细信息

Three kinds of source, and the first is the one no other host can offer. hosted_files reads the site's pages straight off the box we host it on — no crawl, no rate limit, no robots.txt, and no page missed because nothing links to it. It is available only on our own fleet; on a platform we do not run, this returns a source in state failed saying so in as many words rather than quietly falling back. crawl fetches pages over HTTP from an address you give and follows same-site links. It works anywhere, including sites we do not host. manual is text you type — opening hours, a returns policy, an FAQ. Indexing runs immediately and the response says how many pages were read and whether it stopped at a ceiling. truncated is present when it did.

参数

名称类型必填内容简介
siteId (path)UuidSite ID (UUIDv7).

请求正文

名称类型必填内容简介
kindstring<hosted_files, crawl, manual>
urlstringRequired for crawl — the address to start from. Same-site links are followed.
textstringRequired for manual.
titlestringA label you will recognise on the screen.

响应

名称类型必填内容简介
sourceChatAgentSourceOne place the agent's knowledge comes from, and how that last went.
pages_readinteger
truncatedstring⛔ Present only when indexing stopped at a ceiling rather than at the end of the site, and it says how far it got. A truncation nobody reports reads as a complete index.

此端点可能返回的错误

401 · 403 · 404 · 422