Enable Set Sites using the normal head hook automatically get discovery links:<\/p>\n Keep discovery out of human-facing page bodies and footers. Do not add visible\nAPI-help\/Markdown navigation or hidden body links. Agents can discover the APIs\nthrough HTTP headers, the HTML head and dedicated discovery documents. Clients\nthat only expose body text may not discover them; that is an accepted constraint.<\/p>\n Override Standard resource responses receive HTTP discovery links as well. Custom\ncontrollers must call JSON export and Markdown alternates are advertised only when the resource is visible to an\nanonymous visitor and is not excluded by site, resource or category noindex\nsettings. Visibility is checked even when the visitor is an administrator.\nThese exclusions affect discovery, not the permissions of existing API routes.<\/p>\n Visit an ordinary public resource while signed out. Check its HTML links and\nHTTP Check site-specific routes and proxy rules if a discovery URL is unavailable.\nThe module generates URLs through dispatch and uses anonymous metadata. Discovery\ndocuments currently use no-cache headers; existing APIs keep their cache and\n The following routes always use anonymous permissions, including for signed-in\nadministrators. JSON uses the Request All reads accept Search accepts Relationships accept Follow Relationship pages fetch only Other callers can use Markdown includes body HTML, text and header blocks, and one level of page-reference content. All Content responses use ETags and mod_agent_discovery<\/code> on a public publishing site, using the Modules\npage in the admin. It depends on mod_base<\/code> and mod_search<\/code>; it does not require\nmod_seo<\/code>. Activation adds these routes:<\/p>\nURL<\/th> Purpose<\/th><\/tr><\/thead> \/api-help<\/code><\/td>Readable API explanation with a clickable public search example<\/td><\/tr> \/openapi.json<\/code><\/td>OpenAPI 3.1.0 description of concise reads and compatibility search\/export<\/td><\/tr> \/.well-known\/api-catalog<\/code><\/td>Linkset catalog pointing to the public API and its descriptions<\/td><\/tr> \/llms.txt<\/code><\/td>Concise site guidance and links for agents<\/td><\/tr><\/tbody><\/table>\n mod_agent_discovery.description<\/code> to an optional public description of the\nsite's information. The site title comes from the existing site configuration.\nNo API key is needed to read public content. The existing APIs continue to apply\nthe caller's permissions; this module does not grant access or change query limits.<\/p>\nLink from pages<\/h2>\n
{% all include "_html_head.tpl" %}\n<\/code><\/pre>\nagent_api_help.tpl<\/code> to customize the explanation for the site.\nIts discovery<\/code> variable contains public metadata and documented operations.\nThe sitemap is linked when mod_seo_sitemap<\/code> is active.<\/p>\nz_context:set_resource_headers\/2<\/code> to receive them.\nExisting Link values, including WebSub advertisements, are preserved when the\ndiscovery observer runs. Later custom observers must also merge Link values.<\/p>\nVerify the site<\/h2>\n
Link<\/code> header, follow the API-help link, run the search example, and use an\nID from the result to retrieve a full export. Check the export's language-specific\npage_url<\/code> map when citing it. Compare a private and a noindex resource to confirm\nthat no JSON alternate is advertised for either.<\/p>\nX-Robots-Tag: noindex,nofollow<\/code> behavior. No robots.txt or CORS settings are changed.\nDo not infer reuse licenses from public accessibility; retain the site's stated\ncontent and media attribution and licensing terms.<\/p>\nRead concise public content<\/h2>\n
status<\/code>\/result<\/code> envelope.<\/p>\nURL<\/th> Purpose<\/th><\/tr><\/thead> \/api\/agent\/search<\/code><\/td>Browse or full-text search localized summaries<\/td><\/tr> \/api\/agent\/resource\/{id}<\/code><\/td>Read localized Markdown with citation and media links<\/td><\/tr> \/api\/agent\/related\/{id}<\/code><\/td>Follow one hop of outgoing or incoming relationships<\/td><\/tr> \/api\/agent\/schema<\/code><\/td>Discover category or predicate resources<\/td><\/tr> \/agent\/resource\/{id}\/markdown<\/code><\/td>Read a resource as text\/markdown<\/code><\/td><\/tr><\/tbody><\/table>\nAccept: text\/markdown<\/code> on \/id\/{id}<\/code> to negotiate the Markdown URL.\nMarkdown alternates are advertised in HTML head and HTTP Link metadata on public\nindexable pages.<\/p>\nlanguage<\/code>, defaulting to the configured site language. Resources\nreport requested_language<\/code>, actual language<\/code>, available_languages<\/code> and\nis_fallback<\/code>. Missing requested translations fall back to the site language,\nthen the resource's first language. The human page_url<\/code> is the citation URL;\nuri<\/code> is the entity identifier.<\/p>\ntext<\/code> (up to 512 bytes) and cat<\/code> (category name or ID).\nsort<\/code> accepts id<\/code> (resource ID ascending, the default), newest<\/code> (publication\nstart descending), oldest<\/code> (publication start ascending) or modified<\/code>\n(modification time descending). All orders use resource ID ascending as a stable\ntie-breaker. publication_after<\/code> and publication_before<\/code>\naccept valid YYYY-MM-DD<\/code> dates and filter publication start inclusively in UTC.\nThe before date includes the whole day. Reversed ranges are rejected.\nSummaries include the resource's UTC modified<\/code> timestamp.<\/p>\ndirection=in<\/code> or direction=out<\/code> (default) and an optional\npredicate<\/code> ID or name. Unknown or inaccessible predicates return 404.\nSchema accepts kind=category<\/code> (default) or kind=predicate<\/code>. Categories include\ntheir visible parent. Predicates include up to 100 candidate subject\/object\ncategory constraints, filtered by public visibility and noindex. These describe\npublic vocabulary, not validation or access rules; hidden constraints are never\nexported. constraints_truncated<\/code> marks the bound. Names, localized descriptions\nand resource URIs identify the site's vocabulary without exposing database fields.\nPaged reads accept page<\/code> and pagelen<\/code>: default 20, maximum 100 per page,\nwith page * pagelen <= 10000<\/code>. Unknown, duplicate and malformed query parameters\nare rejected. Queries have a five-second execution budget.<\/p>\nnext_url<\/code> until next<\/code> is false. These are candidate pages<\/strong>: public\nvisibility and site\/resource\/category noindex filtering can leave a page empty\nwhile later pages still contain results. No totals are exposed. limit_reached<\/code>\nis true if exploration stops at the 10,000-candidate limit; refine the search.\nDirect reads allow public noindex content. Missing, private and unpublished\nresources all return 404. The facade never returns arbitrary resource properties.<\/p>\npagelen + 1<\/code> candidate edges from PostgreSQL,\nthrough m_edge:get_page\/6<\/code>, before visibility\/noindex filtering. An optional\npredicate is applied in SQL. Ordering is by predicate ID and endpoint resource ID,\nnot editorial edge sequence. This matches the existing outgoing\/incoming unique\nedge indexes and avoids sorting all incoming edges of a popular keyword.\nOffsets still cost work proportional to the skipped candidates; the existing\n10,000-candidate traversal limit remains. Concurrent edge edits can shift pages.<\/p>\n\/api\/model\/edge\/get\/o\/{id}\/{predicate}<\/code>, s<\/code>, o_props<\/code>\nor s_props<\/code> with optional offset<\/code> and limit<\/code> query parameters. Without either,\nthese methods retain their full-list behavior. With either, offset defaults to\n0 and limit to 20 (maximum 100, offset plus limit at most 10,000). They return the\nexisting edge tuples or property lists, in editorial seq, id<\/code> order, with the\nsame caller ACL semantics and no continuation metadata. The predicate is required.\nSQL selects the requested subset, but sequence ordering may require a sort.\nThe agent facade uses the separate Erlang get_page\/6<\/code> helper for indexed\nendpoint ordering, filtered candidate pages and continuation metadata.<\/p>\nsurvey_...<\/code> blocks are intentionally skipped without marking the content\nincomplete. Unsupported blocks and oversized\ncontent are explicitly marked; check content_complete<\/code> and follow the source\npage for omitted material. Inline images use the existing Markdown conversion.\nThe Markdown response ends with a concise “API access” section linking to API\nhelp, the OpenAPI specification and this resource's JSON endpoint in the content's\nactual language. These links are not added to human-facing HTML or to the JSON\nresource's editorial markdown<\/code> field. They assist clients that already fetch\nthe Markdown representation; header\/head discovery remains unchanged.\nIf an editorial template cannot render, the representation returns HTTP 503 with\nCache-Control: no-store<\/code>, rather than successful empty content. This can occur\nwhile the site's templates are becoming available after startup; retry later.\nLegitimate empty content still renders successfully.\nJSON includes a resource's own downloadable medium when available, exposing only\nMIME type, dimensions, size and download URL. Preserve supplied licensing and\nattribution. Page references include a source link, language and supplied licensing\/attribution;\nprivate or unpublished references are omitted. Nested non-survey blocks are not\nexpanded recursively and mark the representation incomplete. This prevents cycles\nand large graphs from turning a resource read into an unbounded export.<\/p>\nCache-Control: public, max-age=0, must-revalidate<\/code>.\nEach conditional request rebuilds the representation after anonymous permission\nand publication checks. No server-side representation cache is introduced.\nErrors use no-store: 400 for invalid requests, 404 for unavailable resources,\n413 for responses over 1 MiB, 503 for backend failures and 504 for timeouts.\nBoth GET and HEAD are supported; no credentialed CORS policy is added.<\/p>\n