Develop

docs/about.ttl

@@ -2,6 +2,7 @@
 @prefix rdf:    <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
 @prefix dh:     <https://w3id.org/atomgraph/linkeddatahub/document-hierarchy#> .
 @prefix dct:    <http://purl.org/dc/terms/> .
+@prefix schema: <https://schema.org/> .
 
 <> a dh:Item ;
     dct:title "About LinkedDataHub" ;
@@ -11,7 +12,13 @@
 
 <#overview-video> a ldh:Object ;
     dct:title "LinkedDataHub 6 overview" ;
-    rdf:value <https://www.youtube.com/watch?v=g3i-yH95WZA> .
+    rdf:value <https://www.youtube.com/watch?v=g3i-yH95WZA> ;
+    ldh:graph <> .
+
+<https://www.youtube.com/watch?v=g3i-yH95WZA> a schema:VideoObject ;
+    schema:name "LinkedDataHub 6 overview" ;
+    schema:url <https://www.youtube.com/watch?v=g3i-yH95WZA> ;
+    schema:thumbnailUrl <https://i.ytimg.com/vi/g3i-yH95WZA/hqdefault.jpg> .
 
 <#content> a ldh:XHTML ;
     rdf:value """<div xmlns="http://www.w3.org/1999/xhtml">
@@ -37,18 +44,14 @@
     </div>
     <div>
         <h2 id="kms">Knowledge management and collaboration</h2>
-        <div class="screenshot-placeholder">
-            <span aria-hidden="true" class="msi">play_circle</span>
-            <p>Screen recording: LinkedDataHub overview</p>
-        </div>
         <p>As an RDF-native CMS, LinkedDataHub provides these end-user features:</p>
         <ul>
             <li><a href="../user-guide/create-data/create-content/" target="_blank">creating structured content</a> from Linked Data and SPARQL resources</li>
             <li><a href="../user-guide/create-data/create-documents/" target="_blank">creating RDF documents (graphs)</a></li>
             <li><a href="../user-guide/browse-data/" target="_blank">browsing Linked Data</a></li>
             <li><a href="../user-guide/add-data/" target="_blank">copying RDF resources</a> into the local dataspace</li>
             <li><a href="../user-guide/browse-data/#text-search" target="_blank">searching</a> and <a href="../user-guide/query-data/" target="_blank">querying</a> SPARQL endpoints</li>
-            <li><a href="../reference/data-model/resources/#views" target="_blank">creating collections from SPARQL queries</a></li>
+            <li><a href="../reference/data-model/views/" target="_blank">creating collections from SPARQL queries</a></li>
             <li><a href="../user-guide/edit-content/" target="_blank">editing content</a> in place, with RDFa annotations linking text to resources</li>
             <li><a href="../user-guide/import-data/import-csv-data/" target="_blank">importing CSV data</a> by mapping it to RDF using SPARQL</li>
             <li><a href="../reference/data-model/resources/#charts" target="_blank">visualizing query results as graphs, charts etc.</a></li>

docs/extending.ttl

@@ -12,7 +12,7 @@
 <#content> a ldh:XHTML ;
     rdf:value """<div xmlns="http://www.w3.org/1999/xhtml">
     <p class="ldh-body-lg">Extending LinkedDataHub with your own code: XSLT stylesheets, Java endpoints, Docker images</p>
-    <p>Everything in the <a href="../tutorial/">tutorial</a> and the <a href="../user-guide/">user guide</a> happens through the UI, the CLI and RDF files. The pages in this section pick up where low-code ends: they show how to override the rendering with your own <a href="change-layout/">XSLT stylesheet</a> and how to <a href="build-apps/">package an app</a> with custom Java endpoints and a dedicated Docker image.</p>
+    <p>Everything in the <a href="../tutorial/">tutorial</a> and the <a href="../user-guide/">user guide</a> happens through the UI, the CLI and RDF files. The pages in this section pick up where low-code ends: they show how to override the rendering with your own <a href="change-layout/">XSLT stylesheet</a>, how to <a href="create-packages/">create a package</a> for your own vocabulary, and how to <a href="build-apps/">package an app</a> with custom Java endpoints and a dedicated Docker image.</p>
 </div>"""^^rdf:XMLLiteral .
 
 <#select-children> a ldh:Object ;

docs/extending/create-packages.ttl

@@ -0,0 +1,73 @@
+@prefix ldh:    <https://w3id.org/atomgraph/linkeddatahub#> .
+@prefix rdf:    <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
+@prefix dh:     <https://w3id.org/atomgraph/linkeddatahub/document-hierarchy#> .
+@prefix dct:    <http://purl.org/dc/terms/> .
+
+<> a dh:Item ;
+    dct:title "Create packages" ;
+    dct:description "Bundle the ontology and the stylesheet for your own vocabulary as a reusable package" ;
+    rdf:_1 <#content> .
+
+<#content> a ldh:XHTML ;
+    rdf:value """<div xmlns="http://www.w3.org/1999/xhtml">
+    <p class="ldh-body-lg">Bundle the ontology and the stylesheet for your own vocabulary as a reusable package</p>
+    <p>A <a href="../../reference/administration/packages/">package</a> is a namespace ontology and an XSLT stylesheet, published as Linked Data behind a package URI.
+        Developers can create custom packages for their own domain vocabularies. The process has four steps.</p>
+    <div>
+        <h2 id="write-ontology">Write package ontology</h2>
+        <p>Create <samp>ns.ttl</samp> describing both the document the registry serves it as and the ontology inside it, with the vocabulary import and the views:</p>
+        <pre>&lt;&gt; a dh:Item ;
+    dct:title "Schema.org ontology" ;
+    foaf:primaryTopic &lt;#&gt; .
+
+&lt;#&gt; a owl:Ontology ;
+    owl:imports &lt;https://schema.org/&gt; .
+
+# Attach view to a property
+schema:knows ldh:view :PersonKnows .
+
+:PersonKnows a ldh:View ;
+    dct:title "Knows" ;
+    spin:query :SelectPersonKnows .
+
+:SelectPersonKnows a sp:Select ;
+    sp:text \"\"\"
+    SELECT DISTINCT ?person
+    WHERE { GRAPH ?graph { $about schema:knows ?person } }
+    ORDER BY ?person
+    \"\"\" .</pre>
+        <div class="ac-alert va-warning" role="status">
+            <span class="ac-alert-ic"><span aria-hidden="true" class="msi outline">warning</span></span>
+            <div class="ac-alert-body">
+                <p class="ac-alert-text">The <code>dct:title</code> on <code>&lt;&gt;</code> is not optional. A write is held to the constraints of the document as it will be written, and a LinkedDataHub document must be titled, so an <samp>ns.ttl</samp> that describes only the ontology is refused <samp>422 Unprocessable Entity</samp> when it is pushed.</p>
+            </div>
+        </div>
+    </div>
+    <div>
+        <h2 id="write-stylesheet">Write XSLT stylesheet</h2>
+        <p>Create the stylesheet — name the file for the vocabulary it covers — with XSLT templates using system modes like <code>ac:*</code>, <code>ldh:*</code> and <code>xhtml:*</code>. See the <a href="../../reference/stylesheets/">Stylesheets reference</a> for template customization patterns.</p>
+        <pre>&lt;xsl:template match="schema:knows" mode="ac:PropertyEditor"/&gt;</pre>
+    </div>
+    <div>
+        <h2 id="publish-metadata">Publish package metadata</h2>
+        <p>Publish the descriptor as Linked Data at your package URI, naming the ontology beside it:</p>
+        <pre>&lt;&gt; a dh:Container ;
+    dct:title "Schema.org" ;
+    foaf:primaryTopic &lt;#this&gt; .
+
+&lt;#this&gt; a lds:Package ;
+    dct:title "Schema.org Package" ;
+    dct:description "Schema.org vocabulary support" ;
+    lds:ontology &lt;ns/#&gt; .</pre>
+        <p>Then upload the stylesheet into the package document and add <code>ac:stylesheet</code> naming that upload, which is where a registry published with <samp>ldh push</samp> gets it:</p>
+        <pre>package="https://packages.example.com/schema.org/"
+
+upload=$(ldh add file --title schema.xsl --file schema.xsl "$package")
+echo "INSERT { &lt;${package}#this&gt; &lt;https://w3id.org/atomgraph/client#stylesheet&gt; &lt;$upload&gt; } WHERE { }" | ldh patch "$package"</pre>
+        <p>A dataspace importing the package dereferences its descriptor, ontology and stylesheet anonymously, so all three have to be readable without a certificate.</p>
+    </div>
+    <div>
+        <h2 id="test-installation">Test installation</h2>
+        <p>Declare the <code>ldh:import</code> triple in a test dataspace's <samp>settings</samp> document — see <a href="../../reference/administration/packages/#management">Management</a> for the command.</p>
+    </div>
+</div>"""^^rdf:XMLLiteral .

docs/get-started.ttl

@@ -35,7 +35,7 @@
     </div>
     <div>
         <h2 id="connect-an-agent">Connect an agent</h2>
-        <p>AI agents such as Claude Code or Cursor work with LinkedDataHub through the same HTTP API, using Web-Algebra's MCP server. Connected that way, an agent can declare a working application in a dataspace — containers, views and charts — instead of generating code for one.</p>
+        <p>AI agents such as Claude Code or Cursor work with LinkedDataHub through the same HTTP API, with the <samp>ldh</samp> Claude Code plugin or over Web-Algebra's MCP server. Connected that way, an agent can declare a working application in a dataspace — containers, views and charts — instead of generating code for one.</p>
         <p class="ldh-body-lg"><a href="agents/">Read instructions on connecting an agent.</a></p>
     </div>
     <div>

docs/get-started/agents.ttl

@@ -5,21 +5,66 @@
 
 <> a dh:Item ;
     dct:title "Connect an agent" ;
-    dct:description "Give an AI agent such as Claude Code or Cursor access to a LinkedDataHub instance over MCP, using Web-Algebra" ;
+    dct:description "Give an AI agent such as Claude Code or Cursor access to a LinkedDataHub instance, with the ldh Claude Code plugin or over MCP with Web-Algebra" ;
     rdf:_1 <#content> .
 
 <#content> a ldh:XHTML ;
     rdf:value """<div xmlns="http://www.w3.org/1999/xhtml">
-    <p class="ldh-body-lg">Give an AI agent such as Claude Code or Cursor access to a LinkedDataHub instance over MCP, using Web-Algebra</p>
-    <p>Agents use the same HTTP API as the browser and the <a href="../../reference/command-line-interface/">CLI</a>. The bridge is
-        <a href="https://github.com/AtomGraph/Web-Algebra" target="_blank">Web-Algebra</a>, which exposes LinkedDataHub operations as
-        <a href="https://modelcontextprotocol.io/" target="_blank">Model Context Protocol</a> (MCP) tools: creating containers and items, adding views,
-        queries and charts, reading and writing RDF, and running SPARQL. An agent connected this way can build a working application in a dataspace —
+    <p class="ldh-body-lg">Give an AI agent such as Claude Code or Cursor access to a LinkedDataHub instance, with the ldh Claude Code plugin or over MCP with Web-Algebra</p>
+    <p>Agents use the same HTTP API as the browser and the <a href="../../reference/command-line-interface/">CLI</a>, and act as a WebID: an agent can read and
+        write only what its WebID is authorized to. There are two ways to connect one. The <a href="#claude-code-plugin">Claude Code plugin</a> gives Claude Code
+        skills for building on LinkedDataHub with the <samp>ldh</samp> CLI. <a href="#mcp">Web-Algebra</a> exposes LinkedDataHub operations to any
+        <a href="https://modelcontextprotocol.io/" target="_blank">Model Context Protocol</a> (MCP) client as tools: creating containers and items, adding views,
+        queries and charts, reading and writing RDF, and running SPARQL. Either way, an agent can build a working application in a dataspace:
         its classes, containers, views and charts are documents like any other.</p>
     <p>The conventions agents should follow when writing to LinkedDataHub are in
-        <a href="https://github.com/AtomGraph/LinkedDataHub/blob/master/AGENTS.md" target="_blank">AGENTS.md</a>.</p>
+        <a href="https://github.com/AtomGraph/LinkedDataHub/blob/master/AGENTS.md" target="_blank">AGENTS.md</a>. To ask questions about the data in a document
+        from the browser instead, use the <a href="../../user-guide/use-the-assistant/">assistant</a>.</p>
     <div>
-        <h2 id="prerequisites">Prerequisites</h2>
+        <h2 id="claude-code-plugin">Claude Code plugin</h2>
+        <p>The <a href="https://github.com/AtomGraph/LinkedDataHub/tree/master/plugins/ldh" target="_blank"><samp>ldh</samp> plugin</a> ships with LinkedDataHub.
+            Its marketplace is the LinkedDataHub repository, so Claude Code is told about it once and then installs the plugin from it. In Claude Code:</p>
+        <pre>/plugin marketplace add AtomGraph/LinkedDataHub
+/plugin install ldh@linkeddatahub</pre>
+        <p>The skills drive the <samp>ldh</samp> CLI, so they need a LinkedDataHub instance, the owner's WebID certificate and the CLI on the <samp>PATH</samp>
+            (see <a href="../../reference/command-line-interface/#install">installing the CLI</a>). Claude loads a skill by itself when a request matches what it
+            covers. To call one explicitly, type <samp>/ldh:</samp> followed by its name.</p>
+        <table>
+            <caption class="ac-vh">Skills of the ldh plugin</caption>
+            <thead>
+                <tr>
+                    <th scope="col">Skill</th>
+                    <th scope="col">Covers</th>
+                </tr>
+            </thead>
+            <tbody>
+                <tr>
+                    <th scope="row"><samp>ldh:dataspace</samp></th>
+                    <td>Provisioning a new dataspace on a self-hosted instance: the configuration files, the restart, and checking that it is up</td>
+                </tr>
+                <tr>
+                    <th scope="row"><samp>ldh:dataset</samp></th>
+                    <td>The namespace ontology and its install, CSV and RDF imports, federation with remote SPARQL endpoints, and building a large dataset standalone and bulk-loading it</td>
+                </tr>
+                <tr>
+                    <th scope="row"><samp>ldh:app</samp></th>
+                    <td>Documents, blocks, views and charts, and an app kept as a directory and installed with <samp>ldh push</samp></td>
+                </tr>
+                <tr>
+                    <th scope="row"><samp>ldh:package</samp></th>
+                    <td>A reusable <a href="../../reference/administration/packages/">package</a>: an ontology and an XSLT stylesheet, published to a registry and imported with <samp>ldh packages add</samp></td>
+                </tr>
+            </tbody>
+        </table>
+        <p>A request can span several skills, and Claude uses them in turn:</p>
+        <blockquote><p>Here are products.csv and categories.csv. Model them with schema.org, import them into the inventory dataspace, and make each category's page list its products.</p></blockquote>
+    </div>
+    <div>
+        <h2 id="mcp">Web-Algebra over MCP</h2>
+        <p><a href="https://github.com/AtomGraph/Web-Algebra" target="_blank">Web-Algebra</a> expresses LinkedDataHub operations, SPARQL queries and Linked Data
+            reads as plans that run as one unit, and its MCP server offers them as tools. Any MCP client can use it.</p>
+    <div>
+        <h3 id="prerequisites">Prerequisites</h3>
         <ul>
             <li>A running LinkedDataHub instance. The examples below assume the default <a href="../setup/">local setup</a> at <samp>https://localhost:4443/</samp>.</li>
             <li>The WebID certificate the agent will act as. For your own instance that is the owner certificate, <samp>ssl/owner/cert.pem</samp>, with the password from
@@ -28,7 +73,7 @@
         </ul>
     </div>
     <div>
-        <h2 id="claude-code">Claude Code</h2>
+        <h3 id="claude-code">Claude Code</h3>
         <p>Register Web-Algebra as an MCP server, run from the LinkedDataHub directory:</p>
         <pre>claude mcp add web-algebra \\
   --env CERT_PEM_PATH="$PWD/ssl/owner/cert.pem" \\
@@ -37,7 +82,7 @@
         <p>Start <samp>claude</samp> and check that <samp>/mcp</samp> lists <samp>web-algebra</samp> as connected.</p>
     </div>
     <div>
-        <h2 id="cursor-and-claude-desktop">Cursor and Claude Desktop</h2>
+        <h3 id="cursor-and-claude-desktop">Cursor and Claude Desktop</h3>
         <p>Both read an <samp>mcpServers</samp> entry: <samp>~/.cursor/mcp.json</samp> for Cursor, <samp>claude_desktop_config.json</samp> for Claude Desktop.
             Use absolute paths and the literal password, since these files are not run by a shell.</p>
         <pre>{
@@ -55,7 +100,7 @@
         <p>If the client cannot find <samp>uv</samp>, replace <samp>"uv"</samp> with its absolute path, as printed by <samp>which uv</samp>.</p>
     </div>
     <div>
-        <h2 id="http-transport">Streamable HTTP transport</h2>
+        <h3 id="http-transport">Streamable HTTP transport</h3>
         <p>Clients that connect to an MCP URL instead of launching a process can use Web-Algebra's HTTP server:</p>
         <pre>CERT_PEM_PATH="$PWD/ssl/owner/cert.pem" \\
 CERT_PASSWORD="$(cat secrets/owner_cert_password.txt)" \\
@@ -70,6 +115,7 @@ uv run --with web-algebra uvicorn web_algebra.server:app</pre>
             </div>
         </div>
     </div>
+    </div>
     <div>
         <h2 id="try-it">Try it</h2>
         <p>Ask the agent for something small that it has to declare rather than code. With the <a href="https://github.com/AtomGraph/LinkedDataHub-Apps/tree/master/demo/northwind-traders" target="_blank">Northwind Traders</a> demo installed, for example:</p>

docs/reference.ttl

@@ -13,7 +13,7 @@
     rdf:value """<div xmlns="http://www.w3.org/1999/xhtml">
     <p class="ldh-body-lg">Feature-by-feature reference for the LinkedDataHub platform, its data model, HTTP API and administration</p>
     <p>The concept pages describe the <a href="dataspace/">dataspace</a>, its <a href="dataset/">dataset</a> and the <a href="data-model/">data model</a>.
-        The interface pages cover the <a href="user-interface/">user interface</a>, the <a href="http-api/">HTTP API</a> and the <a href="command-line-interface/">command line interface</a>.
+        The interface pages cover the <a href="user-interface/">user interface</a>, the <a href="http-api/">HTTP API</a>, the <a href="command-line-interface/">command line interface</a> and the <a href="assistant/">assistant</a>.
         Customization is covered by the <a href="stylesheets/">stylesheets</a> and <a href="imports/">imports</a> pages,
         and operations by the <a href="configuration/">configuration</a>, <a href="triplestores/">triplestores</a>, <a href="versioning/">versioning</a> and <a href="administration/">administration</a> pages.
         Start with the dataspace and data model pages; an application or deployment moving off 5.x starts at

docs/reference/administration/acl.ttl

@@ -14,7 +14,7 @@
     <p>LinkedDataHub access control is based on the W3C <a href="https://www.w3.org/wiki/WebAccessControl" target="_blank">ACL ontology</a>.</p>
     <div>
         <h2 id="access-modes">Access modes</h2>
-        <p>There are four access modes (classes of operation) that map to <a href="../../http-api/">HTTP methods</a>:</p>
+        <p>There are four access modes (classes of operation) that map to <a href="../../http-api/documents/#methods">HTTP methods</a>:</p>
         <table>
             <caption class="ac-vh">Access modes and the HTTP methods they permit</caption>
             <thead>