@canonry/canonry 4.158.0 → 4.159.0

package/dist/{chunk-YP4K2KXH.js → chunk-5AFWHKKI.js}

@@ -67,6 +67,7 @@ import {
   MeasurementEvidenceShapes,
   MeasurementPlanValidationError,
   MeasurementRunScopeError,
+  PROJECTS_WRITE_SCOPE,
   READ_ONLY_SCOPE,
   ResearchQueryStatuses,
   ResearchRunStatuses,
@@ -92,6 +93,8 @@ import {
   SchedulableRunKinds,
   SiteAuditTrendDirections,
   SiteCrawlFetchedStates,
+  SiteHealthLinkKinds,
+  SiteHealthTemplateDetections,
   TrafficEventConfidences,
   TrafficEventKinds,
   TrafficEvidenceKinds,
@@ -221,6 +224,7 @@ import {
   contentTargetsResponseDtoSchema,
   cosineSimilarity,
   createApiKeyRequestSchema,
+  createTemplateLinkPairIndex,
   createUserRequestSchema,
   createdApiKeyDtoSchema,
   dedupeReportActions,
@@ -313,6 +317,7 @@ import {
   isBrowserProvider,
   isReadOnlyKey,
   isRetryableHttpError,
+  isTemplateLinkRatio,
   isTrendBaseline,
   isVertexGroundingRedirect,
   isoDateDaysBeforeInTimeZone,
@@ -424,6 +429,7 @@ import {
   normalizeModelId,
   normalizeProjectAliases,
   normalizeProjectDomain,
+  normalizeProjectName,
   normalizeQueryText,
   normalizeSiteAuditRunRequest,
   normalizeUrlPath,
@@ -431,6 +437,7 @@ import {
   notFound,
   notImplemented,
   notificationDtoSchema,
+  observeTemplateLinkEdges,
   onboardingTelemetryEventSchema,
   operationInProgress,
   orderLocationsDefaultFirst,
@@ -446,6 +453,7 @@ import {
   parseWindow,
   pickClusterRepresentative,
   projectConfigSchema,
+  projectCreateRequestSchema,
   projectDtoSchema,
   projectOverviewDtoSchema,
   projectReportDtoSchema,
@@ -491,9 +499,11 @@ import {
   segmentCrawlerHits,
   serializeRunError,
   settingsDtoSchema,
+  siteAuditLivePageHealthSchema,
   siteAuditPageFactorSchema,
   siteAuditPagesResponseSchema,
   siteAuditRequestIdentity,
+  siteAuditRunProgressSchema,
   siteAuditRunRequestSchema,
   siteAuditRunResponseSchema,
   siteAuditScoreSchema,
@@ -509,10 +519,12 @@ import {
   siteCrawlStructureResponseSchema,
   siteCrawlSummarySchema,
   siteHealthChangesResponseSchema,
+  siteHealthLinkKindSchema,
   siteHealthPathResponseSchema,
   siteHealthScansResponseSchema,
   siteHealthStateSchema,
   siteHealthSubgraphResponseSchema,
+  siteHealthTemplateDetectionSchema,
   snapshotDiffResponseSchema,
   snapshotListResponseSchema,
   snapshotReportSchema,
@@ -524,6 +536,8 @@ import {
   surfaceClassFromCompetitorType,
   surfaceClassLabel,
   telemetryEventAcceptedDtoSchema,
+  templateLinkDetection,
+  templateLinkRatio,
   textContainsAnyBrandAlias,
   textContainsBrandAlias,
   textContainsDomain,
@@ -563,7 +577,7 @@ import {
   wordpressSchemaDeployResultDtoSchema,
   wordpressSchemaStatusResultDtoSchema,
   wordpressStatusDtoSchema
-} from "./chunk-DEVAEC2S.js";
+} from "./chunk-ZL3VY435.js";
 
 // src/intelligence-service.ts
 import { eq as eq58, desc as desc26, asc as asc11, and as and46, ne as ne7, or as or12, inArray as inArray20, gte as gte14, lte as lte11 } from "drizzle-orm";
@@ -1313,6 +1327,13 @@ var siteCrawlSnapshots = sqliteTable("site_crawl_snapshots", {
   deadLinkState: text("dead_link_state").notNull().default("disabled"),
   deadLinksChecked: integer("dead_links_checked").notNull().default(0),
   deadLinksFound: integer("dead_links_found").notNull().default(0),
+  /**
+   * Whether nav/header/footer link detection ran for this scan, and if not,
+   * why. NULL means the scan predates detection, which reads report as
+   * `unavailable-legacy-scan` rather than letting an empty template-link list
+   * pass for "this site has no nav".
+   */
+  templateDetection: text("template_detection"),
   createdAt: text("created_at").notNull(),
   updatedAt: text("updated_at").notNull()
 }, (table) => [
@@ -1352,6 +1373,13 @@ var siteCrawlPages = sqliteTable("site_crawl_pages", {
   canonicalNodeKey: text("canonical_node_key"),
   indexabilityState: text("indexability_state").notNull().default("unknown"),
   indexabilityReasons: text("indexability_reasons", { mode: "json" }).$type().notNull().default([]),
+  /**
+   * Derived Site Health state, computed once at publish time by the contract's
+   * `deriveSiteHealthState`. Stored so reads can filter on an index instead of
+   * materializing every page row; NULL means a snapshot published before this
+   * column existed, which reads must report rather than treat as a state.
+   */
+  healthState: text("health_state"),
   auditState: text("audit_state").notNull().default("pending"),
   auditScore: real("audit_score"),
   auditFields: text("audit_fields", { mode: "json" }).$type().notNull().default({}),
@@ -1370,6 +1398,11 @@ var siteCrawlPages = sqliteTable("site_crawl_pages", {
   index("idx_site_crawl_pages_read").on(table.projectId, table.runId, table.attemptId, table.inventoryEligible, table.auditScore, table.url),
   index("idx_site_crawl_pages_parent").on(table.projectId, table.runId, table.attemptId, table.parentPath, table.path),
   index("idx_site_crawl_pages_url").on(table.projectId, table.runId, table.attemptId, table.url),
+  index("idx_site_crawl_pages_health").on(table.projectId, table.runId, table.attemptId, table.healthState, table.path, table.nodeKey),
+  // The in-progress Page Health preview reads only completed audits below its
+  // threshold, ordered worst-first. Keep its stable tie-breaker in the index
+  // so a growing crawl never needs a transient sort before its small limit.
+  index("idx_site_crawl_pages_live_preview").on(table.projectId, table.runId, table.attemptId, table.auditState, table.auditScore, table.nodeKey),
   foreignKey({
     name: "site_crawl_pages_attempt_fk",
     columns: [table.projectId, table.runId, table.attemptId],
@@ -1393,6 +1426,29 @@ var siteCrawlEdges = sqliteTable("site_crawl_edges", {
   followableOccurrences: integer("followable_occurrences").notNull().default(1),
   nofollowOccurrences: integer("nofollow_occurrences").notNull().default(0),
   anchors: text("anchors", { mode: "json" }).$type().notNull().default([]),
+  /**
+   * True when EVERY (target page, anchor) pair on this link appears on at
+   * least `TEMPLATE_LINK_RATIO_THRESHOLD` of the attempt's fetched pages: a
+   * nav, header, or footer link rather than an editorial one. Computed once at
+   * publish time by the contract's `classifyTemplateLinks`, so the map, the
+   * API filters, and the agents all read the same decision.
+   *
+   * One row aggregates every anchor between the same two pages, so a link that
+   * is BOTH a footer link and an in-prose link is one row with two anchors. It
+   * is editorial, because the in-prose link exists.
+   *
+   * NULL means this row was never classified, which reads report as
+   * `unavailable-legacy-scan`. It never means "not a template link": the
+   * too-few-pages guard writes an explicit `false` and records its reason on
+   * the snapshot instead.
+   */
+  isTemplate: integer("is_template", { mode: "boolean" }),
+  /**
+   * Share of fetched pages carrying this link's LEAST ubiquitous anchor, which
+   * is the one that decides `is_template`. NULL when the link has no
+   * measurable pair (an unresolved target or no anchor).
+   */
+  templateRatio: real("template_ratio"),
   createdAt: text("created_at").notNull(),
   updatedAt: text("updated_at").notNull()
 }, (table) => [
@@ -1401,6 +1457,17 @@ var siteCrawlEdges = sqliteTable("site_crawl_edges", {
   index("idx_site_crawl_edges_inbound").on(table.projectId, table.runId, table.attemptId, table.targetNodeKey, table.edgeKey),
   index("idx_site_crawl_edges_source_url").on(table.projectId, table.runId, table.attemptId, table.sourceUrl),
   index("idx_site_crawl_edges_target_url").on(table.projectId, table.runId, table.attemptId, table.targetUrl),
+  // Backs the `linkKind` filter on the internal-link reads, whose ORDER BY is
+  // the trailing edge key, so a filtered page terminates at LIMIT instead of
+  // sorting every match in a temp b-tree.
+  index("idx_site_crawl_edges_template").on(
+    table.projectId,
+    table.runId,
+    table.attemptId,
+    table.internal,
+    table.isTemplate,
+    table.edgeKey
+  ),
   // Publish-time graph sampling filters this exact scope and consumes the
   // highest-occurrence internal anchors first. The mixed-direction suffix
   // avoids a million-row temporary ORDER BY at the maximum crawl budget.
@@ -1429,8 +1496,17 @@ var siteCrawlGraphLayouts = sqliteTable("site_crawl_graph_layouts", {
   failureCode: text("failure_code"),
   totalNodes: integer("total_nodes").notNull().default(0),
   totalEdges: integer("total_edges").notNull().default(0),
+  /** Share of `totalEdges` classified as nav, header, or footer links. */
+  totalTemplateEdges: integer("total_template_edges").notNull().default(0),
   nodeCount: integer("node_count").notNull().default(0),
   edgeCount: integer("edge_count").notNull().default(0),
+  /**
+   * True when template links were kept out of the ForceAtlas2 physics, so
+   * these positions describe content structure. Layouts published before
+   * template detection keep `false`: the migration classifies their links, but
+   * it deliberately does not rewrite immutable coordinates.
+   */
+  templateLinksExcluded: integer("template_links_excluded", { mode: "boolean" }).notNull().default(false),
   createdAt: text("created_at").notNull(),
   updatedAt: text("updated_at").notNull()
 }, (table) => [
@@ -1480,6 +1556,13 @@ var siteCrawlGraphEdges = sqliteTable("site_crawl_graph_edges", {
   targetNodeKey: text("target_node_key").notNull(),
   followable: integer("followable", { mode: "boolean" }).notNull(),
   occurrences: integer("occurrences").notNull(),
+  /**
+   * Denormalized from `site_crawl_edges` the same way `followable` and
+   * `occurrences` are, so a bounded map read never joins the full link table.
+   * Template links stay IN the sample: the map hides them by default, and a
+   * viewer switching them on must not trigger a refetch or a re-layout.
+   */
+  isTemplate: integer("is_template", { mode: "boolean" }).notNull().default(false),
   createdAt: text("created_at").notNull()
 }, (table) => [
   uniqueIndex("idx_site_crawl_graph_edges_attempt_edge").on(table.projectId, table.runId, table.attemptId, table.edgeKey),
@@ -2867,6 +2950,138 @@ function relinkOrphanedSnapshotQueryIds(tx) {
     }
   }
 }
+function dropSiteCrawlSelfLinks(tx) {
+  tx.run(sql2.raw("DELETE FROM site_crawl_edges WHERE source_url = target_url"));
+  tx.run(sql2.raw("DELETE FROM site_crawl_graph_edges WHERE source_node_key = target_node_key"));
+}
+function backfillSiteCrawlPageHealthState(tx) {
+  const BATCH = 500;
+  for (; ; ) {
+    const rows = tx.all(sql2.raw(`
+      SELECT id, node_key, fetch_state, indexability_state, indexability_reasons, canonical_node_key
+      FROM site_crawl_pages
+      WHERE health_state IS NULL
+      LIMIT ${BATCH}
+    `));
+    if (rows.length === 0) return;
+    for (const row of rows) {
+      const healthState = deriveSiteHealthState({
+        fetchState: row.fetch_state,
+        indexabilityState: row.indexability_state,
+        indexabilityReasons: parseJsonColumn(row.indexability_reasons, []),
+        canonicalNodeKey: row.canonical_node_key,
+        nodeKey: row.node_key
+      });
+      tx.run(sql2`UPDATE site_crawl_pages SET health_state = ${healthState} WHERE id = ${row.id}`);
+    }
+    if (rows.length < BATCH) return;
+  }
+}
+function* streamSiteCrawlTemplateLinkEdges(tx, attemptId, batchSize = 2e3) {
+  let after = "";
+  for (; ; ) {
+    const rows = tx.all(sql2`
+      SELECT edge_key, source_node_key, target_node_key, anchors
+      FROM site_crawl_edges
+      WHERE attempt_id = ${attemptId} AND edge_key > ${after}
+      ORDER BY edge_key
+      LIMIT ${batchSize}
+    `);
+    if (rows.length === 0) return;
+    yield rows.map((row) => ({
+      edgeKey: row.edge_key,
+      sourceNodeKey: row.source_node_key,
+      targetNodeKey: row.target_node_key,
+      anchors: parseJsonColumn(row.anchors, [])
+    }));
+    after = rows[rows.length - 1].edge_key;
+    if (rows.length < batchSize) return;
+  }
+}
+function* chunked(values, size = 500) {
+  for (let offset = 0; offset < values.length; offset += size) {
+    yield values.slice(offset, offset + size);
+  }
+}
+function backfillSiteCrawlTemplateLinks(tx) {
+  const attempts = tx.all(sql2.raw(`
+    SELECT
+      a.id                AS attempt_id,
+      a.project_id        AS project_id,
+      a.run_id            AS run_id,
+      COALESCE(NULLIF(a.pages_fetched, 0), s.pages_fetched, 0) AS pages_fetched
+    FROM site_crawl_attempts AS a
+    LEFT JOIN site_crawl_snapshots AS s
+      ON s.project_id = a.project_id AND s.run_id = a.run_id AND s.attempt_id = a.id
+  `));
+  for (const attempt of attempts) {
+    const detection = templateLinkDetection(attempt.pages_fetched);
+    const scope = { attemptId: attempt.attempt_id, projectId: attempt.project_id, runId: attempt.run_id };
+    tx.run(sql2`
+      UPDATE site_crawl_edges SET is_template = 0, template_ratio = NULL
+      WHERE project_id = ${scope.projectId} AND run_id = ${scope.runId} AND attempt_id = ${scope.attemptId}
+    `);
+    tx.run(sql2`
+      UPDATE site_crawl_graph_edges SET is_template = 0
+      WHERE project_id = ${scope.projectId} AND run_id = ${scope.runId} AND attempt_id = ${scope.attemptId}
+    `);
+    if (detection === SiteHealthTemplateDetections.applied) {
+      const index2 = createTemplateLinkPairIndex();
+      for (const batch of streamSiteCrawlTemplateLinkEdges(tx, scope.attemptId)) {
+        observeTemplateLinkEdges(index2, batch);
+      }
+      const keysByRatio = /* @__PURE__ */ new Map();
+      const templateEdgeKeys = [];
+      for (const batch of streamSiteCrawlTemplateLinkEdges(tx, scope.attemptId)) {
+        for (const edge of batch) {
+          const ratio = templateLinkRatio(index2, attempt.pages_fetched, edge);
+          if (ratio == null) continue;
+          const group = keysByRatio.get(ratio);
+          if (group) group.push(edge.edgeKey);
+          else keysByRatio.set(ratio, [edge.edgeKey]);
+          if (isTemplateLinkRatio(ratio)) templateEdgeKeys.push(edge.edgeKey);
+        }
+      }
+      for (const [ratio, edgeKeys] of keysByRatio) {
+        const isTemplate = isTemplateLinkRatio(ratio) ? 1 : 0;
+        for (const chunk of chunked(edgeKeys)) {
+          tx.run(sql2`
+            UPDATE site_crawl_edges SET is_template = ${isTemplate}, template_ratio = ${ratio}
+            WHERE project_id = ${scope.projectId} AND run_id = ${scope.runId} AND attempt_id = ${scope.attemptId}
+              AND edge_key IN (SELECT value FROM json_each(${JSON.stringify(chunk)}))
+          `);
+        }
+      }
+      for (const chunk of chunked(templateEdgeKeys)) {
+        tx.run(sql2`
+          UPDATE site_crawl_graph_edges SET is_template = 1
+          WHERE project_id = ${scope.projectId} AND run_id = ${scope.runId} AND attempt_id = ${scope.attemptId}
+            AND edge_key IN (SELECT value FROM json_each(${JSON.stringify(chunk)}))
+        `);
+      }
+    }
+    tx.run(sql2`
+      UPDATE site_crawl_graph_layouts
+      SET total_template_edges = (
+        SELECT COUNT(*)
+        FROM site_crawl_edges AS e
+        INNER JOIN site_crawl_pages AS source_page
+          ON source_page.project_id = e.project_id AND source_page.run_id = e.run_id
+          AND source_page.attempt_id = e.attempt_id AND source_page.node_key = e.source_node_key
+        INNER JOIN site_crawl_pages AS target_page
+          ON target_page.project_id = e.project_id AND target_page.run_id = e.run_id
+          AND target_page.attempt_id = e.attempt_id AND target_page.node_key = e.target_node_key
+        WHERE e.project_id = ${scope.projectId} AND e.run_id = ${scope.runId} AND e.attempt_id = ${scope.attemptId}
+          AND e.internal = 1 AND e.relation = 'anchor' AND e.is_template = 1
+      )
+      WHERE project_id = ${scope.projectId} AND run_id = ${scope.runId} AND attempt_id = ${scope.attemptId}
+    `);
+    tx.run(sql2`
+      UPDATE site_crawl_snapshots SET template_detection = ${detection}
+      WHERE project_id = ${scope.projectId} AND run_id = ${scope.runId} AND attempt_id = ${scope.attemptId}
+    `);
+  }
+}
 function backfillQuerySnapshotServedModel(tx) {
   tx.run(sql2.raw(`
     UPDATE query_snapshots
@@ -5809,6 +6024,127 @@ var MIGRATION_VERSIONS = [
       `CREATE INDEX IF NOT EXISTS idx_traffic_event_receipts_expires
         ON traffic_event_receipts(source_id, expires_at)`
     ]
+  },
+  {
+    version: 130,
+    name: "site-crawl-pages-persisted-health-state",
+    // Site Health state is derived from fetch state, indexability, the
+    // crawler's reasons, and canonical identity together. Filtering it by
+    // recomputing in JS meant reading every page row on every request. It is
+    // now written once at publish time by the SAME contract function the map
+    // and the agents use, so reads are an ordinary indexed WHERE.
+    //
+    // Backfilled in TypeScript inside this transaction (see
+    // `backfillSiteCrawlPageHealthState`) so existing scans are filterable
+    // immediately. The derivation is never rewritten in SQL: the backfill
+    // calls the contract's own function, so there is no second implementation
+    // to drift. `unavailable-legacy-scan` therefore survives only as the
+    // honest answer for a row this could not reach.
+    //
+    // Idempotent: `ALTER TABLE ADD COLUMN` errors with "duplicate column
+    // name" on retry, which the runner already swallows, and the backfill
+    // only touches NULL rows.
+    statements: [
+      `ALTER TABLE site_crawl_pages ADD COLUMN health_state TEXT`,
+      // Trailing columns match the ORDER BY the page list actually issues
+      // (`sort=path`), so a filtered read terminates at LIMIT instead of
+      // sorting every match in a temp b-tree on every cursor page.
+      `CREATE INDEX IF NOT EXISTS idx_site_crawl_pages_health
+        ON site_crawl_pages(project_id, run_id, attempt_id, health_state, path, node_key)`
+    ],
+    run: backfillSiteCrawlPageHealthState
+  },
+  {
+    version: 131,
+    name: "site-crawl-template-link-classification",
+    // A real site's internal links are bimodal: a (target page, anchor) pair
+    // sits on nearly every page or on nearly none. The ubiquitous half is nav,
+    // header, and footer chrome, and drawing it buries the content structure
+    // the map exists to show. Classification is computed once at publish time
+    // by the contract's own function, so the map, the API filters, and the
+    // agents cannot disagree about which links are chrome.
+    //
+    // `is_template` and `template_detection` are deliberately nullable:
+    // NULL means "never classified" and reads report it as
+    // `unavailable-legacy-scan`. The too-few-pages guard writes an explicit
+    // `false` plus its reason on the snapshot, so an empty template-link list
+    // can never pass for "this site has no nav".
+    //
+    // Idempotent: `ALTER TABLE ADD COLUMN` errors with "duplicate column
+    // name", which the runner already swallows, and the backfill reclassifies
+    // each attempt from its own stored rows.
+    statements: [
+      `ALTER TABLE site_crawl_edges ADD COLUMN is_template INTEGER`,
+      `ALTER TABLE site_crawl_edges ADD COLUMN template_ratio REAL`,
+      `ALTER TABLE site_crawl_graph_edges ADD COLUMN is_template INTEGER NOT NULL DEFAULT 0`,
+      `ALTER TABLE site_crawl_graph_layouts ADD COLUMN total_template_edges INTEGER NOT NULL DEFAULT 0`,
+      `ALTER TABLE site_crawl_graph_layouts ADD COLUMN template_links_excluded INTEGER NOT NULL DEFAULT 0`,
+      `ALTER TABLE site_crawl_snapshots ADD COLUMN template_detection TEXT`,
+      // Trailing edge key matches the ORDER BY the internal-link reads issue,
+      // so a filtered page terminates at LIMIT.
+      `CREATE INDEX IF NOT EXISTS idx_site_crawl_edges_template
+        ON site_crawl_edges(project_id, run_id, attempt_id, internal, is_template, edge_key)`
+    ],
+    run: backfillSiteCrawlTemplateLinks
+  },
+  {
+    version: 132,
+    name: "site-crawl-drop-self-links",
+    // A page linking to itself is not a link to or from another page, and the
+    // crawl engine already excludes self-loops from a page's inbound and
+    // outbound metrics. The edge tables kept them, so the stored edges
+    // disagreed with the page rows produced by the same crawl: a self-loop
+    // appeared in BOTH neighbour lists, so every self-linking page read one
+    // higher in each direction than its own tiles.
+    //
+    // The writer now drops them, which fixes new scans. This clears the ones
+    // already stored so an existing scan is correct without a re-crawl.
+    // Comparing the stored normalized URLs is exactly the rule the writer
+    // applies (`isSelfLink`), expressed as the equality SQL can do without
+    // reimplementing any derivation.
+    //
+    // Idempotent: a re-run deletes nothing once the rows are gone.
+    statements: [],
+    run: dropSiteCrawlSelfLinks
+  },
+  {
+    version: 133,
+    name: "site-crawl-reclassify-template-links-per-anchor",
+    // v131 classified a link as chrome when its MOST ubiquitous anchor was
+    // ubiquitous. One stored link row aggregates every anchor the crawl saw
+    // between the same two pages, so on any site with a comprehensive footer
+    // (which is most sites) an in-prose link to a footer-linked target shared
+    // a row with that page's footer link and inherited its ratio. The
+    // editorial link was marked chrome, hidden from the map, and dropped from
+    // every content count. Observed on canonry.ai: /aeo-methodology had 24
+    // outbound links, all marked chrome at ratio 0.96, with anchors like
+    // "Explore the Canonry platform" that are plainly editorial.
+    //
+    // The rule is now the LEAST ubiquitous anchor: chrome only when every
+    // anchor on the link is chrome. This re-runs the same backfill hook, which
+    // reclassifies each attempt from its own stored rows through the corrected
+    // contract function, so an existing scan is right without a re-crawl.
+    //
+    // Statements stay empty because v131 already added every column; only the
+    // values were wrong. Re-running is safe: the hook resets each attempt to
+    // "classified, not chrome" before it writes, so it is idempotent and its
+    // result depends on nothing but the stored links.
+    statements: [],
+    run: backfillSiteCrawlTemplateLinks
+  },
+  {
+    version: 134,
+    name: "site-crawl-live-page-health-preview-index",
+    // The live Page Health preview is deliberately small and stable while a
+    // crawl writes more pages: completed audits below the display threshold,
+    // ordered worst-first by score and then node key. Cover the exact WHERE +
+    // ORDER BY shape so the read stops at LIMIT without sorting every current
+    // crawl row in a temporary b-tree. Its standard composite-index SQL is
+    // also PostgreSQL-compatible; the query shape has no SQLite-only syntax.
+    statements: [
+      `CREATE INDEX IF NOT EXISTS idx_site_crawl_pages_live_preview
+        ON site_crawl_pages(project_id, run_id, attempt_id, audit_state, audit_score, node_key)`
+    ]
   }
 ];
 function addRunsMeasurementPlanVersionForeignKey(tx) {
@@ -8680,8 +9016,14 @@ function isOverviewRead(url) {
     "citations/visibility"
   ])).has(rest);
 }
-function isTechnicalAeoRead(url) {
+function isTechnicalAeoRead(request, url) {
   const rest = projectRouteRest(url);
+  if (rest === "runs") {
+    return queryValue(request, "kind") === RunKinds["site-audit"];
+  }
+  if (rest && /^technical-aeo\/runs\/[^/]+\/(?:progress|page-health-preview)$/.test(rest)) {
+    return true;
+  }
   return rest === "technical-aeo" || rest === "technical-aeo/pages" || rest === "technical-aeo/trend" || rest === "technical-aeo/crawl" || rest === "technical-aeo/graph" || rest === "technical-aeo/subgraph" || rest === "technical-aeo/path" || rest === "technical-aeo/changes" || rest === "technical-aeo/crawl/pages/audit" || rest === "technical-aeo/crawl/pages" || rest === "technical-aeo/runs" || rest === "technical-aeo/structure" || rest === "technical-aeo/internal-links" || rest === "technical-aeo/internal-links/neighbors" || rest === "technical-aeo/dead-links";
 }
 function isReportRead(url) {
@@ -8698,7 +9040,7 @@ function enforceEmbedProjectTabs(request, configuredTabs) {
   const url = request.url.split("?")[0];
   if (isProjectShellRead(request, url)) return;
   if (tabs.includes("overview") && isOverviewRead(url)) return;
-  if (tabs.includes("technical-aeo") && isTechnicalAeoRead(url)) return;
+  if (tabs.includes("technical-aeo") && isTechnicalAeoRead(request, url)) return;
   if (tabs.includes("report") && isReportRead(url)) return;
   throw forbidden("This endpoint is not available for the configured embed tabs.");
 }
@@ -9047,6 +9389,104 @@ function pruneProviderModelsForProviders(models, providers) {
 
 // ../api-routes/src/projects.ts
 async function projectRoutes(app, opts) {
+  app.post("/projects", async (request, reply) => {
+    requireAdminSession(request);
+    if (request.principal?.projectId) {
+      throw forbidden("This API key is limited to one project, and cannot create projects for this install.");
+    }
+    requireScope(request, PROJECTS_WRITE_SCOPE);
+    const parsedBody = projectCreateRequestSchema.safeParse(request.body);
+    if (!parsedBody.success) {
+      throw validationError("Invalid project payload", {
+        issues: parsedBody.error.issues.map((issue) => ({
+          path: issue.path.join("."),
+          message: issue.message
+        }))
+      });
+    }
+    const body = parsedBody.data;
+    const name = normalizeProjectName(body.name);
+    if (!name) {
+      throw validationError("Project name must contain at least one letter or number after normalization.");
+    }
+    const canonicalDomain = hostOf(body.canonicalDomain);
+    if (!canonicalDomain) {
+      throw validationError("canonicalDomain must be a valid hostname or http(s) URL.");
+    }
+    const validNames = opts.providerAdapters?.map((adapter) => adapter.name) ?? [];
+    if (validNames.length && body.providers?.length) {
+      const invalid = body.providers.filter((p) => !validNames.includes(p));
+      if (invalid.length) {
+        throw validationError(`Invalid provider(s): ${invalid.join(", ")}. Must be one of: ${validNames.join(", ")}`, {
+          invalidProviders: invalid,
+          validProviders: validNames
+        });
+      }
+    }
+    const nextProviders = body.providers ?? [];
+    const providerModels = pruneProviderModelsForProviders(
+      validateProviderModels(body.providerModels ?? {}, opts.providerAdapters),
+      nextProviders
+    );
+    assertProviderModelScope(request, {}, providerModels, nextProviders);
+    const nextLocations = body.locations ?? [];
+    const duplicateLabels = findDuplicateLocationLabels(nextLocations);
+    if (duplicateLabels.length > 0) {
+      throw validationError(`Duplicate location labels are not allowed: ${duplicateLabels.join(", ")}`, {
+        duplicateLabels
+      });
+    }
+    const nextDefaultLocation = body.defaultLocation ?? null;
+    if (!hasLocationLabel(nextLocations, nextDefaultLocation)) {
+      throw validationError(`defaultLocation "${nextDefaultLocation}" must match a configured location label`, {
+        defaultLocation: nextDefaultLocation
+      });
+    }
+    const normalizedCollision = app.db.select({
+      name: projects.name
+    }).from(projects).all().find((project) => normalizeProjectName(project.name) === name);
+    if (normalizedCollision) throw alreadyExists("Project", name);
+    const id = crypto6.randomUUID();
+    const now = (/* @__PURE__ */ new Date()).toISOString();
+    const nextAliases = normalizeProjectAliases(body.displayName, body.aliases ?? []);
+    const inserted = app.db.transaction((tx) => {
+      const result = tx.insert(projects).values({
+        id,
+        name,
+        displayName: body.displayName,
+        canonicalDomain,
+        ownedDomains: body.ownedDomains ?? [],
+        aliases: nextAliases,
+        country: body.country,
+        language: body.language,
+        tags: body.tags ?? [],
+        labels: body.labels ?? {},
+        providers: nextProviders,
+        providerModels,
+        measurement: body.measurement ?? DEFAULT_MEASUREMENT_CONFIG,
+        locations: nextLocations,
+        defaultLocation: nextDefaultLocation,
+        autoExtractBacklinks: body.autoExtractBacklinks ?? false,
+        configSource: body.configSource ?? "api",
+        configRevision: 1,
+        createdAt: now,
+        updatedAt: now
+      }).onConflictDoNothing().run();
+      if (result.changes !== 1) return false;
+      writeAuditLog(tx, {
+        projectId: id,
+        actor: "api",
+        action: "project.created",
+        entityType: "project",
+        entityId: id
+      });
+      return true;
+    });
+    if (!inserted) throw alreadyExists("Project", name);
+    opts.onProjectUpserted?.(id, name);
+    const created = app.db.select().from(projects).where(eq4(projects.id, id)).get();
+    return reply.status(201).send(formatProject(created));
+  });
   app.put("/projects/:name", async (request, reply) => {
     const { name } = request.params;
     const parsedBody = projectUpsertRequestSchema.safeParse(request.body);
@@ -27389,6 +27829,7 @@ var SCHEMA_TABLE = {
   NotificationDto: notificationDtoSchema,
   OnboardingTelemetryEvent: onboardingTelemetryEventSchema,
   OrganicEvidenceDto: organicEvidenceDtoSchema,
+  ProjectCreateRequest: projectCreateRequestSchema,
   ProjectDto: projectDtoSchema,
   ProjectUpsertRequest: projectUpsertRequestSchema,
   ProjectConfig: projectConfigSchema,
@@ -27402,7 +27843,9 @@ var SCHEMA_TABLE = {
   SchedulableRunKind: schedulableRunKindSchema,
   ScheduleDto: scheduleDtoSchema,
   SettingsDto: settingsDtoSchema,
+  SiteAuditLivePageHealthDto: siteAuditLivePageHealthSchema,
   SiteAuditPagesResponseDto: siteAuditPagesResponseSchema,
+  SiteAuditRunProgressDto: siteAuditRunProgressSchema,
   SiteAuditRunResponseDto: siteAuditRunResponseSchema,
   SiteAuditScoreDto: siteAuditScoreSchema,
   SiteAuditTrendResponseDto: siteAuditTrendResponseSchema,
@@ -27890,6 +28333,12 @@ var crawlLimitParameter = {
   description: "Page size. Defaults to 100 (50 for structure/neighbors); maximum 200 (100 for structure/neighbors).",
   schema: { type: "integer", minimum: 1, maximum: 200 }
 };
+var linkKindParameter = {
+  name: "linkKind",
+  in: "query",
+  description: 'Restrict links to `content` (excludes nav, header, and footer links) or `template` (only those). Defaults to `all`. A scan whose links were never classified matches neither, which is why every link response also carries `templateDetection`: an empty `content` list means "could not tell" when detection is not `applied`.',
+  schema: { type: "string", enum: ["all", "content", "template"], default: "all" }
+};
 var measurementOverviewSortParameter = {
   name: "sort",
   in: "query",
@@ -28031,6 +28480,27 @@ var routeCatalog = [
       404: errorResponse("Project or connection not found.")
     }
   },
+  {
+    method: "post",
+    path: "/api/v1/projects",
+    summary: "Create a project",
+    description: "Creates a new project with create-only semantics for the domain-first launchpad. The server normalizes the project name and canonical domain; a normalized-name collision returns 409 and never updates the existing project.",
+    tags: ["projects"],
+    requestBody: {
+      required: true,
+      content: {
+        "application/json": {
+          schema: { $ref: "#/components/schemas/ProjectCreateRequest" }
+        }
+      }
+    },
+    responses: {
+      201: jsonResponse("Project created.", "ProjectDto"),
+      400: errorResponse("Invalid project payload, normalized name, or canonical domain."),
+      403: errorResponse("The caller lacks broad projects.write authority."),
+      409: errorResponse("A project with the normalized name already exists.")
+    }
+  },
   {
     method: "put",
     path: "/api/v1/projects/{name}",
@@ -33408,13 +33878,14 @@ var routeCatalog = [
     method: "get",
     path: "/api/v1/projects/{name}/technical-aeo/graph",
     summary: "Get a persisted Site Health graph projection",
-    description: "Returns the deterministic graph projection computed once when the latest complete or selected historical crawl was published. ForceAtlas2 positions and the exact internal-anchor edge sample are persisted, so reads run no layout physics and never rescan the crawl edge table. The response is bounded to 20,000 nodes / 50,000 edges; `layout`, `omittedNodes`, `omittedEdges`, and `sampled` disclose legacy/unavailable layouts and intentional truncation.",
+    description: "Returns the deterministic graph projection computed once when the latest complete or selected historical crawl was published. ForceAtlas2 positions and the exact internal-anchor edge sample are persisted, so reads run no layout physics and never rescan the crawl edge table. Nav, header, and footer links are excluded from the layout physics but retained in the sample and tagged `isTemplate`, so a viewer can draw them without a refetch and without any node moving. The response is bounded to 20,000 nodes / 50,000 edges; `layout`, `omittedNodes`, `omittedEdges`, and `sampled` disclose legacy/unavailable layouts and intentional truncation, and `templateDetection` says whether template links could be told apart at all.",
     tags: ["technical-aeo"],
     parameters: [
       nameParameter,
       { name: "runId", in: "query", description: "Historical site-audit run ID. Omit for the latest complete crawl.", schema: stringSchema },
       { name: "maxNodes", in: "query", description: "Maximum graph nodes. Defaults to and is capped at 20,000.", schema: { type: "integer", minimum: 1, maximum: 2e4 } },
-      { name: "maxEdges", in: "query", description: "Maximum graph edges. Defaults to and is capped at 50,000.", schema: { type: "integer", minimum: 1, maximum: 5e4 } }
+      { name: "maxEdges", in: "query", description: "Maximum graph edges. Defaults to and is capped at 50,000.", schema: { type: "integer", minimum: 1, maximum: 5e4 } },
+      linkKindParameter
     ],
     responses: {
       200: jsonResponse("Bounded Site Health graph projection returned.", "SiteCrawlGraphResponseDto"),
@@ -33513,7 +33984,8 @@ var routeCatalog = [
       { name: "inventoryEligible", in: "query", description: "Filter Canonry technical inventory eligibility (`true` or `false`).", schema: booleanSchema },
       { name: "fetchState", in: "query", description: "Filter by persisted fetch state.", schema: stringSchema },
       { name: "indexabilityState", in: "query", description: "Filter by crawler-derived indexability state; this is not Google index coverage.", schema: stringSchema },
-      { name: "healthState", in: "query", description: "Filter by the derived Site Health state shared with the dashboard and agents. Unlike `indexabilityState` this folds fetch state, canonical identity, and the crawler reasons into one decision, so `hidden` also covers redirects, robots blocks, non-HTML, and canonical-away pages. Values: `eligible`, `hidden`, `failed`, `unchecked`.", schema: { type: "string", enum: ["eligible", "hidden", "failed", "unchecked"] } },
+      { name: "nodeKey", in: "query", description: "Return only the page with this node key. Combined with a filter it answers whether that exact page is in the filtered set, without paging through the list.", schema: stringSchema },
+      { name: "healthState", in: "query", description: "Filter by the derived Site Health state shared with the dashboard and agents. Unlike `indexabilityState` this folds fetch state, canonical identity, and the crawler reasons into one decision. `hidden` means the SITE told answer engines not to index the page (noindex, canonical-away, robots.txt); a non-HTML file is `resource` and a redirect is `redirect`, because neither is suppressed. Values: `eligible`, `hidden`, `resource`, `redirect`, `failed`, `unchecked`.", schema: { type: "string", enum: ["eligible", "hidden", "resource", "redirect", "failed", "unchecked"] } },
       { name: "auditState", in: "query", description: "Filter by audit state.", schema: stringSchema },
       { name: "sort", in: "query", description: "Sort order. Defaults to `url`.", schema: { type: "string", enum: ["url", "path", "score-asc", "score-desc"] } },
       crawlCursorParameter,
@@ -33546,7 +34018,7 @@ var routeCatalog = [
     method: "get",
     path: "/api/v1/projects/{name}/technical-aeo/internal-links",
     summary: "List persisted internal crawl links",
-    description: "Cursor-paged internal edges for the latest or selected crawl. Optional source/target/followability filters remain project-, run-, and attempt-scoped.",
+    description: "Cursor-paged internal edges for the latest or selected crawl. Optional source/target/followability/link-kind filters remain project-, run-, and attempt-scoped. `total` counts exactly what the requested filters match, and `templateDetection` says whether nav and footer links could be told apart for this scan.",
     tags: ["technical-aeo"],
     parameters: [
       nameParameter,
@@ -33554,6 +34026,7 @@ var routeCatalog = [
       { name: "sourceUrl", in: "query", description: "Restrict to a source URL.", schema: stringSchema },
       { name: "targetUrl", in: "query", description: "Restrict to a target URL.", schema: stringSchema },
       { name: "followable", in: "query", description: "Restrict to followable or nofollow link observations.", schema: booleanSchema },
+      linkKindParameter,
       crawlCursorParameter,
       crawlLimitParameter
     ],
@@ -33573,6 +34046,7 @@ var routeCatalog = [
       { name: "runId", in: "query", description: "Historical site-audit run ID. Omit for the latest crawl.", schema: stringSchema },
       { name: "nodeKey", in: "query", description: "Canonical crawl node key.", schema: stringSchema },
       { name: "url", in: "query", description: "Canonical crawl URL.", schema: stringSchema },
+      linkKindParameter,
       { name: "limit", in: "query", description: "Maximum inbound and outbound edges independently. Defaults to 50; maximum 100.", schema: { type: "integer", minimum: 1, maximum: 100 } }
     ],
     responses: {
@@ -33613,6 +34087,36 @@ var routeCatalog = [
       404: errorResponse("Project not found.")
     }
   },
+  {
+    method: "get",
+    path: "/api/v1/projects/{name}/technical-aeo/runs/{runId}/progress",
+    summary: "Get exact stored Site Health run progress",
+    description: "Returns durable run, crawl-attempt, and terminal graph-layout state for one exact non-probe site-audit run. It performs no network work and intentionally returns raw counters instead of a completion percentage.",
+    tags: ["technical-aeo"],
+    parameters: [
+      nameParameter,
+      { name: "runId", in: "path", required: true, description: "Exact site-audit run ID.", schema: stringSchema }
+    ],
+    responses: {
+      200: jsonResponse("Exact stored Site Health progress returned.", "SiteAuditRunProgressDto"),
+      404: errorResponse("Project or non-probe site-audit run not found.")
+    }
+  },
+  {
+    method: "get",
+    path: "/api/v1/projects/{name}/technical-aeo/runs/{runId}/page-health-preview",
+    summary: "Get a bounded, provisional Page Health preview",
+    description: "Returns a small durable, provisional preview for one exact non-probe site-audit run while it is running: the count of audited pages plus at most 12 actionable low-score examples without finding prose. Queued scans return waiting. Terminal scans return no examples so callers use the immutable Page Health results instead. It performs no network work.",
+    tags: ["technical-aeo"],
+    parameters: [
+      nameParameter,
+      { name: "runId", in: "path", required: true, description: "Exact site-audit run ID.", schema: stringSchema }
+    ],
+    responses: {
+      200: jsonResponse("Bounded provisional Page Health preview returned.", "SiteAuditLivePageHealthDto"),
+      404: errorResponse("Project or non-probe site-audit run not found.")
+    }
+  },
   {
     method: "post",
     path: "/api/v1/projects/{name}/technical-aeo/runs",
@@ -33642,6 +34146,7 @@ var routeCatalog = [
       200: jsonResponse("Site-audit run queued (or the in-flight run returned).", "SiteAuditRunResponseDto"),
       400: errorResponse("Invalid site-audit request."),
       409: errorResponse("A crawl with different effective options is already queued or running."),
+      422: errorResponse("The site-audit executor is unavailable on this deployment."),
       404: errorResponse("Project not found.")
     }
   }
@@ -57062,7 +57567,7 @@ function dedupeStrings(input) {
 
 // ../api-routes/src/technical-aeo.ts
 import crypto44 from "crypto";
-import { and as and44, asc as asc10, count, desc as desc24, eq as eq56, inArray as inArray19, isNotNull as isNotNull3, lt as lt8, or as or11, sql as sql21 } from "drizzle-orm";
+import { and as and44, asc as asc10, count, desc as desc24, eq as eq56, inArray as inArray19, isNotNull as isNotNull3, isNull as isNull5, lt as lt8, or as or11, sql as sql21 } from "drizzle-orm";
 import { alias } from "drizzle-orm/sqlite-core";
 var FETCHED_SITE_CRAWL_STATES = /* @__PURE__ */ new Set([
   ...SiteCrawlFetchedStates,
@@ -57072,6 +57577,22 @@ var FETCHED_SITE_CRAWL_STATES = /* @__PURE__ */ new Set([
   "fetched"
 ]);
 var SURFACEABLE_STATUSES = [RunStatuses.completed, RunStatuses.partial];
+function siteAuditRunPhase(status, attempt, layoutState) {
+  switch (status) {
+    case RunStatuses.queued:
+      return "queued";
+    case RunStatuses.running:
+      return (attempt?.pagesFetched ?? 0) > 0 ? "checking" : "discovering";
+    case RunStatuses.completed:
+      return attempt && layoutState === "pending" ? "arranging-map" : "completed";
+    case RunStatuses.partial:
+      return attempt && layoutState === "pending" ? "arranging-map" : "partial";
+    case RunStatuses.failed:
+      return "failed";
+    case RunStatuses.cancelled:
+      return "cancelled";
+  }
+}
 function emptyScore(projectName) {
   return {
     project: projectName,
@@ -57199,6 +57720,21 @@ function mapCrawlPageAuditEvidence(row) {
     criticalDefects: parsedCriticalDefects.flatMap((result) => result.success ? [result.data] : [])
   };
 }
+var LIVE_PAGE_HEALTH_CANDIDATE_LIMIT = 48;
+var LIVE_PAGE_HEALTH_EXAMPLE_LIMIT = 12;
+function countLivePageHealthChecks(auditFields) {
+  const rawFactors = Array.isArray(auditFields.factors) ? auditFields.factors : [];
+  const rawCriticalDefects = Array.isArray(auditFields.criticalDefects) ? auditFields.criticalDefects : [];
+  let checks = 0;
+  for (const rawFactor of rawFactors) {
+    const factor = siteCrawlAuditFactorSchema.safeParse(rawFactor);
+    if (factor.success && factor.data.applicable !== false && factor.data.status !== "pass") checks++;
+  }
+  for (const rawCriticalDefect of rawCriticalDefects) {
+    if (siteCrawlCriticalDefectSchema.safeParse(rawCriticalDefect).success) checks++;
+  }
+  return checks;
+}
 function mapCrawlEdge(row) {
   return {
     edgeKey: row.edgeKey,
@@ -57212,9 +57748,28 @@ function mapCrawlEdge(row) {
     occurrences: row.occurrences,
     followableOccurrences: row.followableOccurrences,
     nofollowOccurrences: row.nofollowOccurrences,
-    anchors: row.anchors
+    anchors: row.anchors,
+    isTemplate: row.isTemplate,
+    templateRatio: row.templateRatio
   };
 }
+function templateDetectionOf(value) {
+  const parsed = siteHealthTemplateDetectionSchema.safeParse(value);
+  return parsed.success ? parsed.data : SiteHealthTemplateDetections["unavailable-legacy-scan"];
+}
+function parseLinkKind(value) {
+  if (value === void 0 || value === "") return SiteHealthLinkKinds.all;
+  const allowed = siteHealthLinkKindSchema.safeParse(value);
+  if (!allowed.success) {
+    throw validationError(`"linkKind" must be one of: ${siteHealthLinkKindSchema.options.join(", ")}`);
+  }
+  return allowed.data;
+}
+function linkKindFilter(linkKind) {
+  if (linkKind === SiteHealthLinkKinds.content) return eq56(siteCrawlEdges.isTemplate, false);
+  if (linkKind === SiteHealthLinkKinds.template) return eq56(siteCrawlEdges.isTemplate, true);
+  return void 0;
+}
 var graphSourceNode = alias(siteCrawlGraphNodes, "site_crawl_graph_source_node");
 var graphTargetNode = alias(siteCrawlGraphNodes, "site_crawl_graph_target_node");
 var PAGE_CHANGE_PREDICATE = sql21`(
@@ -57510,6 +58065,7 @@ async function technicalAeoRoutes(app, opts) {
   });
   app.get("/projects/:name/technical-aeo/graph", async (request) => {
     const project = resolveProject(app.db, request.params.name);
+    const linkKind = parseLinkKind(request.query.linkKind);
     const target = resolveCrawl(project.id, request.query.runId);
     if (!target) {
       assertKnownAuditRun(project.id, request.query.runId);
@@ -57519,8 +58075,13 @@ async function technicalAeoRoutes(app, opts) {
         runId: null,
         rootNodeKey: null,
         layout: { state: "unavailable", version: null, reason: "no-crawl" },
+        // Nothing was classified because nothing was crawled.
+        templateDetection: SiteHealthTemplateDetections["unavailable-legacy-scan"],
+        linkKind,
         totalNodes: 0,
         totalEdges: 0,
+        totalTemplateEdges: 0,
+        totalContentEdges: 0,
         nodes: [],
         edges: [],
         omittedNodes: 0,
@@ -57529,6 +58090,7 @@ async function technicalAeoRoutes(app, opts) {
       };
     }
     const snapshot = target.snapshot;
+    const templateDetection = templateDetectionOf(snapshot.templateDetection);
     if (!snapshot.detailsAvailable || !snapshot.attemptId) {
       return {
         project: project.name,
@@ -57536,8 +58098,12 @@ async function technicalAeoRoutes(app, opts) {
         runId: snapshot.runId,
         rootNodeKey: null,
         layout: { state: "unavailable", version: null, reason: "details-unavailable" },
+        templateDetection,
+        linkKind,
         totalNodes: 0,
         totalEdges: 0,
+        totalTemplateEdges: 0,
+        totalContentEdges: 0,
         nodes: [],
         edges: [],
         omittedNodes: 0,
@@ -57561,8 +58127,12 @@ async function technicalAeoRoutes(app, opts) {
         runId: snapshot.runId,
         rootNodeKey,
         layout: { state: "unavailable", version: null, reason: "legacy-snapshot" },
+        templateDetection,
+        linkKind,
         totalNodes: 0,
         totalEdges: 0,
+        totalTemplateEdges: 0,
+        totalContentEdges: 0,
         nodes: [],
         edges: [],
         omittedNodes: 0,
@@ -57583,8 +58153,12 @@ async function technicalAeoRoutes(app, opts) {
           version: null,
           reason: persistedLayout.failureCode === "empty-crawl" ? "empty-crawl" : "layout-failed"
         },
+        templateDetection,
+        linkKind,
         totalNodes: totalNodes2,
         totalEdges: totalEdges2,
+        totalTemplateEdges: persistedLayout.totalTemplateEdges,
+        totalContentEdges: Math.max(0, totalEdges2 - persistedLayout.totalTemplateEdges),
         nodes: [],
         edges: [],
         omittedNodes: totalNodes2,
@@ -57626,12 +58200,14 @@ async function technicalAeoRoutes(app, opts) {
       ...row,
       healthState: deriveSiteHealthState({ ...row, indexabilityReasons, canonicalNodeKey })
     }));
+    const graphLinkKindFilter = linkKind === SiteHealthLinkKinds.all ? void 0 : eq56(siteCrawlGraphEdges.isTemplate, linkKind === SiteHealthLinkKinds.template);
     const edges = app.db.select({
       edgeKey: siteCrawlGraphEdges.edgeKey,
       sourceNodeKey: siteCrawlGraphEdges.sourceNodeKey,
       targetNodeKey: siteCrawlGraphEdges.targetNodeKey,
       followable: siteCrawlGraphEdges.followable,
-      occurrences: siteCrawlGraphEdges.occurrences
+      occurrences: siteCrawlGraphEdges.occurrences,
+      isTemplate: siteCrawlGraphEdges.isTemplate
     }).from(siteCrawlGraphEdges).innerJoin(graphSourceNode, and44(
       eq56(graphSourceNode.projectId, siteCrawlGraphEdges.projectId),
       eq56(graphSourceNode.runId, siteCrawlGraphEdges.runId),
@@ -57648,10 +58224,12 @@ async function technicalAeoRoutes(app, opts) {
       eq56(siteCrawlGraphEdges.projectId, project.id),
       eq56(siteCrawlGraphEdges.runId, snapshot.runId),
       eq56(siteCrawlGraphEdges.attemptId, snapshot.attemptId),
-      lt8(siteCrawlGraphEdges.sampleRank, maxEdges)
+      lt8(siteCrawlGraphEdges.sampleRank, maxEdges),
+      graphLinkKindFilter
     )).orderBy(asc10(siteCrawlGraphEdges.sampleRank)).all();
     const totalNodes = persistedLayout.totalNodes;
     const totalEdges = persistedLayout.totalEdges;
+    const totalTemplateEdges = persistedLayout.totalTemplateEdges;
     const omittedNodes = Math.max(0, totalNodes - nodes.length);
     const omittedEdges = Math.max(0, totalEdges - edges.length);
     return {
@@ -57659,9 +58237,21 @@ async function technicalAeoRoutes(app, opts) {
       hasCrawlData: true,
       runId: snapshot.runId,
       rootNodeKey,
-      layout: { state: "ready", version: persistedLayout.layoutVersion, computedAt: persistedLayout.createdAt },
+      layout: {
+        state: "ready",
+        version: persistedLayout.layoutVersion,
+        computedAt: persistedLayout.createdAt,
+        templateLinksExcluded: persistedLayout.templateLinksExcluded
+      },
+      templateDetection,
+      linkKind,
       totalNodes,
+      // Deliberately still every graph-compatible link, whatever `linkKind`
+      // asked for. The two counts below split it; none of the three moves
+      // because a caller narrowed the payload.
       totalEdges,
+      totalTemplateEdges,
+      totalContentEdges: Math.max(0, totalEdges - totalTemplateEdges),
       nodes,
       edges,
       omittedNodes,
@@ -58343,41 +58933,52 @@ async function technicalAeoRoutes(app, opts) {
     const target = resolveCrawl(project.id, request.query.runId);
     if (!target) {
       assertKnownAuditRun(project.id, request.query.runId);
-      return { project: project.name, hasCrawlData: false, runId: null, total: 0, nextCursor: null, pages: [] };
+      return { project: project.name, hasCrawlData: false, runId: null, total: 0, nextCursor: null, healthStateFilter: null, pages: [] };
     }
     const snapshot = target.snapshot;
     if (!snapshot.detailsAvailable || !snapshot.attemptId) {
-      return { project: project.name, hasCrawlData: true, runId: snapshot.runId, total: 0, nextCursor: null, pages: [] };
+      return { project: project.name, hasCrawlData: true, runId: snapshot.runId, total: 0, nextCursor: null, healthStateFilter: null, pages: [] };
     }
     const filters = [
       eq56(siteCrawlPages.projectId, project.id),
       eq56(siteCrawlPages.runId, snapshot.runId),
       eq56(siteCrawlPages.attemptId, snapshot.attemptId)
     ];
+    if (request.query.nodeKey) filters.push(eq56(siteCrawlPages.nodeKey, request.query.nodeKey));
     const inventoryEligible = parseBoolean2(request.query.inventoryEligible);
     if (inventoryEligible != null) filters.push(eq56(siteCrawlPages.inventoryEligible, inventoryEligible));
     if (request.query.fetchState) filters.push(eq56(siteCrawlPages.fetchState, request.query.fetchState));
     if (request.query.indexabilityState) filters.push(eq56(siteCrawlPages.indexabilityState, request.query.indexabilityState));
     if (request.query.auditState) filters.push(eq56(siteCrawlPages.auditState, request.query.auditState));
     const healthState = parseSiteHealthState(request.query.healthState);
-    const where = and44(...filters);
     const limit = parseBoundedLimit(request.query.limit, 100, 200);
     const offset = decodeCursor(request.query.cursor);
     const orderBy = request.query.sort === "score-desc" ? [desc24(siteCrawlPages.auditScore), asc10(siteCrawlPages.nodeKey)] : request.query.sort === "score-asc" ? [asc10(siteCrawlPages.auditScore), asc10(siteCrawlPages.nodeKey)] : request.query.sort === "path" ? [asc10(siteCrawlPages.path), asc10(siteCrawlPages.nodeKey)] : [asc10(siteCrawlPages.url), asc10(siteCrawlPages.nodeKey)];
-    let total;
-    let rows;
+    let healthStateFilter = null;
     if (healthState) {
-      const candidates = app.db.select().from(siteCrawlPages).where(where).orderBy(...orderBy).limit(MAX_STRUCTURE_SOURCE_ROWS + 1).all();
-      if (candidates.length > MAX_STRUCTURE_SOURCE_ROWS) {
-        throw validationError(`Persisted crawl exceeds the ${MAX_STRUCTURE_SOURCE_ROWS}-page health-state filter limit`);
-      }
-      const matching = candidates.filter((row) => deriveSiteHealthState(row) === healthState);
-      total = matching.length;
-      rows = matching.slice(offset, offset + limit);
-    } else {
-      total = app.db.select({ value: count() }).from(siteCrawlPages).where(where).get()?.value ?? 0;
-      rows = app.db.select().from(siteCrawlPages).where(where).orderBy(...orderBy).limit(limit).offset(offset).all();
+      const legacyRow = app.db.select({ id: siteCrawlPages.id }).from(siteCrawlPages).where(and44(
+        eq56(siteCrawlPages.projectId, project.id),
+        eq56(siteCrawlPages.runId, snapshot.runId),
+        eq56(siteCrawlPages.attemptId, snapshot.attemptId),
+        isNull5(siteCrawlPages.healthState)
+      )).limit(1).get();
+      healthStateFilter = legacyRow ? "unavailable-legacy-scan" : "applied";
+      if (healthStateFilter === "applied") filters.push(eq56(siteCrawlPages.healthState, healthState));
+    }
+    if (healthStateFilter === "unavailable-legacy-scan") {
+      return {
+        project: project.name,
+        hasCrawlData: true,
+        runId: snapshot.runId,
+        total: 0,
+        nextCursor: null,
+        healthStateFilter,
+        pages: []
+      };
     }
+    const where = and44(...filters);
+    const total = app.db.select({ value: count() }).from(siteCrawlPages).where(where).get()?.value ?? 0;
+    const rows = app.db.select().from(siteCrawlPages).where(where).orderBy(...orderBy).limit(limit).offset(offset).all();
     const nextOffset = offset + rows.length;
     return {
       project: project.name,
@@ -58385,6 +58986,7 @@ async function technicalAeoRoutes(app, opts) {
       runId: snapshot.runId,
       total,
       nextCursor: nextOffset < total ? encodeCursor2(nextOffset) : null,
+      healthStateFilter,
       pages: rows.map(mapCrawlPage)
     };
   });
@@ -58457,20 +59059,41 @@ async function technicalAeoRoutes(app, opts) {
   });
   app.get("/projects/:name/technical-aeo/internal-links", async (request) => {
     const project = resolveProject(app.db, request.params.name);
+    const linkKind = parseLinkKind(request.query.linkKind);
     const target = resolveCrawl(project.id, request.query.runId);
     if (!target) {
       assertKnownAuditRun(project.id, request.query.runId);
-      return { project: project.name, hasCrawlData: false, runId: null, total: 0, nextCursor: null, edges: [] };
+      return {
+        project: project.name,
+        hasCrawlData: false,
+        runId: null,
+        total: 0,
+        nextCursor: null,
+        templateDetection: SiteHealthTemplateDetections["unavailable-legacy-scan"],
+        linkKind,
+        edges: []
+      };
     }
     const snapshot = target.snapshot;
+    const templateDetection = templateDetectionOf(snapshot.templateDetection);
     if (!snapshot.detailsAvailable || !snapshot.attemptId) {
-      return { project: project.name, hasCrawlData: true, runId: snapshot.runId, total: 0, nextCursor: null, edges: [] };
+      return {
+        project: project.name,
+        hasCrawlData: true,
+        runId: snapshot.runId,
+        total: 0,
+        nextCursor: null,
+        templateDetection,
+        linkKind,
+        edges: []
+      };
     }
     const filters = [
       eq56(siteCrawlEdges.projectId, project.id),
       eq56(siteCrawlEdges.runId, snapshot.runId),
       eq56(siteCrawlEdges.attemptId, snapshot.attemptId),
-      eq56(siteCrawlEdges.internal, true)
+      eq56(siteCrawlEdges.internal, true),
+      linkKindFilter(linkKind)
     ];
     if (request.query.sourceUrl) filters.push(eq56(siteCrawlEdges.sourceUrl, request.query.sourceUrl));
     if (request.query.targetUrl) filters.push(eq56(siteCrawlEdges.targetUrl, request.query.targetUrl));
@@ -58488,11 +59111,14 @@ async function technicalAeoRoutes(app, opts) {
       runId: snapshot.runId,
       total,
       nextCursor: nextOffset < total ? encodeCursor2(nextOffset) : null,
+      templateDetection,
+      linkKind,
       edges: rows.map(mapCrawlEdge)
     };
   });
   app.get("/projects/:name/technical-aeo/internal-links/neighbors", async (request) => {
     const project = resolveProject(app.db, request.params.name);
+    const linkKind = parseLinkKind(request.query.linkKind);
     if (!request.query.nodeKey && !request.query.url) {
       throw validationError("Provide nodeKey or url");
     }
@@ -58505,6 +59131,8 @@ async function technicalAeoRoutes(app, opts) {
         runId: null,
         nodeKey: request.query.nodeKey ?? null,
         url: request.query.url ?? null,
+        templateDetection: SiteHealthTemplateDetections["unavailable-legacy-scan"],
+        linkKind,
         inbound: [],
         outbound: [],
         inboundTruncated: false,
@@ -58512,6 +59140,7 @@ async function technicalAeoRoutes(app, opts) {
       };
     }
     const snapshot = target.snapshot;
+    const templateDetection = templateDetectionOf(snapshot.templateDetection);
     if (!snapshot.detailsAvailable || !snapshot.attemptId) {
       return {
         project: project.name,
@@ -58519,6 +59148,8 @@ async function technicalAeoRoutes(app, opts) {
         runId: snapshot.runId,
         nodeKey: request.query.nodeKey ?? null,
         url: request.query.url ?? null,
+        templateDetection,
+        linkKind,
         inbound: [],
         outbound: [],
         inboundTruncated: false,
@@ -58529,7 +59160,8 @@ async function technicalAeoRoutes(app, opts) {
       eq56(siteCrawlEdges.projectId, project.id),
       eq56(siteCrawlEdges.runId, snapshot.runId),
       eq56(siteCrawlEdges.attemptId, snapshot.attemptId),
-      eq56(siteCrawlEdges.internal, true)
+      eq56(siteCrawlEdges.internal, true),
+      linkKindFilter(linkKind)
     ];
     const inboundMatch = request.query.nodeKey && request.query.url ? or11(eq56(siteCrawlEdges.targetNodeKey, request.query.nodeKey), eq56(siteCrawlEdges.targetUrl, request.query.url)) : request.query.nodeKey ? eq56(siteCrawlEdges.targetNodeKey, request.query.nodeKey) : eq56(siteCrawlEdges.targetUrl, request.query.url);
     const outboundMatch = request.query.nodeKey && request.query.url ? or11(eq56(siteCrawlEdges.sourceNodeKey, request.query.nodeKey), eq56(siteCrawlEdges.sourceUrl, request.query.url)) : request.query.nodeKey ? eq56(siteCrawlEdges.sourceNodeKey, request.query.nodeKey) : eq56(siteCrawlEdges.sourceUrl, request.query.url);
@@ -58542,6 +59174,8 @@ async function technicalAeoRoutes(app, opts) {
       runId: snapshot.runId,
       nodeKey: request.query.nodeKey ?? null,
       url: request.query.url ?? null,
+      templateDetection,
+      linkKind,
       inbound: inboundRows.slice(0, limit).map(mapCrawlEdge),
       outbound: outboundRows.slice(0, limit).map(mapCrawlEdge),
       inboundTruncated: inboundRows.length > limit,
@@ -58631,12 +59265,121 @@ async function technicalAeoRoutes(app, opts) {
       }))
     };
   });
+  app.get("/projects/:name/technical-aeo/runs/:runId/progress", async (request) => {
+    const project = resolveProject(app.db, request.params.name);
+    const run = app.db.select().from(runs).where(and44(
+      eq56(runs.id, request.params.runId),
+      eq56(runs.projectId, project.id),
+      eq56(runs.kind, RunKinds["site-audit"]),
+      notProbeRun()
+    )).get();
+    if (!run) throw notFound("Site audit run", request.params.runId);
+    const attempt = app.db.select().from(siteCrawlAttempts).where(and44(
+      eq56(siteCrawlAttempts.projectId, project.id),
+      eq56(siteCrawlAttempts.runId, run.id)
+    )).orderBy(desc24(siteCrawlAttempts.attemptNumber), desc24(siteCrawlAttempts.updatedAt)).limit(1).get();
+    const persistedLayout = attempt ? app.db.select().from(siteCrawlGraphLayouts).where(and44(
+      eq56(siteCrawlGraphLayouts.projectId, project.id),
+      eq56(siteCrawlGraphLayouts.runId, run.id),
+      eq56(siteCrawlGraphLayouts.attemptId, attempt.id)
+    )).get() : void 0;
+    const layoutState = persistedLayout?.state === "ready" || persistedLayout?.state === "unavailable" ? persistedLayout.state : run.status === RunStatuses.completed || run.status === RunStatuses.partial ? "unavailable" : "pending";
+    return {
+      project: project.name,
+      runId: run.id,
+      status: run.status,
+      phase: siteAuditRunPhase(run.status, attempt, layoutState),
+      attempt: attempt ? {
+        id: attempt.id,
+        state: attempt.state,
+        pagesDiscovered: attempt.pagesDiscovered,
+        pagesFetched: attempt.pagesFetched,
+        pagesEligible: attempt.pagesEligible,
+        pagesErrored: attempt.pagesErrored,
+        edgesDiscovered: attempt.edgesDiscovered,
+        lastUpdatedAt: attempt.updatedAt,
+        startedAt: attempt.startedAt,
+        finishedAt: attempt.finishedAt,
+        error: attempt.error
+      } : null,
+      layout: {
+        state: layoutState,
+        layoutVersion: persistedLayout?.layoutVersion ?? null,
+        failureCode: persistedLayout?.failureCode ?? null,
+        updatedAt: persistedLayout?.updatedAt ?? null
+      },
+      error: run.error ?? attempt?.error ?? null
+    };
+  });
+  app.get("/projects/:name/technical-aeo/runs/:runId/page-health-preview", async (request) => app.db.transaction((tx) => {
+    const project = tx.select().from(projects).where(eq56(projects.name, request.params.name)).get();
+    if (!project) throw notFound("Project", request.params.name);
+    const run = tx.select().from(runs).where(and44(
+      eq56(runs.id, request.params.runId),
+      eq56(runs.projectId, project.id),
+      eq56(runs.kind, RunKinds["site-audit"]),
+      notProbeRun()
+    )).get();
+    if (!run) throw notFound("Site audit run", request.params.runId);
+    const attempt = tx.select().from(siteCrawlAttempts).where(and44(
+      eq56(siteCrawlAttempts.projectId, project.id),
+      eq56(siteCrawlAttempts.runId, run.id)
+    )).orderBy(desc24(siteCrawlAttempts.attemptNumber), desc24(siteCrawlAttempts.updatedAt)).limit(1).get();
+    const state = run.status === RunStatuses.queued ? "waiting" : run.status === RunStatuses.running ? "collecting" : "terminal";
+    const base = {
+      project: project.name,
+      runId: run.id,
+      status: run.status,
+      state,
+      attemptId: attempt?.id ?? null,
+      updatedAt: attempt?.updatedAt ?? null
+    };
+    if (!attempt) {
+      return { ...base, pagesAudited: 0, examples: [] };
+    }
+    const auditedWhere = and44(
+      eq56(siteCrawlPages.projectId, project.id),
+      eq56(siteCrawlPages.runId, run.id),
+      eq56(siteCrawlPages.attemptId, attempt.id),
+      eq56(siteCrawlPages.auditState, "success")
+    );
+    const pagesAudited = tx.select({ value: count() }).from(siteCrawlPages).where(auditedWhere).get()?.value ?? 0;
+    if (state !== "collecting") {
+      return { ...base, pagesAudited, examples: [] };
+    }
+    const candidates = tx.select({
+      nodeKey: siteCrawlPages.nodeKey,
+      url: siteCrawlPages.url,
+      auditScore: siteCrawlPages.auditScore,
+      auditFields: siteCrawlPages.auditFields
+    }).from(siteCrawlPages).where(and44(
+      auditedWhere,
+      isNotNull3(siteCrawlPages.auditScore),
+      lt8(siteCrawlPages.auditScore, 70)
+    )).orderBy(asc10(siteCrawlPages.auditScore), asc10(siteCrawlPages.nodeKey)).limit(LIVE_PAGE_HEALTH_CANDIDATE_LIMIT).all();
+    const examples = candidates.flatMap((candidate) => {
+      if (candidate.auditScore == null) return [];
+      const checksNeedingAttention = countLivePageHealthChecks(candidate.auditFields);
+      return checksNeedingAttention > 0 ? [{
+        nodeKey: candidate.nodeKey,
+        url: candidate.url,
+        auditScore: candidate.auditScore,
+        checksNeedingAttention
+      }] : [];
+    }).slice(0, LIVE_PAGE_HEALTH_EXAMPLE_LIMIT);
+    return { ...base, pagesAudited, examples };
+  }));
   app.post("/projects/:name/technical-aeo/runs", async (request) => {
     const project = resolveProject(app.db, request.params.name);
     const parsed = siteAuditRunRequestSchema.safeParse(request.body ?? {});
     if (!parsed.success) {
       throw validationError(parsed.error.issues[0]?.message ?? "Invalid site-audit request");
     }
+    if (!opts.onSiteAuditRequested) {
+      throw missingDependency("Site Health execution is not available on this deployment.", {
+        reason: "no-site-audit-handler"
+      });
+    }
     const effectiveRequest = normalizeSiteAuditRunRequest(parsed.data);
     const identityKey = siteAuditRequestIdentity(effectiveRequest);
     const result = app.db.transaction((tx) => {
@@ -58687,7 +59430,7 @@ async function technicalAeoRoutes(app, opts) {
       return { runId, status: RunStatuses.queued, created: true };
     });
     if (result.created) {
-      opts.onSiteAuditRequested?.(result.runId, project.id, {
+      opts.onSiteAuditRequested(result.runId, project.id, {
         sitemapUrl: effectiveRequest.sitemapUrl ?? void 0,
         limit: parsed.data.limit,
         maxPages: effectiveRequest.maxPages,
@@ -60092,6 +60835,7 @@ export {
   latestQueryBasketRevision,
   ensureCurrentQueryBasketRevision,
   queueRunIfProjectIdle,
+  resolveMeasurementSitemapTarget,
   measurementRunCompleteness,
   nextRunFromCron,
   resolveWebhookTarget,