feat(api): unified FTS search

Single GET /api/search?q=&space_id=&kinds=&limit=&offset= unions FTS hits across pages / refs / source_docs / messages with a `kind` discriminator and ts_rank ordering. Each branch's to_tsvector matches the GIN index expression on its source table so indexes are used. Messages have no space_id and are excluded when a space filter is set. Hybrid vector / RRF lands in Plan 3. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-06-01 02:04:57 +10:00
parent ec96e4e2e3
commit 69e26ada98
5 changed files with 289 additions and 0 deletions
--- a/lib/db/repos/search.js
+++ b/lib/db/repos/search.js
@@ -0,0 +1,83 @@
+import { pool } from '../pool.js';
+
+// FTS-only search across pages / refs / source_docs / messages, unioned
+// with a `kind` discriminator and ranked by ts_rank. Each branch's
+// to_tsvector expression matches the GIN index on its table so the
+// indexes are used. Vector / hybrid RRF search lands in Plan 3.
+//
+// Notes:
+// - messages have no space_id → the messages branch is dropped when a
+//   space_id filter is present.
+// - source_docs inherit space_id from their owning resource via join.
+
+const PAGES_TSV = `to_tsvector('english', p.title || ' ' || coalesce(p.body_md,''))`;
+const REFS_TSV  = `to_tsvector('english', coalesce(r.title,'') || ' ' || coalesce(r.summary,'') || ' ' || coalesce(r.body_text,''))`;
+const SD_TSV    = `to_tsvector('english', sd.name || ' ' || coalesce(sd.body_text,''))`;
+const MSG_TSV   = `to_tsvector('english', m.body)`;
+
+function buildBranches({ kinds, spaceFilterPresent }) {
+  const branches = [];
+  const wantPage = !kinds || kinds.includes('page');
+  const wantRef  = !kinds || kinds.includes('ref');
+  const wantSD   = !kinds || kinds.includes('source_doc');
+  const wantMsg  = (!kinds || kinds.includes('message')) && !spaceFilterPresent;
+
+  if (wantPage) {
+    branches.push(`
+      SELECT 'page'::text AS kind, p.id, p.space_id, p.title AS title_or_snippet,
+             ts_rank(${PAGES_TSV}, q.tsq) AS rank
+        FROM pages p, q
+        WHERE ${PAGES_TSV} @@ q.tsq
+          AND ($2::uuid IS NULL OR p.space_id = $2)
+    `);
+  }
+  if (wantRef) {
+    branches.push(`
+      SELECT 'ref'::text AS kind, r.id, r.space_id,
+             coalesce(r.title, r.source_url, '(untitled)') AS title_or_snippet,
+             ts_rank(${REFS_TSV}, q.tsq) AS rank
+        FROM refs r, q
+        WHERE ${REFS_TSV} @@ q.tsq
+          AND ($2::uuid IS NULL OR r.space_id = $2)
+    `);
+  }
+  if (wantSD) {
+    branches.push(`
+      SELECT 'source_doc'::text AS kind, sd.id, res.space_id, sd.name AS title_or_snippet,
+             ts_rank(${SD_TSV}, q.tsq) AS rank
+        FROM source_docs sd
+        JOIN resources res ON res.id = sd.resource_id, q
+        WHERE ${SD_TSV} @@ q.tsq
+          AND ($2::uuid IS NULL OR res.space_id = $2)
+    `);
+  }
+  if (wantMsg) {
+    branches.push(`
+      SELECT 'message'::text AS kind, m.id, NULL::uuid AS space_id,
+             substring(m.body, 1, 200) AS title_or_snippet,
+             ts_rank(${MSG_TSV}, q.tsq) AS rank
+        FROM messages m, q
+        WHERE ${MSG_TSV} @@ q.tsq
+    `);
+  }
+  return branches;
+}
+
+export async function fts({ q, space_id = null, kinds = null, limit = 50, offset = 0 } = {}) {
+  if (!q || typeof q !== 'string') return [];
+  const normalizedKinds = Array.isArray(kinds) && kinds.length ? kinds : null;
+  const spaceFilterPresent = space_id !== null && space_id !== undefined;
+  const branches = buildBranches({ kinds: normalizedKinds, spaceFilterPresent });
+  if (branches.length === 0) return [];
+
+  const sql = `
+    WITH q AS (SELECT plainto_tsquery('english', $1) AS tsq)
+    SELECT * FROM (
+      ${branches.join('\n      UNION ALL\n      ')}
+    ) u
+    ORDER BY rank DESC
+    LIMIT $3 OFFSET $4
+  `;
+  const { rows } = await pool.query(sql, [q, space_id, limit, offset]);
+  return rows;
+}