@jhizzard/termdeck 0.10.2 → 0.10.3

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jhizzard/termdeck",
-  "version": "0.10.2",
+  "version": "0.10.3",
   "description": "Browser-based terminal multiplexer with metadata overlays, panel flashback memory recall, and AI-aware session management",
   "bin": {
     "termdeck": "./packages/cli/src/index.js"
@@ -42,6 +42,9 @@
     "ws": "^8.16.0",
     "yaml": "^2.3.4"
   },
+  "devDependencies": {
+    "@anthropic-ai/sdk": "^0.39.0"
+  },
   "keywords": [
     "terminal",
     "multiplexer",

package/packages/client/public/graph.html

@@ -51,6 +51,7 @@
     <div class="graph-empty" id="graphEmpty" hidden>
       <h3 id="graphEmptyTitle">No memories yet</h3>
       <p id="graphEmptyBody">This project has no <code>memory_items</code> rows. Run a Claude Code session in this project; the session-end hook will populate Mnestra and edges will be inferred on the next nightly cron.</p>
+      <button type="button" class="graph-empty-action" id="graphEmptyAllProjects" hidden>View All Projects</button>
     </div>
 
     <svg class="graph-svg" id="graphSvg" role="img" aria-label="Force-directed knowledge graph">

package/packages/client/public/graph.js

@@ -186,6 +186,51 @@
     window.history.replaceState(null, '', url);
   }
 
+  // -------- Stateful UI reset (Sprint 41 T3) ------------------------------
+
+  // Clear the SVG render and tear down the running simulation. Called at the
+  // start of every fetchGraph so a re-fetch from a different mode/project
+  // can't paint over a stale render.
+  function clearGraphSvg() {
+    if (state.sim) {
+      state.sim.stop();
+      state.sim = null;
+    }
+    const r = root();
+    if (r) {
+      while (r.firstChild) r.removeChild(r.firstChild);
+    }
+    state.nodeSel = null;
+    state.edgeSel = null;
+    state.labelSel = null;
+  }
+
+  // Ephemeral toast in the top-right of the graph stage. Used to surface
+  // truncation warnings on the All Projects view. Auto-dismisses; calling
+  // again before dismissal replaces the message and resets the timer.
+  function showToast(msg, durationMs = 6000) {
+    const stageEl = stage();
+    if (!stageEl) return;
+    let el = document.getElementById('graphToast');
+    if (!el) {
+      el = document.createElement('div');
+      el.id = 'graphToast';
+      el.className = 'graph-toast';
+      stageEl.appendChild(el);
+    }
+    el.textContent = msg;
+    el.hidden = false;
+    // Force a reflow so the .show transition fires when toggling rapid-fire.
+    void el.offsetWidth;
+    el.classList.add('show');
+    if (showToast._timer) clearTimeout(showToast._timer);
+    showToast._timer = setTimeout(() => {
+      el.classList.remove('show');
+      // Hide after the transition completes so it can't catch clicks.
+      setTimeout(() => { el.hidden = true; }, 220);
+    }, durationMs);
+  }
+
   // -------- API ------------------------------------------------------------
 
   async function api(path) {
@@ -208,14 +253,32 @@
   }
 
   async function fetchGraph() {
+    // Sprint 41 T3 — reset all stateful UI before the new fetch starts so a
+    // re-fetch from a different mode/project starts from a clean slate. Fixes
+    // the three-way race where "Loading graph…" + "No memories yet" + a stale
+    // node render all paint over each other after a mode/project switch.
+    hideEmpty();
+    clearGraphSvg();
+    state.nodes = [];
+    state.edges = [];
     setLoading('Loading graph…');
     try {
       let data;
-      if (state.mode === 'memory') {
+      if (state.mode === 'project' && state.project === '__all__') {
+        data = await api('/api/graph/all');
+        if (data.enabled === false) return showDisabled(data);
+        state.nodes = data.nodes || [];
+        state.edges = data.edges || [];
+        if (data.truncated) {
+          showToast(
+            `Showing ${state.nodes.length} most-recent of ${data.totalAvailable} memories — narrow by project to see specific clusters.`,
+          );
+        }
+      } else if (state.mode === 'memory') {
         data = await api(`/api/graph/memory/${encodeURIComponent(state.memoryId)}?depth=${state.depth}`);
         if (data.enabled === false) return showDisabled(data);
-        state.nodes = data.nodes;
-        state.edges = data.edges;
+        state.nodes = data.nodes || [];
+        state.edges = data.edges || [];
         // Use the root memory's project as the view's "current project" for
         // the legend / drawer / fallback color when nodes span projects.
         if (data.root && data.root.project) state.project = data.root.project;
@@ -224,8 +287,8 @@
         state.project = name;
         data = await api(`/api/graph/project/${encodeURIComponent(name)}`);
         if (data.enabled === false) return showDisabled(data);
-        state.nodes = data.nodes;
-        state.edges = data.edges;
+        state.nodes = data.nodes || [];
+        state.edges = data.edges || [];
       }
       writeUrlState();
       hideLoading();
@@ -238,6 +301,9 @@
       renderGraph();
       updateStats();
     } catch (err) {
+      // Even on error, drop the empty-state overlay so the failure message
+      // shows alone instead of stacking on top of "No memories yet".
+      hideEmpty();
       setLoading(`Failed: ${err.message}`);
     }
   }
@@ -256,14 +322,30 @@
   }
   function showEmpty() {
     $('graphEmpty').hidden = false;
-    if (state.mode === 'project') {
-      $('graphEmptyTitle').textContent = `No memories in "${state.project}"`;
+    const allBtn = $('graphEmptyAllProjects');
+    if (state.mode === 'project' && state.project === '__all__') {
+      $('graphEmptyTitle').textContent = 'No memories yet';
+      $('graphEmptyBody').innerHTML =
+        'No <code>memory_items</code> rows in the database. Run a Claude Code session and the session-end hook will populate Mnestra; edges will be inferred on the next nightly cron.';
+      if (allBtn) allBtn.hidden = true;
+    } else if (state.mode === 'project') {
+      $('graphEmptyTitle').textContent = `No memories tagged "${state.project}"`;
+      $('graphEmptyBody').innerHTML =
+        'Your <code>memory_items</code> may be mis-tagged under a parent directory. ' +
+        'Try the All Projects view, or check the actual distribution with ' +
+        '<code>SELECT project, count(*) FROM memory_items GROUP BY project</code>.';
+      if (allBtn) allBtn.hidden = false;
     } else {
       $('graphEmptyTitle').textContent = 'No neighbors yet';
+      $('graphEmptyBody').textContent =
+        'This memory has no edges in memory_relationships. The next graph-inference cron run will infer them if any are warranted.';
+      if (allBtn) allBtn.hidden = true;
     }
   }
   function hideEmpty() {
     $('graphEmpty').hidden = true;
+    const allBtn = $('graphEmptyAllProjects');
+    if (allBtn) allBtn.hidden = true;
   }
   function showDisabled(data) {
     hideLoading();
@@ -610,28 +692,31 @@
     readUrlState();
     await loadConfig();
 
-    // Project picker
+    // Project picker. The "All projects" option is always present so the user
+    // can recover from mis-tagged data (Sprint 41 T3); per-project options are
+    // appended after it from /api/config.
     const sel = $('graphProject');
     sel.innerHTML = '';
-    if (state.projects.length === 0) {
+    sel.disabled = false;
+    const allOpt = document.createElement('option');
+    allOpt.value = '__all__';
+    allOpt.textContent = 'All projects';
+    sel.appendChild(allOpt);
+    for (const p of state.projects) {
       const opt = document.createElement('option');
-      opt.value = '';
-      opt.textContent = '— add a project first —';
+      opt.value = p;
+      opt.textContent = p;
       sel.appendChild(opt);
-      sel.disabled = true;
-    } else {
-      for (const p of state.projects) {
-        const opt = document.createElement('option');
-        opt.value = p;
-        opt.textContent = p;
-        sel.appendChild(opt);
-      }
-      // Pick state.project, or first project from config.
-      if (!state.project && state.mode !== 'memory') {
-        state.project = state.projects[0];
+    }
+    // Pick state.project, or fall back to the first configured project, or to
+    // __all__ when nothing is configured. Memory mode inherits its project
+    // from the root node and skips this resolution.
+    if (state.mode !== 'memory') {
+      if (!state.project) {
+        state.project = state.projects.length > 0 ? state.projects[0] : '__all__';
       }
-      if (state.project) sel.value = state.project;
     }
+    if (state.project) sel.value = state.project;
 
     sel.addEventListener('change', () => {
       state.mode = 'project';
@@ -640,6 +725,17 @@
       fetchGraph();
     });
 
+    const emptyAllBtn = $('graphEmptyAllProjects');
+    if (emptyAllBtn) {
+      emptyAllBtn.addEventListener('click', () => {
+        state.mode = 'project';
+        state.project = '__all__';
+        state.memoryId = null;
+        sel.value = '__all__';
+        fetchGraph();
+      });
+    }
+
     $('graphSearch').addEventListener('input', (e) => applySearch(e.target.value));
     $('graphReheat').addEventListener('click', () => {
       if (state.sim) state.sim.alpha(0.6).restart();

package/packages/client/public/style.css

@@ -3468,6 +3468,59 @@
       border-radius: 3px;
       font-size: 11px;
     }
+    /* Sprint 41 T3 — interactive button inside the .graph-empty overlay.
+       The parent has pointer-events: none so taps on the SVG behind pass
+       through; the button must opt back in for the click to register. */
+    .graph-empty .graph-empty-action {
+      pointer-events: auto;
+      margin-top: 6px;
+      background: var(--tg-surface);
+      color: var(--tg-text);
+      border: 1px solid var(--tg-border-active);
+      border-radius: var(--tg-radius-sm);
+      padding: 6px 14px;
+      font-family: var(--tg-mono);
+      font-size: 12px;
+      cursor: pointer;
+      transition: background 0.12s ease, border-color 0.12s ease, color 0.12s ease;
+    }
+    .graph-empty .graph-empty-action:hover {
+      background: var(--tg-surface-hover);
+      border-color: var(--tg-accent);
+      color: var(--tg-text-bright);
+    }
+    .graph-empty .graph-empty-action:focus-visible {
+      outline: 2px solid var(--tg-accent);
+      outline-offset: 2px;
+    }
+
+    /* Sprint 41 T3 — ephemeral toast for graph-side notifications
+       (truncation warnings, etc). Sits in the top-right of #graphStage. */
+    .graph-toast {
+      position: absolute;
+      top: 14px;
+      right: 14px;
+      max-width: 360px;
+      background: rgba(15, 17, 23, 0.95);
+      color: var(--tg-text);
+      border: 1px solid var(--tg-border-active);
+      border-radius: var(--tg-radius-sm);
+      padding: 10px 14px;
+      font-family: var(--tg-mono);
+      font-size: 12px;
+      line-height: 1.4;
+      box-shadow: 0 4px 14px rgba(0, 0, 0, 0.35);
+      backdrop-filter: blur(6px);
+      z-index: 60;
+      opacity: 0;
+      transform: translateY(-6px);
+      transition: opacity 0.18s ease, transform 0.18s ease;
+      pointer-events: none;
+    }
+    .graph-toast.show {
+      opacity: 1;
+      transform: translateY(0);
+    }
 
     .graph-tooltip {
       position: fixed;

package/packages/server/src/graph-routes.js

@@ -26,6 +26,10 @@ const NODE_LABEL_LEN = 200;
 const MAX_DEPTH = 4;
 const DEFAULT_DEPTH = 2;
 const MAX_NODES_PER_PROJECT = 2000;
+// Sprint 41 T3 — All Projects view cap. Same ceiling as the per-project cap;
+// the global view trades cluster-fidelity for breadth and warns the client via
+// `truncated`/`totalAvailable` when the ceiling clips the corpus.
+const MAX_NODES_GLOBAL = 2000;
 
 function snippet(content, len = NODE_LABEL_LEN) {
   if (!content) return '';
@@ -158,6 +162,60 @@ async function fetchProjectGraph(pool, projectName) {
   return { nodes, edges };
 }
 
+async function fetchAllGraph(pool) {
+  // Sprint 41 T3 — backs the "All projects" picker option in /graph.html.
+  // Returns the most-recent MAX_NODES_GLOBAL active+non-archived memories plus
+  // every edge whose endpoints both land in the result set. `totalAvailable`
+  // and `truncated` let the client surface a toast when the corpus overflows
+  // the cap.
+  const totalSql = `
+    SELECT COUNT(*)::int AS c
+      FROM memory_items
+     WHERE is_active = TRUE AND archived = FALSE
+  `;
+  const totalRes = await pool.query(totalSql);
+  const totalAvailable = Number(totalRes.rows[0]?.c || 0);
+
+  const nodesSql = `
+    WITH all_nodes AS (
+      SELECT ${NODE_COLUMNS_SQL}
+        FROM memory_items
+       WHERE is_active = TRUE AND archived = FALSE
+       ORDER BY created_at DESC
+       LIMIT ${MAX_NODES_GLOBAL}
+    )
+    SELECT
+      n.*,
+      COALESCE((
+        SELECT COUNT(*)::int
+          FROM memory_relationships r
+         WHERE r.source_id = n.id OR r.target_id = n.id
+      ), 0) AS degree
+      FROM all_nodes n
+  `;
+  const nodesRes = await pool.query(nodesSql);
+  const nodes = nodesRes.rows.map(rowToNode);
+  if (nodes.length === 0) {
+    return { nodes: [], edges: [], totalAvailable, truncated: false };
+  }
+  const ids = nodes.map((n) => n.id);
+
+  const edgesSql = `
+    SELECT ${EDGE_COLUMNS_BASE_SQL}, ${EDGE_COLUMNS_T2_SQL}
+      FROM memory_relationships
+     WHERE source_id = ANY($1::uuid[])
+       AND target_id = ANY($1::uuid[])
+  `;
+  const edgesRes = await pool.query(edgesSql, [ids]);
+  const edges = edgesRes.rows.map(rowToEdge);
+  return {
+    nodes,
+    edges,
+    totalAvailable,
+    truncated: totalAvailable > MAX_NODES_GLOBAL,
+  };
+}
+
 async function fetchNeighborhood(pool, rootId, depth) {
   // Inline recursive CTE so T4 ships independently of T1's
   // expand_memory_neighborhood RPC. When that RPC lands, the CTE here can be
@@ -467,6 +525,46 @@ function createGraphRoutes({ app, getPool }) {
     }
   });
 
+  // Sprint 41 T3 — All Projects view. Sibling of /api/graph/project/:name with
+  // no project filter; cap is MAX_NODES_GLOBAL and the response carries
+  // `truncated`/`totalAvailable` so the client can surface a "showing N of M"
+  // toast when the corpus overflows.
+  app.get('/api/graph/all', async (req, res) => {
+    const pool = getPool();
+    if (!pool) {
+      return res.json(disabledPayload({
+        nodes: [],
+        edges: [],
+        totalAvailable: 0,
+        truncated: false,
+      }));
+    }
+    try {
+      const { nodes, edges, totalAvailable, truncated } = await fetchAllGraph(pool);
+      const byType = {};
+      for (const e of edges) {
+        byType[e.kind] = (byType[e.kind] || 0) + 1;
+      }
+      res.json({
+        enabled: true,
+        stats: {
+          nodes: nodes.length,
+          edges: edges.length,
+          byType,
+          truncated,
+          totalAvailable,
+        },
+        nodes,
+        edges,
+        totalAvailable,
+        truncated,
+      });
+    } catch (err) {
+      console.warn('[graph] /api/graph/all failed:', err.message);
+      res.status(500).json({ error: 'graph all query failed', detail: err.message });
+    }
+  });
+
   app.get('/api/graph/memory/:id', async (req, res) => {
     const id = req.params.id;
     if (!UUID_RE.test(id)) {
@@ -540,6 +638,7 @@ module.exports = {
   createGraphRoutes,
   // Exported for tests + reuse:
   fetchProjectGraph,
+  fetchAllGraph,
   fetchNeighborhood,
   fetchStats,
   fetchInferenceStats,
@@ -550,6 +649,7 @@ module.exports = {
   UUID_RE,
   PROJECT_RE,
   MAX_NODES_PER_PROJECT,
+  MAX_NODES_GLOBAL,
   MAX_DEPTH,
   DEFAULT_DEPTH,
 };

package/packages/server/src/setup/mnestra-migrations/012_project_tag_re_taxonomy.sql

@@ -0,0 +1,397 @@
+-- Sprint 41 T2 — chopin-nashville re-taxonomy.
+--
+-- Why this exists:
+--   Sprint 39 T3 migration 011_project_tag_backfill.sql moved 192 rows out of
+--   the chopin-nashville bucket (1,139 → 947) using a conservative 5-bucket
+--   keyword pass. 947 rows still remain mis-tagged. Sprint 41 widens the
+--   keyword sets per the new project taxonomy (T1 owns the canonical doc at
+--   docs/PROJECT-TAXONOMY.md) and adds three buckets that 011 did not have:
+--     - chopin-in-bohemia (festival, distinct from the Nashville competition)
+--     - chopin-scheduler  (the SchedulingApp / "Maestro" project — single
+--                          project under two names per orchestrator
+--                          mid-inject clarification 2026-04-28 12:51 ET)
+--     - claimguard        (Gorgias-ticket-monitor / ClaimGuard-AI work)
+--
+--   The remaining residue after this migration is what Sprint 41 T4 hands to
+--   the LLM-classification runner. Conservative wins: rows with no clear
+--   keyword signal STAY chopin-nashville for T4 to handle.
+--
+-- What this migration does NOT do:
+--   - Does NOT touch mnestra_session_memory / mnestra_project_memory / etc.
+--     (legacy rag-events tables; different write path; Sprint 42+ cleanup).
+--   - Does NOT consolidate duplicate tags like 'gorgias' vs
+--     'gorgias-ticket-monitor', 'pvb' vs 'PVB', or 'mnestra' vs 'engram'.
+--     Visible in `SELECT project, count(*) FROM memory_items GROUP BY
+--     project` but a separate cleanup pass.
+--   - Does NOT re-tag rows whose only signal is the legitimate
+--     chopin-nashville vocabulary (competition / performance / jury /
+--     sponsor / applicant / repertoire / Acceptd / NICPC / laureate). Those
+--     are the rows the chopin-nashville tag SHOULD claim.
+--   - Does NOT touch source_session_id → transcript_path → cwd resolution.
+--     The briefing scoped that as a possible additional signal, not
+--     required. Keyword bucketing + T4's LLM pass hits the < 100 target.
+--
+-- Heuristic — content keyword bucketing:
+--   The migration runs UPDATEs sequentially. Earlier buckets claim ambiguous
+--   multi-project rows first; later buckets only see rows that no earlier
+--   bucket has already re-tagged. Order is broadest-first (largest expected
+--   bucket size first):
+--
+--     1. termdeck            — termdeck, mnestra, "4+1 sprint", xterm,
+--                              node-pty, flashback, memory_items,
+--                              memory_relationships
+--     2. rumen               — rumen, rumen-tick, "insight synthesis"
+--     3. podium              — podium
+--     4. chopin-in-bohemia   — bohemia, "chopin in bohemia", "2026 festival"
+--     5. chopin-scheduler    — scheduling, schedulingapp, \mMaestro\M
+--                              (Maestro is the working name; chopin-scheduler
+--                              is the canonical tag — alias confirmed 2026-04-28
+--                              by Joshua; case-sensitive word-boundary token
+--                              avoids matching unrelated "[maestro]" log
+--                              prefixes)
+--     6. pvb                 — PVB, petvetbid, "pet vet bid"
+--     7. claimguard          — claimguard, gorgias-ticket-monitor,
+--                              "gorgias ticket monitor"
+--     8. dor                 — \mDOR\M, /DOR/, ~/Documents/DOR, dor.config,
+--                              "Rust LLM gateway", openclaw
+--                              (reused verbatim from 011's tightened
+--                              pattern; word-boundary uppercase rules out
+--                              "dormant", "vendored", "indoor", etc.)
+--
+-- Idempotence:
+--   Every UPDATE is gated by `WHERE project = 'chopin-nashville'`. After the
+--   first run, those rows have a different project tag, so re-running this
+--   migration is a no-op (zero rows updated per bucket). RAISE NOTICE on a
+--   re-run will print zeros, which is the expected idempotent signal.
+--
+-- Application:
+--   THIS MIGRATION IS NOT EXECUTED BY THE LANE THAT WROTE IT. Orchestrator
+--   reviews the RAISE NOTICE counts after applying. Apply via the bundled
+--   migration runner at packages/server/src/setup/migration-runner.js (which
+--   uses node-postgres client.query — psql metacommands like \gset are NOT
+--   available, so the count probes use GET DIAGNOSTICS ROW_COUNT inside DO
+--   blocks). Manual fallback:
+--     `psql "$DATABASE_URL" -f 012_project_tag_re_taxonomy.sql`.
+
+BEGIN;
+
+-- ============================================================
+-- AUDIT BEFORE
+-- ============================================================
+DO $$
+DECLARE
+  before_chopin       int;
+  before_termdeck     int;
+  before_rumen        int;
+  before_podium       int;
+  before_bohemia      int;
+  before_scheduler    int;
+  before_pvb          int;
+  before_claimguard   int;
+  before_dor          int;
+BEGIN
+  SELECT count(*) INTO before_chopin     FROM memory_items WHERE project = 'chopin-nashville';
+  SELECT count(*) INTO before_termdeck   FROM memory_items WHERE project = 'termdeck';
+  SELECT count(*) INTO before_rumen      FROM memory_items WHERE project = 'rumen';
+  SELECT count(*) INTO before_podium     FROM memory_items WHERE project = 'podium';
+  SELECT count(*) INTO before_bohemia    FROM memory_items WHERE project = 'chopin-in-bohemia';
+  SELECT count(*) INTO before_scheduler  FROM memory_items WHERE project = 'chopin-scheduler';
+  SELECT count(*) INTO before_pvb        FROM memory_items WHERE project = 'pvb';
+  SELECT count(*) INTO before_claimguard FROM memory_items WHERE project = 'claimguard';
+  SELECT count(*) INTO before_dor        FROM memory_items WHERE project = 'dor';
+  RAISE NOTICE '[012-retaxonomy] BEFORE  chopin-nashville=% termdeck=% rumen=% podium=% chopin-in-bohemia=% chopin-scheduler=% pvb=% claimguard=% dor=%',
+    before_chopin, before_termdeck, before_rumen, before_podium, before_bohemia, before_scheduler, before_pvb, before_claimguard, before_dor;
+END $$;
+
+-- ============================================================
+-- BUCKET 1 — termdeck (broadest first; claims ambiguous multi-project rows)
+--
+-- Widened from 011's 3-keyword set [termdeck | mnestra | "4+1 sprint"] to
+-- include TermDeck-internal vocabulary that almost never appears outside the
+-- TermDeck stack (xterm, node-pty), the Flashback subsystem name, and the
+-- memory_* table identifiers (which are spoken about in TermDeck/Mnestra
+-- context overwhelmingly — graph-routes, mnestra-bridge, the migrations
+-- themselves).
+-- ============================================================
+DO $$
+DECLARE
+  rows_updated int;
+BEGIN
+  UPDATE memory_items SET project = 'termdeck'
+   WHERE project = 'chopin-nashville'
+     AND (
+       content ILIKE '%termdeck%'
+       OR content ILIKE '%mnestra%'
+       OR content ILIKE '%4+1 sprint%'
+       OR content ILIKE '%xterm%'
+       OR content ILIKE '%node-pty%'
+       OR content ILIKE '%flashback%'
+       OR content ILIKE '%memory_items%'
+       OR content ILIKE '%memory_relationships%'
+     );
+  GET DIAGNOSTICS rows_updated = ROW_COUNT;
+  RAISE NOTICE '[012-retaxonomy] bucket 1 (termdeck): % rows re-tagged', rows_updated;
+END $$;
+
+-- ============================================================
+-- BUCKET 2 — rumen
+--
+-- 011 used [rumen] alone. 012 widens to include rumen-tick (the Rumen
+-- cron-tick subsystem) and "insight synthesis" (Rumen's product vocabulary).
+-- ============================================================
+DO $$
+DECLARE
+  rows_updated int;
+BEGIN
+  UPDATE memory_items SET project = 'rumen'
+   WHERE project = 'chopin-nashville'
+     AND (
+       content ILIKE '%rumen%'
+       OR content ILIKE '%rumen-tick%'
+       OR content ILIKE '%insight synthesis%'
+     );
+  GET DIAGNOSTICS rows_updated = ROW_COUNT;
+  RAISE NOTICE '[012-retaxonomy] bucket 2 (rumen): % rows re-tagged', rows_updated;
+END $$;
+
+-- ============================================================
+-- BUCKET 3 — podium
+--
+-- Same single-keyword pattern as 011. Podium-specific vocabulary doesn't
+-- have synonyms that justify widening. (The Chopin in Bohemia festival
+-- mentions Podium often, but bucket 1's broadest-first ordering means
+-- podium-AND-bohemia rows where podium is the dominant tag claim it here;
+-- bohemia-only rows fall to bucket 4.)
+-- ============================================================
+DO $$
+DECLARE
+  rows_updated int;
+BEGIN
+  UPDATE memory_items SET project = 'podium'
+   WHERE project = 'chopin-nashville'
+     AND content ILIKE '%podium%';
+  GET DIAGNOSTICS rows_updated = ROW_COUNT;
+  RAISE NOTICE '[012-retaxonomy] bucket 3 (podium): % rows re-tagged', rows_updated;
+END $$;
+
+-- ============================================================
+-- BUCKET 4 — chopin-in-bohemia (NEW in 012)
+--
+-- The 2026 festival is a distinct project from the Chopin Nashville
+-- competition. Keywords: bohemia (substring; festival-specific), "chopin in
+-- bohemia" (full phrase, near-zero false positives), "2026 festival" (date+
+-- project disambiguator).
+--
+-- Note: rows that mention "Chopin Nashville" AND "Bohemia" together (rare —
+-- maybe cross-project planning notes) will already have been claimed by
+-- earlier buckets if they also mention TermDeck/Rumen/Podium tooling.
+-- Otherwise they land here, which is the right call: the "current festival
+-- being planned" is Bohemia 2026.
+-- ============================================================
+DO $$
+DECLARE
+  rows_updated int;
+BEGIN
+  UPDATE memory_items SET project = 'chopin-in-bohemia'
+   WHERE project = 'chopin-nashville'
+     AND (
+       content ILIKE '%bohemia%'
+       OR content ILIKE '%chopin in bohemia%'
+       OR content ILIKE '%2026 festival%'
+     );
+  GET DIAGNOSTICS rows_updated = ROW_COUNT;
+  RAISE NOTICE '[012-retaxonomy] bucket 4 (chopin-in-bohemia): % rows re-tagged', rows_updated;
+END $$;
+
+-- ============================================================
+-- BUCKET 5 — chopin-scheduler (NEW in 012; absorbs Maestro alias)
+--
+-- Per orchestrator clarification 2026-04-28 12:51 ET: "Maestro" is the
+-- working/branding name for the chopin-scheduler project. Same project,
+-- two names. The on-disk path is SchedulingApp/, so the keywords cover both
+-- the path-style identifier (scheduling, schedulingapp) and the branding
+-- alias (\mMaestro\M — POSIX word-boundary, case-sensitive Capitalized
+-- token).
+--
+-- The case-sensitive Maestro pattern matters: lowercase "maestro" can
+-- appear in unrelated content (log prefixes like "[maestro]" if any tool
+-- ever named itself that, generic music vocabulary). Capitalized Maestro
+-- with word boundaries is much closer to "the project name" intent.
+-- ============================================================
+DO $$
+DECLARE
+  rows_updated int;
+BEGIN
+  UPDATE memory_items SET project = 'chopin-scheduler'
+   WHERE project = 'chopin-nashville'
+     AND (
+       content ILIKE '%scheduling%'
+       OR content ILIKE '%schedulingapp%'
+       OR content ~ '\mMaestro\M'
+     );
+  GET DIAGNOSTICS rows_updated = ROW_COUNT;
+  RAISE NOTICE '[012-retaxonomy] bucket 5 (chopin-scheduler): % rows re-tagged', rows_updated;
+END $$;
+
+-- ============================================================
+-- BUCKET 6 — pvb (case-insensitive PVB / petvetbid markers)
+--
+-- Same pattern as 011 bucket 4. PVB is small in the chopin-nashville bucket
+-- (Sprint 39 dry-run found 7 rows; live apply landed 3 because bucket 1
+-- claimed mnestra-AND-PVB rows first). 012's earlier expansion of bucket 1
+-- means this stays small or zero.
+-- ============================================================
+DO $$
+DECLARE
+  rows_updated int;
+BEGIN
+  UPDATE memory_items SET project = 'pvb'
+   WHERE project = 'chopin-nashville'
+     AND (
+       content ILIKE '%PVB%'
+       OR content ILIKE '%petvetbid%'
+       OR content ILIKE '%pet vet bid%'
+     );
+  GET DIAGNOSTICS rows_updated = ROW_COUNT;
+  RAISE NOTICE '[012-retaxonomy] bucket 6 (pvb): % rows re-tagged', rows_updated;
+END $$;
+
+-- ============================================================
+-- BUCKET 7 — claimguard (NEW in 012)
+--
+-- ClaimGuard-AI is the active Unagi project (Joshua's roadmap shows it as
+-- the next 1-2 sprints after Sprint 41 ships). On-disk path is
+-- ~/Documents/Unagi/gorgias-ticket-monitor/. Keywords:
+--   - claimguard (substring; product name, near-zero false positives)
+--   - gorgias-ticket-monitor (the on-disk dir name; near-zero FP)
+--   - "gorgias ticket monitor" (the spoken-form variant)
+--
+-- The bare "gorgias" keyword is intentionally NOT used here because the
+-- pre-existing `gorgias` tag (468 rows) and `gorgias-ticket-monitor` tag
+-- (207 rows) are separate categories — bare "gorgias" content could be
+-- about Gorgias-the-helpdesk-product unrelated to ClaimGuard. The
+-- compound-token discipline keeps the bucket precise.
+-- ============================================================
+DO $$
+DECLARE
+  rows_updated int;
+BEGIN
+  UPDATE memory_items SET project = 'claimguard'
+   WHERE project = 'chopin-nashville'
+     AND (
+       content ILIKE '%claimguard%'
+       OR content ILIKE '%gorgias-ticket-monitor%'
+       OR content ILIKE '%gorgias ticket monitor%'
+     );
+  GET DIAGNOSTICS rows_updated = ROW_COUNT;
+  RAISE NOTICE '[012-retaxonomy] bucket 7 (claimguard): % rows re-tagged', rows_updated;
+END $$;
+
+-- ============================================================
+-- BUCKET 8 — dor (REUSED VERBATIM from 011's tightened pattern)
+--
+-- 011's audit found that the original briefing's `%dor%` ILIKE pattern
+-- produced ~33% false positives (matched "dormant", "vendored", "indoor",
+-- etc.). 011 tightened to:
+--   - POSIX word boundary `\mDOR\M` — case-sensitive uppercase only
+--   - path/identifier markers: /DOR/, ~/Documents/DOR, dor.config,
+--     "Rust LLM gateway" (DOR's tagline)
+--   - openclaw substring (OpenClaw is the slack-channel automation product
+--     that lives next to DOR in Joshua's stack)
+--
+-- 012 reuses this verbatim. After 011 caught 3 dor rows live, residue is
+-- expected to be near-zero — but the bucket stays in case any new rows
+-- accumulated post-Sprint-39 carry the markers.
+-- ============================================================
+DO $$
+DECLARE
+  rows_updated int;
+BEGIN
+  UPDATE memory_items SET project = 'dor'
+   WHERE project = 'chopin-nashville'
+     AND (
+       content ~ '\mDOR\M'
+       OR content ILIKE '%/DOR/%'
+       OR content ILIKE '%~/Documents/DOR%'
+       OR content ILIKE '%dor.config%'
+       OR content ILIKE '%Rust LLM gateway%'
+       OR content ILIKE '%openclaw%'
+     );
+  GET DIAGNOSTICS rows_updated = ROW_COUNT;
+  RAISE NOTICE '[012-retaxonomy] bucket 8 (dor): % rows re-tagged', rows_updated;
+END $$;
+
+-- ============================================================
+-- AUDIT AFTER
+-- ============================================================
+DO $$
+DECLARE
+  after_chopin       int;
+  after_termdeck     int;
+  after_rumen        int;
+  after_podium       int;
+  after_bohemia      int;
+  after_scheduler    int;
+  after_pvb          int;
+  after_claimguard   int;
+  after_dor          int;
+BEGIN
+  SELECT count(*) INTO after_chopin     FROM memory_items WHERE project = 'chopin-nashville';
+  SELECT count(*) INTO after_termdeck   FROM memory_items WHERE project = 'termdeck';
+  SELECT count(*) INTO after_rumen      FROM memory_items WHERE project = 'rumen';
+  SELECT count(*) INTO after_podium     FROM memory_items WHERE project = 'podium';
+  SELECT count(*) INTO after_bohemia    FROM memory_items WHERE project = 'chopin-in-bohemia';
+  SELECT count(*) INTO after_scheduler  FROM memory_items WHERE project = 'chopin-scheduler';
+  SELECT count(*) INTO after_pvb        FROM memory_items WHERE project = 'pvb';
+  SELECT count(*) INTO after_claimguard FROM memory_items WHERE project = 'claimguard';
+  SELECT count(*) INTO after_dor        FROM memory_items WHERE project = 'dor';
+  RAISE NOTICE '[012-retaxonomy] AFTER   chopin-nashville=% termdeck=% rumen=% podium=% chopin-in-bohemia=% chopin-scheduler=% pvb=% claimguard=% dor=%',
+    after_chopin, after_termdeck, after_rumen, after_podium, after_bohemia, after_scheduler, after_pvb, after_claimguard, after_dor;
+  RAISE NOTICE '[012-retaxonomy] Sprint 41 acceptance target: chopin-nashville drops 947 -> < 100 after T2+T4. T2 (this migration) handles deterministic keyword cases; T4 LLM-classifies the residue. If chopin-nashville count after this migration is still > 200, T4 has more rows to chew through; if < 100 already, T4 may have very little to do.';
+END $$;
+
+COMMIT;
+
+-- ============================================================
+-- POST-APPLY: optional verification queries (NOT part of the migration).
+-- Run separately to confirm the new taxonomy holds and to spot-check
+-- false-positive rates per bucket.
+-- ============================================================
+--
+-- 1. Tag distribution after migration:
+--    SELECT project, count(*) FROM memory_items
+--     GROUP BY project ORDER BY count(*) DESC LIMIT 20;
+--
+-- 2. Confirm no chopin-nashville rows match obvious termdeck/rumen/podium
+--    keywords (these should all return 0 if the migration succeeded):
+--    SELECT count(*) FROM memory_items
+--     WHERE project='chopin-nashville'
+--       AND (content ILIKE '%termdeck%' OR content ILIKE '%rumen%'
+--            OR content ILIKE '%podium%' OR content ILIKE '%bohemia%'
+--            OR content ILIKE '%scheduling%' OR content ILIKE '%claimguard%');
+--    -- Expected: 0
+--
+-- 3. Spot-check false-positive rate per bucket (replace 'termdeck' with
+--    each new tag in turn):
+--    SELECT id, left(content, 200) AS preview
+--    FROM memory_items
+--    WHERE project='termdeck' AND id IN (
+--      SELECT id FROM memory_items
+--      WHERE project='termdeck'
+--      ORDER BY updated_at DESC LIMIT 10
+--    );
+--
+-- 4. Confirm the legitimate-chopin-nashville signal is preserved (rows
+--    matching competition/laureate/applicant/Acceptd/NICPC/Bohemia/
+--    repertoire keywords should still be tagged chopin-nashville,
+--    EXCEPT for those that ALSO matched a code-project keyword and got
+--    legitimately re-tagged):
+--    SELECT count(*) FROM memory_items
+--     WHERE project='chopin-nashville'
+--       AND (content ILIKE '%competition%' OR content ILIKE '%laureate%'
+--            OR content ILIKE '%applicant%' OR content ILIKE '%Acceptd%'
+--            OR content ILIKE '%NICPC%' OR content ILIKE '%repertoire%'
+--            OR content ILIKE '%jury%');
+--    -- Expected: most of the residue (~71+ rows from Sprint 39 baseline,
+--    -- possibly higher as more legitimate competition content has
+--    -- accumulated since 2026-04-27).