diff --git a/src/embeddings.js b/src/embeddings.js
index 4e87171..dbfb5fb 100644
--- a/src/embeddings.js
+++ b/src/embeddings.js
@@ -108,99 +108,6 @@ function padEmbeddingForVec0(values) {
   return Buffer.from(padded.buffer);
 }
 
-// backfill store from vec0 for any embeddings that predate multi-model support.
-// only runs when store is completely empty, so we never stamp the wrong model on existing data.
-try {
-  const storeEmpty = !db.prepare(`SELECT 1 FROM article_embedding_store LIMIT 1`).get();
-
-  if (storeEmpty) {
-    // use model from meta if available, fall back to config model
-    const metaRow = db.prepare(`SELECT model FROM article_embedding_meta LIMIT 1`).get();
-    const legacyModel = (metaRow && metaRow.model) || EMBEDDING_MODEL;
-
-    db.prepare(`
-      INSERT OR IGNORE INTO article_embedding_store (article_id, model, embedding)
-      SELECT e.article_id, ?, e.embedding FROM article_embeddings e
-    `).run(legacyModel);
-  }
-} catch (err) {
-  console.error('embedding store backfill failed:', err);
-}
-
-
-// if the config model changed, rebuild the vec0 search index from store.
-// only proceeds if the store actually has embeddings for the new model.
-function rebuildVec0IfModelChanged() {
-  const stale = db.prepare(`
-    SELECT 1 FROM article_embedding_meta WHERE model != ? LIMIT 1
-  `).get(EMBEDDING_MODEL);
-
-  // also sync any store entries that didn't make it into vec0
-  const missing = db.prepare(`
-    SELECT article_id, embedding FROM article_embedding_store
-    WHERE model = ?
-      AND NOT EXISTS (SELECT 1 FROM article_embeddings WHERE article_id = article_embedding_store.article_id)
-  `).all(EMBEDDING_MODEL);
-
-  if (missing.length > 0) {
-    const insertVec = db.prepare(`INSERT OR IGNORE INTO article_embeddings (article_id, embedding) VALUES (?, ?)`);
-    const insertMeta = db.prepare(`INSERT OR IGNORE INTO article_embedding_meta (article_id, model) VALUES (?, ?)`);
-
-    const sync = db.transaction(() => {
-      for (const row of missing) {
-        const vals = new Float32Array(row.embedding.buffer, row.embedding.byteOffset, row.embedding.byteLength / 4);
-        insertVec.run(BigInt(row.article_id), padEmbeddingForVec0(vals));
-        insertMeta.run(row.article_id, EMBEDDING_MODEL);
-      }
-    });
-
-    sync();
-    console.log(`synced ${missing.length} store embeddings into vec0`);
-  }
-
-  if (!stale) return;
-
-  const storeCount = db.prepare(`
-    SELECT COUNT(*) AS n FROM article_embedding_store WHERE model = ?
-  `).get(EMBEDDING_MODEL);
-
-  if (!storeCount || storeCount.n === 0) {
-    console.log(`embedding model changed to ${EMBEDDING_MODEL} but store has no embeddings for it yet — skipping index rebuild`);
-    return;
-  }
-
-  console.log(`embedding model changed to ${EMBEDDING_MODEL}, rebuilding search index...`);
-
-  db.exec('BEGIN');
-
-  try {
-    db.exec('DELETE FROM article_embeddings');
-    db.exec('DELETE FROM article_embedding_meta');
-
-    const rows = db.prepare(`
-      SELECT article_id, embedding FROM article_embedding_store WHERE model = ?
-    `).all(EMBEDDING_MODEL);
-
-    const insertVec = db.prepare(`INSERT INTO article_embeddings (article_id, embedding) VALUES (?, ?)`);
-    const insertMeta = db.prepare(`INSERT INTO article_embedding_meta (article_id, model) VALUES (?, ?)`);
-
-    for (const row of rows) {
-      const vals = new Float32Array(row.embedding.buffer, row.embedding.byteOffset, row.embedding.byteLength / 4);
-      const padded = new Float32Array(VEC0_DIM);
-      padded.set(vals);
-      insertVec.run(BigInt(row.article_id), Buffer.from(padded.buffer));
-      insertMeta.run(row.article_id, EMBEDDING_MODEL);
-    }
-
-    db.exec('COMMIT');
-    console.log(`rebuilt search index with ${rows.length} embeddings`);
-  } catch (err) {
-    db.exec('ROLLBACK');
-    throw err;
-  }
-}
-
-rebuildVec0IfModelChanged();
 
 let embeddingBackfillRunning = false;
 const embeddingJobsRunning = new Set();