Transactions

vectlite supports atomic transactions for batched writes. All operations within a transaction are committed together or rolled back on failure.

Python

Use the context manager for automatic commit/rollback:

with db.transaction() as tx:
    tx.upsert("doc1", emb1, {"source": "a"})
    tx.upsert("doc2", emb2, {"source": "b"})
    tx.delete("old_doc")
# Commits atomically on exit; rolls back on exception

Manual commit/rollback:

tx = db.transaction()
try:
    tx.upsert("doc1", emb1, {"source": "a"})
    tx.insert_many([
        {"id": "d2", "vector": emb2, "metadata": {"source": "b"}},
        {"id": "d3", "vector": emb3, "metadata": {"source": "c"}},
    ])
    tx.commit()
except Exception:
    tx.rollback()
    raise

Node.js

const tx = db.transaction()
try {
  tx.upsert('doc1', emb1, { source: 'a' })
  tx.upsert('doc2', emb2, { source: 'b' })
  tx.delete('old_doc')
  tx.commit()
} catch (err) {
  tx.rollback()
  throw err
}

Batch Operations

Transactions support bulk inserts and upserts:

with db.transaction() as tx:
    count = tx.insert_many([
        {"id": "d1", "vector": emb1, "metadata": {"source": "a"}},
        {"id": "d2", "vector": emb2, "metadata": {"source": "b"}},
    ])
    print(f"Inserted {count} records")

Transaction Size

You can check how many operations are queued:

with db.transaction() as tx:
    tx.upsert("doc1", emb1)
    tx.upsert("doc2", emb2)
    print(len(tx))  # 2