diff --git a/migrations/versions/b5ac48010165_add_missing_document_metadata_column_to_.py b/migrations/versions/b5ac48010165_add_missing_document_metadata_column_to_.py
new file mode 100644
index 000000000..c0f8d6b0c
--- /dev/null
+++ b/migrations/versions/b5ac48010165_add_missing_document_metadata_column_to_.py
@@ -0,0 +1,50 @@
+"""add missing document_metadata column to node and cre
+
+Revision ID: b5ac48010165
+Revises: d4e5f6a7b8c9
+Create Date: 2026-07-24 22:19:01.724833
+
+"""
+
+from alembic import op
+import sqlalchemy as sa
+from sqlalchemy import inspect
+
+
+# revision identifiers, used by Alembic.
+revision = "b5ac48010165"
+down_revision = "d4e5f6a7b8c9"
+branch_labels = None
+depends_on = None
+
+
+def upgrade():
+    # Defensive: some environments (e.g. production) already have this column
+    # applied out-of-band without a corresponding migration ever being
+    # committed, so this must not assume a clean "column doesn't exist" state.
+    inspector = inspect(op.get_bind())
+    node_columns = {c["name"] for c in inspector.get_columns("node")}
+    cre_columns = {c["name"] for c in inspector.get_columns("cre")}
+
+    if "document_metadata" not in node_columns:
+        with op.batch_alter_table("node", schema=None) as batch_op:
+            batch_op.add_column(
+                sa.Column("document_metadata", sa.JSON(), nullable=True)
+            )
+
+    if "document_metadata" not in cre_columns:
+        with op.batch_alter_table("cre", schema=None) as batch_op:
+            batch_op.add_column(
+                sa.Column("document_metadata", sa.JSON(), nullable=True)
+            )
+
+
+def downgrade():
+    # Intentional no-op: upgrade() only adds this column where it was
+    # missing, so on environments where it pre-existed (e.g. production,
+    # applied out-of-band) this migration never created it. Unconditionally
+    # dropping it here would destroy that pre-existing data on downgrade.
+    # There is no reliable way to tell "did *this* migration add the
+    # column" apart from "did it already exist," so the safe choice is to
+    # leave the column alone rather than risk deleting real data.
+    pass