fix: Fix parallel execution of alembic with SQlite (#5404)

Fix parallel execution of alembic with SQlite

src/backend/base/langflow/alembic/env.py

@@ -2,6 +2,7 @@ from logging.config import fileConfig
 
 from alembic import context
 from sqlalchemy import engine_from_config, pool, text
+from sqlalchemy.event import listen
 
 from langflow.services.database.service import SQLModel
 
@@ -51,6 +52,21 @@ def run_migrations_offline() -> None:
         context.run_migrations()
 
 
+def _sqlite_do_connect(
+    dbapi_connection,
+    connection_record,  # noqa: ARG001
+):
+    # disable pysqlite's emitting of the BEGIN statement entirely.
+    # also stops it from emitting COMMIT before any DDL.
+    dbapi_connection.isolation_level = None
+
+
+def _sqlite_do_begin(conn):
+    # emit our own BEGIN
+    conn.exec_driver_sql("PRAGMA busy_timeout = 60000")
+    conn.exec_driver_sql("BEGIN EXCLUSIVE")
+
+
 def run_migrations_online() -> None:
     """Run migrations in 'online' mode.
 
@@ -64,6 +80,11 @@ def run_migrations_online() -> None:
         poolclass=pool.NullPool,
     )
 
+    if connectable.dialect.name == "sqlite":
+        # See https://docs.sqlalchemy.org/en/20/dialects/sqlite.html#serializable-isolation-savepoints-transactional-ddl
+        listen(connectable, "connect", _sqlite_do_connect)
+        listen(connectable, "begin", _sqlite_do_begin)
+
     with connectable.connect() as connection:
         context.configure(connection=connection, target_metadata=target_metadata, render_as_batch=True)