that's fine though, people aren't mad about models you can run locally, even though now it takes 30 seconds instead of 5 to get a response from my ERP bot.