* add integration tests for LanguageModelChat#sendRequest * make sure errors are recreated when making LM requests * disable test with a note for later * fix remote integration tests