If you want software that analyses an image and produces tags that describe the image’s content, you want a combination of image recognition, image interpretation, text generation and a promp generator to link these three parts. There is no way that any single of these applications can be “lightweight”, not to speak of the combination.

Different providers have different spamfilters, different rules regarding html mails, attachment file size, use of tls, policies regarding exspired certificates, and might have different log in procedures, so yes, if there are problems the question which providers are involved.