Is gathering all text being generated appropriate

[turtle0x1]

It seems gathering all user input text is now be gathered as of this commit? eebdcc6

Adding analytics is annoying because we'll have to patch them out / add more firewall rules, but siphoning off user data like this feels very wrong if I've understood it correctly.

This could leak sensitive business information if meta voice was used as part of some kind of internal system, please considere adding a --disable-metrics flag/env or removing the gathering of input text altogether (the other stuff is "by the by" - if you just want trends you shouldn't need "text").

[vatsalaggarwal]

Telemetry is optional. You can set ANONYMIZED_TELEMETRY=False in an .env file at the root level of this repo. We provide these instructions when you run inference, and they are additionally provided in a README here: https://github.com/metavoiceio/metavoice-src/tree/main/fam/telemetry

[vatsalaggarwal]

(The text is useful for us to debug errors folks are facing, etc, but if more people complain we can look into options... fyi @sidroopdaska)

[turtle0x1]

Still will have to patch it out, taking the "text" is just weird - what if we put sensitive business numbers with our company name or user information through this? It shouldn't leave the network "anonymized" or otherwise.

[MethanJess]

@vatsalaggarwal I really don't see how collecting the text can help "debug errors"
If someone is facing a problem with text, then they can just manually report it :/
Even then, I think this has more bad than good to it.

[sidroopdaska]

sure thing @MethanJess & @turtle0x1. I'll add a flag to avoid telemetry altogether. Will share a PR tomorrow.