Hi, I work for a bank, and i'm currently installing promptfoo in order to test some internal chatbots, by policy we are not allowed to share information with external sources, and when i run the red team module with some plugings 90% failed because firewall block request to a.promptfoo.app and promptfoo.app. Why is this? There is anyway to avoid this?

Hi! Not all plugins are bundled with the OSS version. There's an on-prem enterprise version for running Promptfoo behind a firewall: https://www.promptfoo.dev/docs/enterprise/

In order to test, i'm using the community version, what kind of data Will be sent out of my network?

This is the only kind of data we collect: https://www.promptfoo.dev/docs/configuration/telemetry/

Thanks!

@josemakampa Apologies - I wasn't fully clear. The results of the tests do get sent to our servers for grading plus whatever is in the purpose for test generation

There is any way to avoid this?

hi josema, a couple things here - if you bring your own OpenAI API key, it will be used for grading instead of Promptfoo's graders - however, promptfoo servers are still used to generate the initial attacks based on the application details you provided - if you want to fully airgap the solution and not use promptfoo servers, you would need to use the Enterprise On Prem version

Hi all, quickly update, yesterday i used promptfoo and i tunelized the traffic using BurpSuite, using the option for disable telemetry and traffic i observed that promptfott dont sent any sensitive data out of the network

Appreciate the update, thank you!