Commit

From what the commit approval speed tells and its potential demands, does it use jev-like LLMs for it?

OpenAI for now: - omni-moderation-latest — scores text and images for unsafe content. - gpt-6.1-sol — the configured model name for additional harmful-text and minors-in-images checks.

Thanks for the detailed summary! Very interesting stack.

omni-moderation is neat because it can do both text and images, and it's free (rate-limited but /dev/null seems to fit within limits)