Quality Control :(

[Faithfinder]

Some packs don't really meet even a low bar of quality standard (see https://openpeon.com/packs/arnold)

Need a mechanism for quality control. I can see how this can be a PITA though. Could do reports for moderation. Could do a ranking system for community self-regulation. What else?

[garysheng]

Yeah, that arnold pack is a good example of the problem — low-effort AI-generated content that degrades the registry experience.

Options I'm thinking about:

1. Report + moderation queue — lowest friction for submitters, highest friction for us. Probably right for early stage.
2. Community upvotes/downvotes — self-regulating over time, but needs the site to support it. Good long-term play.
3. Submission quality bar — enforce a minimum number of sounds, require a preview audio clip in the PR, maybe a description quality check.
4. Trust tiers with curatorial power — let vetted community members mark packs as featured vs unvetted, and surface that on openpeon.com.

Short-term I'm leaning toward: add a quality flag field to the registry schema (e.g. quality: 'unreviewed' | 'community_approved' | 'flagged'), default all new submissions to unreviewed, and let the site filter by default. That buys us time to build proper moderation.

What do you think — worth a small spec PR to openpeon to formalize the schema?

[garysheng]

Real concern — pack quality variance is visible. A lightweight path forward:

Short-term (MVP):

• Add a quality_tier field to the registry schema (e.g. official, community, experimental)
• Add a GitHub issue template for flagging low-quality packs, handled by maintainers

Medium-term (tracked in openpeon#20):

• Ratings on openpeon.com: minimal thumbs-up/down stored per pack, anonymous/open. No account required.
• Sort by rating on the pack list page

Not doing:

• Gatekeeping submissions through a review queue — that slows the commons and concentrates power with maintainers. Better to ship everything + let the community surface the good ones.

PR welcome if anyone wants to prototype the ratings endpoint.