[Code Review] A Few Small Improvements in ranking_scorer.rs & param.rs #196

Hi team! 👋

Really appreciate the transparency of open-sourcing this. While reading through the code, I spotted a few small things that could be worth tidying up. Happy to send a PR for any of these if helpful!

1. param.rs — File appears truncated at the end

The last param!() call for OonWeightFactor seems to be missing its closing );. Just want to flag it in case it was a publish artifact.

2. ranking_scorer.rs — "pdwell" key in applied_weights_map may be pointing to the wrong field

It currently maps to self.post_unexplored. I think it was meant to map to self.cont_dwell_time — happy to double-check and fix if confirmed.

3. ranking_scorer.rs — Negative-action weights have no sign convention guard

Since negative_sum negates the block/report/mute weights, a small debug_assert! or comment clarifying that these params are expected as positive magnitudes would help prevent accidental misconfiguration down the line.

4. param.rs — ShadowTrafficPhoenixClusterRates typed as Vec<String>

Since these are numeric traffic rates (similar to the f64 typed ShadowTrafficDefaultPercent), changing this to Vec<f64> would give a bit more type safety and avoid manual parsing at call-sites.

5. ranking_scorer.rs — bidirectional_boost_eligible may want to cover quote-tweets too

Replies and retweets are excluded from the boost — wondering if quote-tweets (quoted_tweet_id) should follow the same logic for consistency. If intentional, a short comment would make the design decision clearer.

Thanks for your time — feel free to close if any of these are already on your radar! 🙏