Duplicated-retriever

[Guts-Blood]

in memory_system.py line 473 and 489 there are duplicated call for self.retriever.search(query, k) and they should return the same results. Thus the later process should be useless. I think from your comments you should use another embedding retrieve methods in 489?

[onestardao]

You’re right to flag this.

A quick read suggests the two calls to self.retriever.search(query, k) (around 473 and 489) would return the same candidate set unless one of these is meant to:

use a different k or filter (e.g., namespace/index),

run with different scoring (cosine vs. dot / normalized vs. raw),

gate a second-pass re-rank on the first pass results, or

apply side effects (e.g., caching/metrics) between calls.

If none of the above is intended, the second call is likely redundant and could be replaced by reusing the first result (and, if needed, applying a distinct re-ranker). Do you expect pass-2 to differ (different index/embedding space or re-rank params)? If you share that intent, it’ll be easier to propose a small patch (reuse results + optional re-rank) without changing behavior.

[mustbeasaltyfish]

Yes，Me finding the same problem