diff --git a/RMuseum/RMuseum.xml b/RMuseum/RMuseum.xml
index 6dbdb2d8..1b6ad8f9 100644
--- a/RMuseum/RMuseum.xml
+++ b/RMuseum/RMuseum.xml
@@ -23142,6 +23142,36 @@
one must not be able to disable the other.
+
+
+ Common Persian function words stripped out before keyword-matching a query against
+ verse text (see SelectPreviewVerses) — the kind of words that appear in nearly every
+ query regardless of topic ("شعری در مورد ... پیدا کن") and would otherwise match
+ almost any couplet in almost any poem, defeating the whole point of looking for a
+ RELEVANT couplet rather than an arbitrary one. Not remotely exhaustive Persian
+ stopword coverage — just the words that actually show up in how people phrase this
+ kind of query, extended as real queries reveal gaps.
+
+
+
+
+ Splits the query on whitespace (deliberately NOT on ZWNJ — "بیوفایی" should survive
+ as one token, not fracture into "بی" + "وفایی", where "بی" alone is a common enough
+ prefix to false-positive-match all over the place), strips surrounding punctuation,
+ drops stopwords and anything too short to be a meaningful keyword on its own.
+
+
+
+
+ Looks for the couplet (Right/Left verse pair) whose combined text contains the most
+ query keywords, and returns it (plus, if there's room within previewVerseCount, the
+ couplet immediately following it, for a little reading continuity rather than a
+ single isolated pair). Falls back to the poem's opening verses — the previous,
+ always-the-same-lines behavior — if no keyword appears anywhere in the scanned
+ verses, or if there were no real keywords to search for at all (a query that was
+ entirely stopwords, or empty after stripping them).
+
+
Walks the whole category subtree rooted at rootCatId (breadth-first, level by level)