From ae7d2289e8dc1d8936046651e8f825e7eff52d15 Mon Sep 17 00:00:00 2001
From: Trenton Holmes <797416+stumpylog@users.noreply.github.com>
Date: Sat, 29 Aug 2026 15:06:47 -0700
Subject: [PATCH] sort similar documents by weight, use unrestricted-user check
 consistently (existed before, but good to be consistent)

---
 src/paperless_ai/ai_classifier.py | 16 +++++++++-------
 src/paperless_ai/taxonomy.py      | 12 ++++++++----
 2 files changed, 17 insertions(+), 11 deletions(-)

diff --git a/src/paperless_ai/ai_classifier.py b/src/paperless_ai/ai_classifier.py
index f86dbfe32..e52a7cd7f 100644
--- a/src/paperless_ai/ai_classifier.py
+++ b/src/paperless_ai/ai_classifier.py
@@ -204,18 +204,20 @@ def get_taxonomy_context(
     ai_config = AIConfig()
     try:
         if ai_config.llm_embedding_backend:
-            # None means "no restriction" to retrieve_similar_nodes. A superuser
-            # (like no user at all) can see every document, so skip materializing
-            # every visible pk into a Python list and passing it through as an IN
-            # filter: for a large library that is a wasted quadratic scan in the
-            # vector store at best, and past ~32,763 documents a hard
-            # sqlite3.OperationalError (SQLite's bound-parameter limit) at worst.
+            # None means "no restriction" to retrieve_similar_nodes. An
+            # unrestricted user (no user at all, or an active superuser -- see
+            # user_is_unrestricted) can see every document, so skip
+            # materializing every visible pk into a Python list and passing it
+            # through as an IN filter: for a large library that is a wasted
+            # quadratic scan in the vector store at best, and past ~32,763
+            # documents a hard sqlite3.OperationalError (SQLite's
+            # bound-parameter limit) at worst.
             # permitted_object_ids() has its own superuser shortcut that would
             # return every Document's id anyway, so this changes nothing about
             # which documents are considered -- only how we get there.
             visible_document_ids = (
                 None
-                if user is None or user.is_superuser
+                if user_is_unrestricted(user)
                 else list(permitted_object_ids(user, Document, "view_document"))
             )
             nodes = retrieve_similar_nodes(
diff --git a/src/paperless_ai/taxonomy.py b/src/paperless_ai/taxonomy.py
index dedf0fdba..dd599691e 100644
--- a/src/paperless_ai/taxonomy.py
+++ b/src/paperless_ai/taxonomy.py
@@ -126,10 +126,14 @@ def _node_document_weights(nodes: list["NodeWithScore"]) -> list[SimilarDocument
             weights[int(document_id)] += float(node.score or 0.0)
         except (TypeError, ValueError):  # pragma: no cover
             continue
-    return [
-        SimilarDocument(document_id=document_id, weight=weight)
-        for document_id, weight in weights.items()
-    ]
+    return sorted(
+        (
+            SimilarDocument(document_id=document_id, weight=weight)
+            for document_id, weight in weights.items()
+        ),
+        key=lambda similar: similar["weight"],
+        reverse=True,
+    )
 
 
 def _visible_ranked_candidates(