From 959bce47b21fb18e1915522b191d5ae39bc998a6 Mon Sep 17 00:00:00 2001
From: Raphael Southall <69826455+raphasouthall@users.noreply.github.com>
Date: Sat, 22 Aug 2026 13:19:54 +0100
Subject: [PATCH] fix(synthesize): drop blank member tags from the consolidated
 learning

Live run #1830 inherited an empty-string tag from an old harvest row;
_cluster_tags now strips and skips blank or non-string tags.
---
 src/neurostack/synthesize.py | 3 ++-
 tests/test_synthesize.py     | 4 ++--
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/src/neurostack/synthesize.py b/src/neurostack/synthesize.py
index 3e5154b..ac91d14 100644
--- a/src/neurostack/synthesize.py
+++ b/src/neurostack/synthesize.py
@@ -209,7 +209,8 @@ def _cluster_tags(members: list[dict], limit: int = 8) -> list[str]:
     for m in members:
         try:
             for t in json.loads(m.get("tags") or "[]"):
-                if not t.startswith(_SUPERSEDED_PREFIX):
+                t = t.strip() if isinstance(t, str) else ""
+                if t and not t.startswith(_SUPERSEDED_PREFIX):
                     counts[t] = counts.get(t, 0) + 1
         except (TypeError, ValueError):
             continue
diff --git a/tests/test_synthesize.py b/tests/test_synthesize.py
index c7f7450..f1e513d 100644
--- a/tests/test_synthesize.py
+++ b/tests/test_synthesize.py
@@ -80,8 +80,8 @@ def test_strip_fences(self):
 
     def test_cluster_tags_union_ranked_and_marked(self):
         members = [
-            {"tags": json.dumps(["aks", "azure", "superseded_by:9"])},
-            {"tags": json.dumps(["aks"])},
+            {"tags": json.dumps(["aks", "azure", "superseded_by:9", ""])},
+            {"tags": json.dumps(["aks", "  "])},
             {"tags": None},
         ]
         tags = _cluster_tags(members)