fix(search): hybrid keyword + semantic search — exact title/body matches rank first

src/fabledassistant/services/knowledge.py

@@ -122,34 +122,74 @@ async def _semantic_knowledge_search(
     limit: int,
     offset: int,
 ) -> tuple[list[dict], int]:
-    """Semantic search over knowledge objects, with SQL filters applied post-rank."""
+    """Hybrid search: keyword matches first (title/body ILIKE), then semantic results.
+
+    Exact keyword matches always rank above semantic-only matches so that
+    searching for a name like "Weston" surfaces the note with that title
+    before conceptually related notes.
+    """
+    # 1. Keyword search — title and body ILIKE
+    keyword_notes: list[Note] = []
+    try:
+        async with async_session() as session:
+            pattern = f"%{q}%"
+            base = (
+                select(Note)
+                .where(Note.user_id == user_id)
+                .where(Note.title.ilike(pattern) | Note.body.ilike(pattern))
+            )
+            if note_type == "task":
+                base = base.where(Note.status.isnot(None))
+            elif note_type:
+                base = base.where(Note.note_type == note_type).where(Note.status.is_(None))
+            for tag in tags:
+                base = base.where(Note.tags.contains([tag]))
+            # Title matches first, then body-only matches, newest first within each
+            base = base.order_by(
+                Note.title.ilike(pattern).desc(),
+                Note.updated_at.desc(),
+            ).limit(limit * 2)
+            keyword_notes = list((await session.execute(base)).scalars().all())
+    except Exception:
+        logger.warning("Keyword search failed", exc_info=True)
+
+    # 2. Semantic search — conceptual similarity
+    semantic_notes: list[Note] = []
     try:
         from fabledassistant.services.embeddings import semantic_search_notes
-        # Fetch a larger candidate set to allow for filtering
         is_task_filter = True if note_type == "task" else (False if note_type else None)
         candidates = await semantic_search_notes(
             user_id=user_id,
             query=q,
-            limit=min(200, limit * 8),
+            limit=min(200, limit * 4),
             threshold=0.3,
             is_task=is_task_filter,
         )
+        for _score, note in candidates:
+            if note_type == "task" and not note.is_task:
+                continue
+            elif note_type and note_type != "task" and note.entity_type != note_type:
+                continue
+            if tags and not all(t in (note.tags or []) for t in tags):
+                continue
+            semantic_notes.append(note)
     except Exception:
-        logger.warning("Semantic search unavailable, falling back to SQL", exc_info=True)
+        logger.warning("Semantic search unavailable, using keyword results only", exc_info=True)
-        return await query_knowledge(user_id, note_type, tags, "modified", None, limit, offset)
 
-    results = []
+    # 3. Merge — keyword matches first, then semantic (deduplicated)
-    for _score, note in candidates:
+    seen_ids: set[int] = set()
-        if note_type == "task" and not note.is_task:
+    merged: list[Note] = []
-            continue
+    for note in keyword_notes:
-        elif note_type and note_type != "task" and note.entity_type != note_type:
+        if note.id not in seen_ids:
-            continue
+            seen_ids.add(note.id)
-        if tags and not all(t in (note.tags or []) for t in tags):
+            merged.append(note)
-            continue
+    for note in semantic_notes:
-        results.append(note)
+        if note.id not in seen_ids:
+            seen_ids.add(note.id)
+            merged.append(note)
 
-    total = len(results)
+    total = len(merged)
-    page_items = results[offset: offset + limit]
+    page_items = merged[offset: offset + limit]
     return [_note_to_item(n) for n in page_items], total