Semantic search

backend/app/api/v1/messages.py

@@ -26,6 +26,8 @@ from sqlalchemy import func, or_, select
 from sqlalchemy.ext.asyncio import AsyncSession
 from sqlalchemy.orm import selectinload
 
+from app.services.search_service import SearchService
+
 from app.config import get_settings
 from app.core.exceptions import ForbiddenError, NotFoundError
 from app.database import get_db
@@ -184,8 +186,11 @@ async def list_messages(
     is_starred: Optional[bool] = Query(None),
     is_archived: Optional[bool] = Query(False),
     is_trashed: Optional[bool] = Query(False),
-    search: Optional[str] = Query(None, max_length=200),
+    search: Optional[str] = Query(None, max_length=500),
     pec_type: Optional[str] = Query(None),
+    # Filtri data (ISO 8601, es. 2026-01-01T00:00:00Z)
+    date_from: Optional[datetime] = Query(None, description="Data minima (received_at o sent_at)"),
+    date_to: Optional[datetime] = Query(None, description="Data massima (received_at o sent_at)"),
     # Paginazione
     page: int = Query(1, ge=1),
     page_size: int = Query(50, ge=1, le=200),
@@ -195,7 +200,8 @@ async def list_messages(
 
     - `is_archived=False` (default) esclude i messaggi archiviati.
     - `is_trashed=False` (default) esclude i messaggi nel cestino.
-    - `search` cerca su subject, from_address, to_addresses.
+    - `search` usa ricerca full-text (tsvector) con fallback ILIKE.
+    - `date_from` / `date_to` filtrano per data ricezione o invio.
     - `vbox_id` filtra per Virtual Box assegnata all'utente corrente.
     """
     # Determinare le caselle visibili (normale check permessi)
@@ -284,16 +290,30 @@ async def list_messages(
     if is_trashed is not None:
         q = q.where(Message.is_trashed == is_trashed)
 
+    # ── Full-text search (FTS con fallback ILIKE per messaggi non indicizzati) ───
     if search:
-        term = f"%{search}%"
+        from sqlalchemy import case as sa_case
+
+        tsquery = func.websearch_to_tsquery("italian", search)
+        term_like = f"%{search}%"
         q = q.where(
             or_(
-                Message.subject.ilike(term),
-                Message.from_address.ilike(term),
-                Message.body_text.ilike(term),
+                Message.search_vector.op("@@")(tsquery),
+                # Fallback per messaggi non ancora indicizzati dal worker
+                Message.search_vector.is_(None) & or_(
+                    Message.subject.ilike(term_like),
+                    Message.from_address.ilike(term_like),
+                    Message.body_text.ilike(term_like),
+                ),
             )
         )
 
+    # ── Filtri data ───────────────────────────────────────────────────────────
+    if date_from:
+        q = q.where(or_(Message.received_at >= date_from, Message.sent_at >= date_from))
+    if date_to:
+        q = q.where(or_(Message.received_at <= date_to, Message.sent_at <= date_to))
+
     # Applica le regole della Virtual Box (AND tra le regole)
     for rule in vbox_rules:
         q = _apply_vbox_rule(q, rule.field, rule.operator, rule.value)
@@ -302,13 +322,23 @@ async def list_messages(
     count_q = select(func.count()).select_from(q.subquery())
     total = (await db.execute(count_q)).scalar_one()
 
-    # Ordinamento e paginazione
+    # Ordinamento: se c'e' una ricerca, ordina per rilevanza FTS, poi data
+    if search:
+        from sqlalchemy import case as sa_case
+
+        tsquery_ord = func.websearch_to_tsquery("italian", search)
+        rank_expr = sa_case(
+            (Message.search_vector.isnot(None), func.ts_rank(Message.search_vector, tsquery_ord)),
+            else_=0.0,
+        )
+        order_clauses = [rank_expr.desc(), Message.received_at.desc().nullslast(), Message.created_at.desc()]
+    else:
+        order_clauses = [Message.received_at.desc().nullslast(), Message.created_at.desc()]
+
+    # Paginazione
     q = (
         q.options(selectinload(Message.labels))
-        .order_by(
-            Message.received_at.desc().nullslast(),
-            Message.created_at.desc(),
-        )
+        .order_by(*order_clauses)
         .offset((page - 1) * page_size)
         .limit(page_size)
     )