fix(sync): guard empty project list in watch loop

When constrained_project names a project not in the DB, or every active
project is filtered out by cloud-mode rules, _select_projects_to_watch()
returns []. watchfiles.awatch() raises ValueError on an empty path list,
which the outer handler catches with a 5-second sleep — producing a
tight error-log loop until a project appears.

Addresses review feedback on #759: sleep 30 s and retry instead, giving
the project list time to populate (e.g. after the user adds a project).
Logs include the constrained_project value so misconfiguration is easy
to diagnose.

Signed-off-by: phernandez <paul@basicmachines.co>

Author: phernandez

src/basic_memory/sync/watch_service.py

@@ -211,6 +211,21 @@ async def run(self):  # pragma: no cover
 
                 projects = await self._select_projects_to_watch()
 
+                # Trigger: no projects selected (e.g. constrained_project names a
+                #          project not in the DB, or every project was filtered out)
+                # Why: watchfiles.awatch() requires at least one path. Calling it
+                #      with an empty list raises ValueError, which the outer handler
+                #      catches with a 5s sleep — producing a tight error-log loop.
+                # Outcome: sleep longer and retry, giving the project list a chance
+                #          to populate (e.g. after a project is added).
+                if not projects:
+                    logger.warning(
+                        "No projects to watch; sleeping before retry "
+                        f"(constrained_project={self.constrained_project!r})"
+                    )
+                    await asyncio.sleep(30)
+                    continue
+
                 project_paths = [project.path for project in projects]
                 logger.debug(f"Starting watch cycle for directories: {project_paths}")