docs(kb): document retrieve() payload contract and kb_id precedence

The method had no docstring, so the precedence between `kb_names` and the
`kb_id` path parameter was only discoverable by reading the body. Spell out
that an explicit `kb_names` wins, which is what keeps the legacy multi-base
`POST /api/kb/retrieve` endpoint working.

Author: JosephTian876

astrbot/dashboard/services/knowledge_base_service.py

@@ -761,6 +761,26 @@ async def list_chunks_from_dashboard_query(
         )
 
     async def retrieve(self, data: object) -> dict[str, Any]:
+        """Retrieve chunks from one or more knowledge bases.
+
+        Args:
+            data: Request payload. ``query`` is required. The knowledge bases
+                are taken from ``kb_names`` when present, which is how the
+                legacy ``POST /api/kb/retrieve`` endpoint targets several bases
+                at once; otherwise the ``kb_id`` that the REST route reads from
+                its path is resolved to the matching name. ``top_k`` defaults
+                to 5, and ``debug`` additionally renders a t-SNE plot.
+
+        Returns:
+            A dict with the ``results`` list, its ``total`` and the ``query``
+            that produced it, plus ``visualization`` (or ``visualization_error``)
+            when ``debug`` is set.
+
+        Raises:
+            KnowledgeBaseServiceError: If ``query`` is missing, the given
+                ``kb_id`` does not exist, or neither ``kb_names`` nor ``kb_id``
+                identifies a knowledge base.
+        """
         payload = self._payload(data)
         query = payload.get("query")
         kb_names = payload.get("kb_names")