Merge branch 'main' into rag-metadata-support

llama_stack/providers/inline/tool_runtime/rag/memory.py

@@ -106,7 +106,9 @@ class MemoryToolRuntimeImpl(ToolsProtocolPrivate, ToolRuntime, RAGToolRuntime):
         query_config: RAGQueryConfig | None = None,
     ) -> RAGQueryResult:
         if not vector_db_ids:
-            return RAGQueryResult(content=None)
+            raise ValueError(
+                "No vector DBs were provided to the knowledge search tool. Please provide at least one vector DB ID."
+            )
 
         query_config = query_config or RAGQueryConfig()
         query = await generate_rag_query(

llama_stack/providers/remote/vector_io/chroma/chroma.py

@@ -26,8 +26,7 @@ from .config import ChromaVectorIOConfig as RemoteChromaVectorIOConfig
 
 log = logging.getLogger(__name__)
 
-
-ChromaClientType = chromadb.AsyncHttpClient | chromadb.PersistentClient
+ChromaClientType = chromadb.api.AsyncClientAPI | chromadb.api.ClientAPI
 
 
 # this is a helper to allow us to use async and non-async chroma clients interchangeably