bug fix, generate openapi spec

llama_stack/distribution/routers/routers.py

@@ -440,7 +440,7 @@ class ToolRuntimeRouter(ToolRuntime):
     ) -> None:
         self.routing_table = routing_table
 
-        # TODO: this should be in sync with "get_all_api_endpoints()"
+        # HACK ALERT this should be in sync with "get_all_api_endpoints()"
         # TODO: make sure rag_tool vs builtin::memory is correct everywhere
         self.rag_tool = self.RagToolImpl(routing_table)
         setattr(self, "rag_tool.query_context", self.rag_tool.query_context)

llama_stack/distribution/server/endpoints.py

@@ -38,6 +38,8 @@ def get_all_api_endpoints() -> Dict[Api, List[ApiEndpoint]]:
     for api, protocol in protocols.items():
         endpoints = []
         protocol_methods = inspect.getmembers(protocol, predicate=inspect.isfunction)
+
+        # HACK ALERT
         if api == Api.tool_runtime:
             for tool_group in SpecialToolGroup:
                 sub_protocol = toolgroup_protocols[tool_group]

llama_stack/providers/inline/agents/meta_reference/agent_instance.py

@@ -970,7 +970,7 @@ async def execute_tool_call_maybe(
 
     result = await tool_runtime_api.invoke_tool(
         tool_name=name,
-        args=dict(
+        kwargs=dict(
             session_id=session_id,
             **tool_call_args,
         ),