fix(backend): handle reasoning event correctly

backend/src/inference.py

@@ -318,7 +318,7 @@ async def openrouter_chat_completion(app, messages, model, tools=[], opts={}, ag
         if name == 'reasoning':
             if reasoning_stream == None:
                 reasoning_stream = Stream(app)
-                agent_run.emit_event('output', reasoning_stream)
+                agent_run.emit_event('reasoning', reasoning_stream)
             reasoning_stream.feed(value)
         if name == 'tool-call':
             agent_run.emit_event('tool-call', value)