feat(ai-core): trigger memory extraction after each conversation round

The WS hub and platform bridge now call Pipeline.FinishRound when the
response stream completes, recording the assistant reply and invoking
the LLM memory extractor asynchronously.

backend/cmd/cyrene/main.go

@@ -388,10 +388,12 @@ func loadAdaptersFromConfig(router *bridge.Router, cfg *bridge.BridgeConfig, orc
 	}
 
 	router.SetMessageHandler(func(msg *bridge.UnifiedMessage) {
-		tokenCh, _, err := orch.ProcessInput(context.Background(), "bridge-"+msg.Source, msg.Content)
+		sessionID := "bridge-" + msg.Source
+		tokenCh, _, err := orch.ProcessInput(context.Background(), sessionID, msg.Content)
 		if err != nil { return }
 		var fc string
 		for t := range tokenCh { fc += t.Content }
+		orch.FinishRound(sessionID, msg.SenderID, fc)
 		if fc != "" {
 			router.SendTo(msg.Platform, msg.Source, fc, bridge.SendOptions{MaxLength: 500, MinGap: 3 * time.Second})
 		}