fix: wire agent group selection through to simulation start

Step 2 -> Step 3 -> Simulation API now passes selected_agent_ids.
Backend filters reddit_profiles.json and twitter_profiles.csv
to only include selected agents before starting simulation.

Flow: Step2 checkboxes -> emit('next-step', {selectedAgentIds}) ->
router query -> Step3 props -> startSimulation API -> filter profiles

backend/app/api/simulation.py

@@ -1448,6 +1448,41 @@ def generate_profiles():
 
 # ============== 模拟运行控制接口 ==============
 
+def _filter_simulation_agents(simulation_id: str, selected_agent_ids: list):
+    """
+    Filter simulation profiles to only include selected agents.
+    Modifies the profile files in-place before simulation starts.
+    """
+    import csv
+    sim_dir = os.path.join(Config.OASIS_SIMULATION_DATA_DIR, simulation_id)
+    
+    # Filter Reddit profiles (JSON)
+    reddit_path = os.path.join(sim_dir, 'reddit_profiles.json')
+    if os.path.exists(reddit_path):
+        with open(reddit_path, 'r', encoding='utf-8') as f:
+            profiles = json.load(f)
+        
+        if isinstance(profiles, list):
+            filtered = [p for i, p in enumerate(profiles) if i in selected_agent_ids]
+            with open(reddit_path, 'w', encoding='utf-8') as f:
+                json.dump(filtered, f, ensure_ascii=False, indent=2)
+            logger.info(f"Reddit profiles: {len(profiles)} -> {len(filtered)}")
+    
+    # Filter Twitter profiles (CSV)
+    twitter_path = os.path.join(sim_dir, 'twitter_profiles.csv')
+    if os.path.exists(twitter_path):
+        with open(twitter_path, 'r', encoding='utf-8') as f:
+            reader = csv.DictReader(f)
+            rows = list(reader)
+            fieldnames = reader.fieldnames
+        
+        filtered = [r for i, r in enumerate(rows) if i in selected_agent_ids]
+        with open(twitter_path, 'w', encoding='utf-8', newline='') as f:
+            writer = csv.DictWriter(f, fieldnames=fieldnames)
+            writer.writeheader()
+            writer.writerows(filtered)
+        logger.info(f"Twitter profiles: {len(rows)} -> {len(filtered)}")
+
 @simulation_bp.route('/start', methods=['POST'])
 def start_simulation():
     """
@@ -1503,6 +1538,7 @@ def start_simulation():
         max_rounds = data.get('max_rounds')  # 可选：最大模拟轮数
         enable_graph_memory_update = data.get('enable_graph_memory_update', False)  # 可选：是否启用图谱记忆更新
         force = data.get('force', False)  # 可选：强制重新开始
+        selected_agent_ids = data.get('selected_agent_ids')  # 可选：选中的Agent ID列表
 
         # 验证 max_rounds 参数
         if max_rounds is not None:
@@ -1600,6 +1636,11 @@ def start_simulation():
             
             logger.info(f"启用图谱记忆更新: simulation_id={simulation_id}, graph_id={graph_id}")
         
+        # Filter agents if selected_agent_ids provided
+        if selected_agent_ids:
+            logger.info(f"Filtering agents: selected {len(selected_agent_ids)} agents")
+            _filter_simulation_agents(simulation_id, selected_agent_ids)
+        
         # 启动模拟
         run_state = SimulationRunner.start_simulation(
             simulation_id=simulation_id,